The next phase of artificial intelligence (AI) adoption is shifting from model capability to execution, as enterprises struggle to turn powerful systems into usable, reliable applications.

This gap is becoming the industry’s defining challenge, particularly in complex environments where accuracy, trust, and workflow integration are critical. As companies move beyond experimentation, the emphasis is shifting toward embedding AI into core business processes rather than running isolated pilot projects.

“There’s actually a pretty big gap between the model capability and the user of the model. What people now call agents is essentially about building real use cases for end customers to truly unleash that capability,” said Emily Xue, Head of Enterprise AI at Scale AI.

“To close that gap, you need to understand what the model can do and how to steer it, but also deeply understand the customer. Connecting these two sides is not easy.”

The shift toward execution is forcing enterprises to rethink workflows, data integration, and how success is measured. Rather than focusing solely on model performance, organizations are prioritizing reliability, explainability, and operational fit.

This pressure is most visible in healthcare, where AI systems must operate within strict safety and operational constraints.

“We treat reliability and the quality of our deliveries very highly. We focus on high-stakes, high-impact domains where trust really matters,” Xue said. “We work with one of the largest medical centers in the United States, Mayo Clinic, and help reduce administrative workload and the cognitive load on physicians.”

“Doctors often need to review hundreds of pages of medical records before treating a patient. Our agent helps extract the relevant information and present it in a clear way, reducing a lot of review time.”

Beyond clinical workflows, deployments are also streamlining patient intake. Context-aware systems reduce repetitive questioning and improve patient experience. These early use cases are helping build trust in environments where errors carry significant consequences.

From data to deployment

Xue told TechJournal.uk in an interview during the AI and Business Innovation Summit, organized by Economist Impact, in London on March 25. The discussion focused on how enterprises can operationalize AI at scale and move beyond fragmented tool adoption.

Scale AI, based in San Francisco, was established in 2016 as a data annotation provider supporting early machine learning systems, particularly in computer vision and autonomous driving.

“Scale AI started with data labeling for very data-heavy machine learning models like vision systems. That was the foundation of our business,” she said.

The company later expanded into supporting large language models (LLMs) through reinforcement learning with human feedback (RLHF), a training method in which human evaluators guide model outputs to improve performance, reasoning, and safety alignment.

“What truly makes ChatGPT work is RLHF, where human feedback is used in post-training to improve model quality and safety alignment,” Xue said. “Scale AI was actually the first data partner to work with OpenAI, providing annotated human data to help large language models become usable.”

Scale AI has since evolved into a broader AI infrastructure provider. Its services now span data annotation, model evaluation, and enterprise software for building and deploying AI applications. Its Safety, Evaluation and Alignment Lab focuses on model alignment and benchmarking, while its platforms support both the private and public sectors and customers.

Beyond early language tasks, the company now supplies increasingly complex training data, including reasoning datasets used to improve model performance on multi-step problems. It also supports customers with deployment engineering, helping integrate models into enterprise systems, connect data pipelines, and meet security and compliance requirements.

Xue said Scale AI works with global customers, including large UK-based enterprises and organizations in the Middle East, reflecting demand for localized, domain-specific AI implementations.

The company supports the full lifecycle of AI adoption, including data preparation, model evaluation, system integration, and production deployment. Its role increasingly centers on helping enterprises operationalize AI systems within existing infrastructure and regulatory constraints.

It operates across multiple business lines, including enterprise AI deployment, government applications, and foundational model support. It also runs large-scale data operations through subsidiaries specializing in training data for computer vision and large language models.

The company’s founder, Alexandr Wang, is now the Chief AI Officer at Meta Platforms. He identified early the importance of high-quality training data and human feedback in building scalable AI systems.

This positioning has enabled Scale AI to act as a bridge between model developers and enterprise users. As organizations look to translate general-purpose models into domain-specific applications, demand for this intermediary role is increasing. The company’s expansion into Europe and the Middle East reflects that trend.

Governance challenges

As foundation models become more widely adopted, questions around governance, ethics, and ownership are becoming more prominent.

“Foundation models have, to some extent, become a public asset. It’s almost like electricity in how people rely on it,” Xue said.

“They need to bring public value and hold a high ethical standard. But what defines ‘good’ is still evolving, and there are large gray areas.”

“There are still areas we don’t fully understand, like user privacy and data ownership. These are questions that will need regulatory and industry alignment over time.”

The debate is being shaped by growing government involvement in AI development and deployment. This raises questions around access, control, and cross-border data flows. Enterprises operating globally must navigate emerging regulations while maintaining consistent standards for safety and compliance.

Execution reality

Despite concerns about automation, enterprise AI adoption is increasing demand for skilled workers rather than reducing it.

“AI is improving the velocity and quality of delivery for people already in the company. We actually have more demand than we can meet,” Xue said. “It’s not about reducing people. It’s about needing both people and AI to make teams more effective.”

“Code generation is only one part of the problem. You still need to understand use cases, design products, test systems, and monitor performance in production.”

Enterprise deployments require multidisciplinary teams that combine technical expertise with domain knowledge, particularly in regulated industries. This is driving demand for roles that sit at the intersection of engineering, operations, and business strategy.

Looking ahead, the company’s priority is to translate AI capabilities into measurable business outcomes.

“Our goal this year is to make AI actually produce business impact,” Xue said. “To do that, we need to build foundational technologies and work closely with high-stakes customers.”

She said success will depend on balancing domain expertise with execution discipline as enterprises move from experimentation to scaled deployment. Companies that integrate AI into core workflows, rather than treating it as a standalone tool, are likely to see the most significant gains.

Artificial intelligence (AI) is accelerating cyberattacks at a pace many enterprises are structurally unprepared to match, widening a critical gap between threat detection and response.

As attackers automate intrusion techniques and compress timelines from weeks to hours, chief information officers (CIOs) and chief information security officers (CISOs) face a growing execution problem. Organizations understand the risks but cannot act fast enough to mitigate them.

“The way that attackers are leveraging AI today is to take the drudgery out of attacking organizations. They automate all of the different steps within the attack path, and they can operate 24/7 at scale,” Gavin Millard, Vice President of Intelligence at Tenable, told TechJournal.uk in an interview.

“If you’ve got something measured in minutes from an attacker perspective and the mitigation is measured in weeks or months, we’re going to have a big problem,” he said.

This imbalance reflects what security leaders describe as “active inertia,” where enterprises recognize cyber risk but remain constrained by legacy workflows, manual prioritization, and fragmented visibility.

Attackers are not changing their playbook. They are industrializing it. AI is used to automate reconnaissance, identify weak points, and execute attacks continuously without human intervention.

“These AI-powered attacks aren’t novel and they aren’t indefensible. They use the same methodologies attackers have used for years, but they’re automated and amplified,” Millard said.

The shift has significantly reduced the window for defenders. The average time from vulnerability disclosure to exploitation has collapsed from months to days, and in some cases to hours, leaving little margin for traditional remediation cycles.

Millard attended the Cloud and Cybersecurity Expo, part of Tech Show London, on March 4–5.

Shifting priorities

Tenable, founded in 2002 and headquartered in Columbia, Maryland, is best known for creating the widely used Nessus vulnerability assessment technology. The company has expanded into exposure management, extending visibility beyond traditional scanning to include identity, cloud, external attack surfaces, and operational technology (OT), with a focus on prioritizing the risks most likely to be exploited.

“Traditional vulnerability management really just looks at Windows boxes and network devices, whereas exposure management includes identity, external attack surface, cloud infrastructure, and even OT devices,” Millard said.

The challenge is not only visibility but prioritization. Security teams face an overwhelming number of vulnerabilities, yet only a small fraction are actively exploited.

“About 300,000 vulnerabilities have been disclosed over the years, and around 56% are critical or high. The reality is attackers are only using a very small number of those, and we prioritize about 1% as likely to be targeted,” he said.

This approach enables organizations to focus resources on the exposures that matter most, rather than attempting to remediate everything.

It also highlights a broader operational issue. Many enterprises still rely on ticketing systems and manual workflows that cannot keep pace with machine-driven attacks.

Hygiene gaps

Many high-profile cyber incidents originate from basic operational failures rather than sophisticated techniques.

Millard pointed to phishing as a common entry point and said its effectiveness often reflects gaps in foundational controls.

“Phishing is often just the entry point. If your controls are working properly, users shouldn’t even see those emails in the first place,” he said.

He said effective defense depends on consistent execution of core practices, including patching, filtering, and identity protection.

“To defend against phishing, you have to have a multi-layered approach. You need to make sure your endpoints are up to date, have good email filtering, and use multi-factor authentication so credentials are harder to take advantage of,” he said.

In practice, these measures remain unevenly implemented, particularly in large enterprises with complex environments, creating persistent entry points for attackers.

Ransomware is often less a result of advanced tactics than accumulated weaknesses across these basic controls.

Expanding attack surface

The rapid adoption of AI tools is introducing new forms of exposure, often without sufficient governance or visibility.

“People don’t know where AI is. They don’t know the browser plugins using AI or where someone has installed tools like Claude to generate content,” Millard said.

This lack of oversight creates risks ranging from data leakage to unauthorized automation of sensitive processes.

Prompt injection has emerged as a key concern, enabling attackers to manipulate AI systems to reveal information or perform unintended actions.

“Prompt injection is manipulating the prompt to get the AI to do something it’s not designed to do or to reveal information it shouldn’t,” he said. The issue is compounded by the use of experimental tools inside corporate environments.

“Some people install tools like OpenClaw on corporate systems, and it’s really important organizations know where that is and who’s running it,” he added.

Such deployments can bypass established controls and reinforce the need for continuous discovery and governance of AI usage across the enterprise.

OpenClaw, formerly Moltbot and Clawdbot, is an open-source, self-hosted agentic AI platform that acts as a personal assistant executing tasks autonomously. Unlike traditional chatbots, it can manage emails, browse the web, edit files, and control services through interfaces like Slack and WhatsApp. Its rapid adoption has raised security concerns, with experts warning its broad permissions, limited safeguards, and unvetted plugins could expose sensitive corporate data and systems.

Beyond compliance

Regulatory frameworks such as the EU AI Act are beginning to address AI risks, but they remain misaligned with the pace of threat evolution.

“Compliance is not security. It’s slow and doesn’t adapt to technology very effectively,” Millard said.

He said organizations should treat compliance as a baseline rather than a primary defense mechanism.

“If organizations are waiting for compliance to improve their AI security, they’re going to have a bad time. They need to be ahead of it,” he said. “Cybersecurity is now one of the top business risks, and organizations are willing to invest because the impact of an incident can be business-ending.”

Machine-speed defense

Enterprises face increasing pressure to move from reactive security models to automated, intelligence-driven operations.

Millard said the core issue is execution, as many organizations still depend on manual processes that cannot keep pace with automated attacks.

“If you’ve got something that’s measured in minutes from an attacker perspective, and defenders are using human-driven processes like prioritization and ticketing, that gap is where the real risk sits,” he said.

He said organizations must rethink remediation, shifting from lengthy workflows to faster, automated responses driven by contextual data.

“We don’t want to give teams a 300-page report of issues. You want to give them three things they need to fix today and the actions to take to fix them,” he said.

This requires tighter integration between security and IT operations so teams can move from identifying risks to resolving them in near real time.

Rather than attempting to fix everything, enterprises are focusing on exposures most likely to be exploited, supported by automation that reduces response times from weeks to hours.

On March 24, Tenable unveiled Tenable Hexa AI, an agentic AI engine within its Tenable One Exposure Management Platform designed to automate security workflows and convert exposure intelligence into coordinated action to reduce cyber risk.

The launch reflects a broader shift in the threat landscape. AI-driven attacks and faster vulnerability discovery are expanding the attack surface faster than teams can respond, while fragmented tools and a mix of human and automated workflows make coordinated action harder and keep many organizations in reactive cycles.

Artificial intelligence (AI) is advancing at a pace few organizations can match, exposing a widening gap between what the technology can deliver and what businesses can extract from it. The constraint is no longer whether AI works, but whether companies can absorb, deploy, and scale it effectively across their operations.

That gap is becoming a defining challenge as enterprises invest heavily in AI but struggle to translate capability into measurable outcomes. The issue is a structural misalignment between fast-moving technology and slower organizational systems.

“We’re actually seeing the gap widen. We’re seeing the technological innovation and what the technology can actually do is accelerating, but the value that businesses are extracting is somewhat remaining flat. Our goal is working every single day with companies and customers to narrow that gap,” said Nicolai Skabo, EMEA enterprise lead at OpenAI.

“The ones who are truly approaching this from a true business transformation are the ones that are having success,” he said. “They’re rethinking this from the ground up. They’re thinking about their organizations at the capability layer, not just at the technology layer. They are building the skill set, the operating rhythm and the system. The technology can’t do that on its own.”

Enterprise drag

The discussion took place at the AI and Business Innovation Summit, organized by Economist Impact on March 25, in London. Skabo was speaking in a session, moderated by Dexter Thillien, lead analyst for technology and telecoms at the Economist Intelligence Unit, that focused on how organizations adapt to rapidly evolving AI capabilities.

Across financial services, manufacturing, life sciences, and telecommunications, companies are deploying AI at scale, but benefits remain uneven, reflecting a disconnect between technological progress and enterprise readiness.

This disconnect is rooted in what he described as enterprise inertia or drag, in which legacy processes, governance structures, and operating models slow adoption. Many organizations continue to layer AI onto existing workflows rather than redesigning them, thereby limiting the technology’s impact.

“It’s kind of that enterprise inertia or drag. The companies that have the most success are the ones willing to rethink how their business operates at a fundamental level. If you’re not testing and learning quickly, you’re very quickly behind,” he said.

The comparison, he noted, is similar to earlier industrial transformations. Just as factories had to be redesigned to take full advantage of electricity or steam power, businesses today must rethink how work is organized to fully benefit from AI.

Operating model shift

A key barrier lies in how organizations are structured for evaluating and deploying new technologies. Traditional processes for procurement, compliance, and risk management are often too slow to keep pace with AI’s rapid evolution.

“Most companies have not been structured or set up to handle the pace of innovation that has come from AI. You have to change your process for how you evaluate and adopt this. It goes around security, compliance and governance,” Skabo said.

Some organizations are beginning to address this by creating parallel pathways for AI adoption. These accelerated tracks sit outside traditional processes, allowing companies to experiment and deploy new capabilities more quickly while longer-term operating models are updated.

This dual approach enables firms to capture near-term value while modernizing legacy systems that are not fit for AI’s pace.

Leadership plays a critical role in overcoming these structural barriers, but not in isolation. Rather than a generic top-down mandate, Skabo framed executive involvement as a necessary condition for signaling strategic priority and mobilizing resources across the organization.

“It has to be CEO led. It has to be board-led. They need to show up and put their money where their mouth is if this is a true core part of the strategy. If CEOs and boards are treating this as optional, so will the rest of the organization,” he said.

Effective adoption also depends on distributing capability across the organization, with business units identifying practical use cases and refining workflows.

Companies that share knowledge across teams scale successful initiatives more quickly and avoid duplication, combining centralized direction with distributed execution.

Proof of value

Early adopters are beginning to demonstrate tangible gains from AI, particularly in areas such as software development, process automation, and customer-facing operations.

One example cited during the discussion was NatWest, where AI has been integrated across multiple business lines. Nearly a third of the bank’s code was AI-assisted or AI-generated in the past year, contributing to significant productivity improvements.

The bank also reported saving approximately 70,000 hours by automating business processes, reducing operational costs and improving efficiency. In its wealth management division, time spent on administrative tasks has been cut by 30% to 40%, allowing staff to focus more on client engagement.

These results show AI can deliver measurable returns at scale, but only when aligned with broader organizational change.

Beyond individual use cases, organizations are also confronting a more immediate challenge: defining meaningful milestones for AI adoption.

“So I think firstly, it’s about how the organization is adopting and leveraging AI day to day, and in an integrated environment? When we then look at milestones, I don’t have a single metric. These businesses are so different, they have different goals,” Skabo said.

“Whether it is increasing customer satisfaction, doing more with less, or AI coding, the measures of success will be different by company.”

Rather than relying on universal benchmarks, companies are tracking how deeply AI is embedded in day-to-day operations and whether usage has moved beyond pilots into organization-wide workflows.

AI adoption is not a one-off deployment but an ongoing process. What was advanced six months ago may already be outdated, forcing continuous reassessment.

Skabo framed this as a forward-looking readiness challenge for enterprises.

He said a key question for organizations is whether they are ready to adopt the technology coming in six months, including whether they have the team, structure, processes, and clearly defined business problems in place.

“If we’re ready for that, then we’re in a good position to take advantage of this in a great way looking forward,” he said.

Readiness gap

Looking ahead, enterprises must not only implement current AI capabilities but also prepare for what comes next, as today’s tools may be superseded within months.

Skabo said this raises a fundamental question: whether organizations can absorb the next generation of AI tools. Readiness depends on skills, processes, and strategic clarity, not just infrastructure.

He said companies that continuously test and adapt are more likely to close the gap between capability and value, while others risk falling further behind. In practical terms, organizations must speed up evaluation and deployment. Traditional multi-year planning cycles are incompatible with AI’s pace, pushing firms toward shorter cycles and faster deployment.

He also said clearly defined business problems are critical, enabling organizations to adopt new capabilities proactively rather than reactively. These changes reflect a broader shift in how organizations learn and adapt, with continuous testing enabling firms to keep pace with change.

This reinforces Skabo’s central point: the challenge is not access to AI, but the ability to evolve alongside it. In this environment, the competitive advantage lies not in access to technology, but in the ability to transform alongside it.

Artificial intelligence (AI) is no longer simply enhancing productivity. It is redefining how businesses operate at their core. As companies move from deploying tools to orchestrating autonomous systems, executives are being forced to rethink workflows and the fundamental structure of their organizations.

The shift marks a transition from incremental efficiency gains to structural transformation across the enterprise. Early adoption has focused on automating existing processes. Leading firms are now redesigning operations—aligning decision-making, incentives, and metrics with increasingly autonomous AI systems.

“The unit of production within a business fundamentally changes as a result of AI. The bottleneck moves from being the technology to being your organization and the way that you do things,” said Kareem Ayoub, Vice President of AI technical strategy at Google DeepMind.

Many companies remain in an early phase of adoption, applying AI to existing structures without rethinking how work is organized. They continue to pursue existing goals using unchanged processes, thereby limiting the technology's broader impact.

“Leaders are deploying AI in the same organization with the same people, for the same goals, with the same processes. That is building a faster horse. There is extraordinary value in that, but the fundamental insight is that the unit of production changes,” he said.

The implications are already visible in early case studies across industries. Ayoub pointed to a law firm that shifted its focus from billable hours to customer satisfaction, resulting in sustained year-on-year revenue growth of 30% to 40%. The example underscores how AI can reshape not only execution, but also what businesses choose to optimize.

Agentic shift

The discussion took place at the AI and Business Innovation Summit in London on March 25, organized by Economist Impact. The session was moderated by Jonathan Birdwell, Global Head of the Policy and Insights Team at Economist Impact, who framed the debate around the transition of AI from assistive tools to autonomous systems.

Ayoub outlined a near-term evolution toward more advanced “agentic” capabilities. These systems operate independently, collaborate with one another, and increasingly execute tasks without continuous human oversight. He said these developments will accelerate over the next one to two years and fundamentally change how work is executed.

“You will start to have teams of agents that are able to collaborate with each other, decomposing complex problems into different parts,” Ayoub said. “The second capability is headless agents, where you can send work off and check back later, and the third is self-learning agents.”

He said combining these capabilities can deliver strong outcomes.

Ayoub also drew a distinction between autonomy and what he called asynchronous execution. In practice, he said, companies should think less about unsupervised systems and more about how long an agent can run before a human needs to check in. That depends on how verifiable the goal is.

If an agent is handling a software task, for example, it can write tests and work toward completion with relatively limited intervention. If it is producing a marketing campaign, executives are likely to review the work more frequently because tone, brand, and judgment are harder to verify.

Ayoub suggested that this gives companies a practical way to evaluate deployment: reduce the number of interventions needed to move from response to goal.

He cited a recent example in which a researcher deployed 16 agents running continuously for two weeks to rebuild a complex compiler. The work would typically require a full engineering team. The project cost roughly $20,000 in compute resources, illustrating the rapid compression of time and cost enabled by such systems.

Jagged intelligence

Despite these advances, AI performance remains uneven. Ayoub described this as “jagged intelligence,” where systems can outperform humans in complex domains while failing at simpler tasks. He pointed to a coding competition in which more than 150 human teams failed to solve a problem that an AI system completed in under 30 minutes by approaching it from a different strategic angle.

“AI today is really good in some capabilities and falls short in others that seem simpler,” he said. “That jagged intelligence is the central design challenge for every leader.”

He said the same model that invents mathematical theorems may miscount the number of chairs in a room, illustrating the gap in performance. He added that leading companies design around areas where AI performs best.

This approach requires a shift in mindset—from expecting uniform performance to strategically deploying AI where it delivers the most value.

A recurring theme was that technology is no longer the primary constraint.

Organizational design, data infrastructure, and leadership behavior have become the key limiting factors. Companies must ensure their systems are “agent-ready” and avoid building solutions around temporary model limitations, as rapid capability gains can quickly render such investments obsolete.

“The best companies focus on getting their infrastructure and data agent-ready,” Ayoub said. “When models hit silos inside your company, you lose the value of your investment.”

He warned that rapid improvements in model capability can quickly render some investments obsolete.

“The best companies measure outcomes, not activity,” he said. “They invest heavily in identifying the right metric to improve.”

“Raise your hand if you use AI regularly. Now keep your hand raised if you can tell me exactly how that value shows up on your P&L. That gap is probably core to the issue,” he added.

Leadership gap

The most direct challenge was aimed at senior leadership. Executives who do not actively use AI themselves risk undermining their organizations’ ability to adopt it effectively, limiting their ability to evaluate initiatives or guide implementation.

“Leaders not using AI is probably problem number one, two and three within companies today,” Ayoub said.

“If you are managing a portfolio of experiments, it becomes very challenging to assess it if you have not run any yourself,” he said. “It is like steering a ship without a compass.”

He said the best companies have leaders using the technology directly so they understand where it works and where it fails.

Ayoub said that leaders who do not use AI themselves often struggle to accurately judge frontier capabilities. Some remain constrained by legacy systems, while others assess advanced models through free consumer tiers or ask them to perform tasks that were already possible several years ago.

He also pointed to an authenticity gap. When senior leaders visibly use AI in their own work, it signals that experimentation is acceptable across the organization. He said that dynamic played out within his own team: after colleagues noticed that he had been using Gemini heavily, their own use of AI tools rose sharply across engineering, product, assistant and strategy roles.

Looking ahead, AI is expected to become embedded in daily work at every level of the organization. Rather than treating it as a specialized capability, companies will need to integrate it into core operations and decision-making processes.

Ayoub’s closing advice reflected this urgency. Leaders, he said, should begin by applying AI directly to their own work—not as an experiment, but as a routine part of how they operate.

The next phase of AI adoption will not be defined by technological breakthroughs alone, but by how effectively organizations adapt their structures, incentives, and leadership practices to harness its full potential.

Enterprise software is entering a new phase as artificial intelligence (AI) moves beyond rule-based systems into agent-driven architectures that can reason across data, tools, and workflows. The shift replaces rigid, deterministic logic with probabilistic systems that demand new design and operating models.

Instead of relying on predefined rules and APIs, enterprises are adopting natural-language-powered orchestration layers. These systems allow applications to interpret context and act across multiple data sources in real time. As adoption accelerates, AI is moving into core business processes—while exposing structural challenges around data fragmentation, governance and scalability.

“Building applications has changed quite a lot in the last five years. You went from very stiff if-then logic, a lot of APIs and code rules, and now that we’ve entered the age of agentic applications, it’s completely different,” said Grace Adamson, AI Senior Product Marketing Manager at Snowflake.

“All of a sudden, you have this orchestration layer that can use natural language, isn’t restricted to if-then logic, but can still connect and orchestrate across all your data,” she said.

The shift is forcing enterprises to confront a fundamental constraint: AI systems are only as effective as the data they can access and interpret. As organizations scale AI across departments, fragmented datasets and inconsistent context are emerging as key barriers to reliable outcomes.

“Your agent is only really as good as the data it has access to,” Adamson said. “Every organization is trying to scale AI through all the enterprise applications. This means a lot of disconnected data points and disconnected context.”

At the same time, the rise of agentic systems is reshaping how enterprises think about security and control. Traditional governance models, designed around human users, are being reconfigured to accommodate autonomous agents acting on behalf of users and systems.

“All of this still needs to be in the same security perimeter, and all of that is compounded even more complicated if you think about setting up policies that used to be by person, and now all of that is by agent object instead,” Adamson said.

The shift toward agent-level governance introduces new operational complexity, particularly in regulated sectors where traceability, auditability, and access control are critical. Enterprises must now track not only who accessed data but also which agent performed actions, what context was used, and how decisions were made.

Inside the build

Snowflake is a cloud-native data platform that enables enterprises to consolidate, process, and analyze data across multiple environments, while increasingly embedding artificial intelligence and machine learning capabilities directly into its architecture.

This positioning reflects a broader industry push to operationalize AI within core data infrastructure rather than treating it as a separate layer.

This shift—from rule-based systems to agent-driven orchestration, alongside growing concerns around data fragmentation and governance—was a central focus at Snowflake BUILD: The Dev Conference for AI & Apps, held in London on February 5, where executives outlined how agentic architectures are moving from experimentation into production environments.

Adamson focused on the conceptual framework behind agentic applications, while Teresa Nascimento, Senior Solution Engineer at Snowflake, demonstrated how such systems are implemented and deployed in enterprise settings.

Nascimento outlined a three-layer architecture:

• Data layer integrating structured and unstructured sources

• Agent layer responsible for reasoning and orchestration

• Application layer that interfaces with users

At the core is the ability to combine context with action. In one example, a dispute-resolution agent aggregated customer data, transaction records, and unstructured documents to generate recommendations and risk assessments.

“You add those instructions here to make your agent less probabilistic and more deterministic when it’s posed with a question,” she said.

Tools and control

A key focus of the discussion was the role of tools in making AI systems reliable enough for enterprise use. While agents are inherently probabilistic, integrating predefined functions and domain-specific logic can introduce deterministic behavior.

“When you start to create these customized tools, you’re turning a probabilistic workflow into something that has more deterministic features and functions,” Adamson said.

These tools range from simple user-defined functions to complex machine learning models deployed through container environments. By embedding these capabilities into agents, organizations can constrain outputs, enforce business logic, and improve consistency.

“What we recommend at Snowflake is having an agent that’s incredibly specific, so specific agents for specific tasks,” she said. “This is where we’re seeing a lot of improvements in quality.”

Nascimento showed how developers can connect agents to applications via APIs, passing contextual data and instructions to control how responses are generated and formatted. This allows enterprises to tailor outputs to specific use cases, such as compliance workflows or financial analysis.

“You can add as many tools as you want to your agent,” she said.

The increasing use of modular tools is also driving a shift away from general-purpose AI toward specialized, domain-focused agents designed for specific tasks.

Scaling challenge

Despite rapid progress in development tools, the transition from prototype to production remains a hurdle. Building an agentic application can be straightforward, but scaling it across thousands of users introduces new operational demands.

“It is easy to build an agent. It is easy to build an application, but then scaling that application takes a lot of work,” Adamson said.

To address this, vendors are introducing capabilities such as version control, monitoring, and evaluation metrics that track how agents perform over time. These features are critical for ensuring reliability, especially as agents interact with live data and business processes.

“You’re now able to evaluate that across not just in testing… but after deployment as well,” she said.

Looking ahead, the enterprise focus is shifting toward making agentic systems production-ready, balancing flexibility with control, and automation with accountability. As organizations deepen their use of AI, the ability to orchestrate data, tools, and governance at scale is likely to become a defining factor in competitive advantage.

Real-time machine learning is moving from niche experimentation into a core enterprise capability, as organizations seek to make decisions instantly rather than in batch cycles. The shift is being driven by use cases such as fraud detection and personalization, where milliseconds can determine outcomes.

At the same time, enterprises are rethinking how machine learning systems are built and deployed. They are moving away from fragmented toolchains toward unified platforms that integrate data, models, and governance. This convergence is reshaping how artificial intelligence (AI) is operationalized.

“Sometimes you might have a use case that is real-time in nature. What comes to mind is fraud detection or personalization, where you want to meet the customer at the point of need,” said Larry Orimoloye, Principal AI/ML Field CTO at Snowflake. “What this allows you to do is to serve your model at low latency to 10 milliseconds, so you can actually service that model and give decisions in real time.”

The technical barrier to achieving that level of responsiveness has historically been the gap between model development and production deployment. Systems that perform well in testing environments often struggle to deliver consistent, low-latency performance in live settings.

“I could build the best model, train it and get the best accuracies, but when it comes to bringing it into production and hosting it for inferences, that always takes weeks,” said Harley Chen, Solutions Engineer at Snowflake. “Now with this approach, it is really one function, and you can deploy directly into production.”

Real-time shift

At Snowflake BUILD: The Dev Conference for AI & Apps in London on February 5, Orimoloye and Chen outlined how enterprises are moving toward real-time machine learning systems and integrated development environments. The session focused on building, deploying, and scaling online models using a unified architecture.

Snowflake provides a cloud-native data platform that integrates storage, processing, and analytics. Its AI and machine learning capabilities allow organizations to build and deploy models directly within the data environment, without moving data across systems.

The emergence of real-time inference reflects a broader transition in enterprise AI architecture. Models are increasingly expected to operate continuously and respond dynamically to live data streams. This is particularly evident in financial services, e-commerce, and cybersecurity, where delayed insights can translate into missed opportunities or increased risk.

Latency has become a defining metric in this transition. Systems must process large volumes of data while delivering predictions within tight time constraints.

“With our printout of 20 transactions, the overall latency is about 500 milliseconds, including network and authentication, but the per-record latency is under 30 milliseconds,” Chen said. “That is what enables these online use cases to actually work in practice.”

The ability to sustain such performance under load is equally important. Real-world systems must handle bursts of activity without degrading response times.

“I can inject a burst of transactions and run that against the endpoint, and the system will return results almost instantly,” Chen said. “The average inference time is well under 50 milliseconds, even when running multiple requests.”

These capabilities are supported by advances in infrastructure, including distributed computing, containerized workloads, and tighter integration between data storage and processing layers.

Platform convergence

Alongside the push for real-time performance, enterprises are consolidating machine learning workflows into unified environments. This trend addresses longstanding challenges around data movement, security fragmentation, and operational complexity.

“A lot of customers ask why they should build machine learning pipelines directly on a single platform,” Chen said. “You don’t need to move your data back and forth or manage multiple security environments. Your data, your model, your features and your users are all governed under a single role-based access control model.”

The cost and risk associated with moving data between systems have become increasingly difficult to justify, particularly as regulatory requirements tighten and data volumes grow.

“Gone are the days when you need to manage egress or ingress costs or move data across different systems,” Orimoloye said. “Now you have everything within a governed perimeter, which simplifies both security and operations.”

This consolidation enables greater consistency in how models are developed and reused across organizations. Instead of isolated workflows, teams can build shared assets that scale across departments.

“When you build a model, you want to log it once and reuse it across the organization,” Orimoloye said. “That consistency is important when multiple teams are working on the same data and models.”

From experiment to operations

The shift toward unified platforms is also addressing a persistent bottleneck in machine learning: the transition from experimentation to production. Historically, this process required rebuilding models, reconfiguring environments, and coordinating across multiple teams.

“Moving machine learning from experimentation to production should not require rebuilding everything from scratch,” Orimoloye said. “The goal is to go from development to serving live traffic quickly and reliably.”

Integrated environments embed key machine learning operations (MLOps) capabilities directly into the development workflow. These include model tracking, feature management, and observability.

“You can have full transparency around what dataset was used to train the model, and as you iterate, you can see how the model evolves over time,” Chen said. “That visibility is critical when you are deploying models at scale.”

Observability has become a central requirement as models are deployed in dynamic environments. Teams need to understand how models behave over time to maintain accuracy and trust.

“One of the things you can do is look at drift and understand which features are influencing your model,” Orimoloye said. “You can drill down into specific segments to see how performance changes across different parts of your data.”

The ability to trace model lineage and data transformations is also gaining importance, especially in regulated industries.

“You can see not only the versions of the model, but also the metrics and the lineage of those models,” Chen said. “That helps teams understand how models evolve and what data was used.”

Toward agentic workflows

Another emerging trend is the use of AI systems to automate the development of machine learning pipelines. These agentic approaches are beginning to reshape how engineers interact with data and models.

“This is very powerful and a game changer,” Orimoloye said. “You can prompt it to build an end-to-end pipeline, use it to debug your code, or generate workflows using natural language.”

Rather than manually stitching together workflows, developers can rely on systems that decompose tasks, plan execution steps, and generate code.

“It starts to break the overall task down into different steps and plans how to approach each step,” Chen said. “Then it builds the assets step by step, almost like a structured to-do list.”

These capabilities suggest a future where machine learning development becomes more accessible and less dependent on specialized expertise.

Scaling enterprise AI

The convergence of real-time inference, unified platforms, and automated workflows is accelerating enterprise adoption of AI. Organizations are embedding machine learning into core business processes rather than treating it as isolated experimentation.

“One of the recurring themes we hear from customers is that the platform is scalable,” Orimoloye said. “You have the power of CPUs and GPUs, flexibility, cost efficiency, and a highly trusted environment that can support production workloads.”

As enterprises scale AI deployments, the focus is shifting from individual models to systems that operate reliably across the organization. Delivering fast, governed, and reusable machine learning capabilities is becoming a defining factor in competitive advantage.

The transition is still underway, but the direction is clear. AI is moving toward real-time decision-making, supported by integrated architectures that reduce complexity and enable scale.

Enterprise adoption of artificial intelligence (AI) is accelerating, but executives and engineers are increasingly confronting a harder reality: scaling AI safely requires far more than models and compute. As organizations move from experimentation to production, governance, data integrity, and operational resilience are emerging as critical risks.

The shift toward “agentic” AI systems — where software agents autonomously query data, trigger workflows, and make decisions — is amplifying those concerns. Without strong controls, fragmented data environments and weak permissions can undermine accuracy, compliance, and trust in AI-driven outcomes.

“There is no AI strategy without a data strategy. If your data is fragmented and you don’t have clear governance policies and permissions, AI is going to have a hard time delivering value the way you expect,” said Christian Kleinerman, EVP, Product Management, Snowflake.

“If your system stops working when a cloud provider fails, it’s not a reliable architecture,” he said.

He said enterprises should design systems that can withstand failures across infrastructure layers, rather than relying on any single provider or environment.

Snowflake is a fully managed, cloud-native data platform that enables organizations to consolidate data from multiple sources into a centralized environment for analytics, machine learning, and AI application development. Its platform handles structured and unstructured data while embedding governance and security controls across the data lifecycle.

Agentic shift

The growing interest in AI agents reflects a broader transition from analytics to action. Instead of generating insights for human interpretation, enterprise systems are increasingly expected to execute decisions directly — from pricing adjustments to customer engagement workflows.

“The goal is to make it easy to build not just agents, but enterprise-grade agents. These need to orchestrate structured data, unstructured search, and third-party tools, while staying within governance and security boundaries,” Kleinerman said.

That requirement is pushing organizations to rethink how data, models, and applications interact. Rather than exporting data to external tools, many enterprises are attempting to keep computation within controlled environments to reduce exposure and maintain auditability.

“A lot of what we do is to bring computation into the security boundary, so your data doesn’t have to be copied out and you don’t have to reinvent permissions and controls,” Kleinerman said.

The risks are not theoretical. As AI systems gain autonomy, errors can propagate faster and on a greater scale. Without clear oversight, enterprises face potential failures in compliance, data accuracy, and operational continuity.

Event context

These issues were a central focus at the Snowflake BUILD: The Dev Conference for AI & Apps, held in London on Feb. 5, 2026. Product leaders and enterprise users outlined how AI deployment is reshaping data infrastructure priorities.

Speakers emphasized that fragmented data environments remain one of the biggest barriers to enterprise AI success. Organizations often struggle to unify structured and unstructured data, enforce consistent permissions, and maintain visibility across systems.

Kleinerman said many companies are already encountering these constraints. AI initiatives frequently stall when data is scattered across platforms or lacks proper governance, limiting the ability to generate reliable insights or automate decisions.

At the same time, developers are under pressure to accelerate delivery. New tooling and AI-assisted workflows are helping reduce development time, but they also introduce new dependencies that must be managed carefully to avoid operational risks.

Effie Goenawan, Principal Product Manager at Snowflake, said enterprise AI adoption is also shifting toward broader business use.

“Business users don’t need to learn SQL or write code. They can just ask questions, but the real challenge is ensuring those answers are grounded in trusted company data,” she said.

“This is a great time to be a builder. Whether you’re working in a graphical user interface (GUI) or a command-line interface (CLI), you now have options to build, deploy pipelines, and create applications much faster,” said Dash Desai, Principal Developer Advocate at Snowflake.

Desai demonstrated how developers can move between integrated development environments (IDEs), command-line tools, and browser-based environments while working on the same data workflows. This flexibility reduces friction in development, but it also requires teams to maintain consistent governance and oversight across multiple toolchains.

Data challenges

“We wanted to build an agent that is grounded and auditable, and that stays within the security and governance of our central data platform. Accuracy and control were critical for us,” said Kiran Kodandoor Nayak, Principal Software Engineer at Booking.com.

“Data teams were constantly getting questions from business users. These were time-consuming and business-critical, and it prevented them from focusing on more strategic work,” he said.

The company introduced AI-driven data agents only after establishing strict controls around how those systems access and interpret data. Continuous auditing and iterative refinement were required to ensure reliability.

“We started small, and as business users tested it, we continuously audited responses to ensure accuracy. That helped us refine our semantic models and improve trust,” Nayak said.

Similar challenges were described by Neha Patel, Lead Machine Learning Engineer at IAG Loyalty, which operates the Avios rewards program.

“Machine learning is used across the entire member lifecycle. We generate huge volumes of data, and our challenge is turning that into personalized actions at scale,” Patel said.

Her team centralized data and machine learning workloads to improve consistency and reduce operational complexity. Bringing models closer to the data allowed them to streamline pipelines and maintain tighter control over outputs.

Operational pressures

As AI systems move into real-time decision-making, operational requirements are becoming more demanding. Applications such as fraud detection, dynamic pricing, and recommendation engines require low-latency responses and continuous monitoring.

“These use cases — pricing, fraud detection, recommendations — require ultra-low latency serving of predictions. That brings a whole new set of use cases within reach,” said Annissa Alusi, Senior Product Manager at Snowflake.

Meeting those demands requires not only faster infrastructure but also stronger governance. Real-time systems must ensure that data is accurate, permissions are enforced, and decisions can be traced and explained.

At the same time, enterprises are seeking to reduce the burden on developers. AI-assisted tools are increasingly used to automate tasks such as pipeline creation, data transformation, and model deployment. While these tools improve productivity, they also require careful oversight to prevent errors from being introduced at scale.

“A lot of what we do is to simplify the experience and remove the heavy lifting, so you can focus on how to get value out of your data instead of managing infrastructure,” Kleinerman said.

Control vs speed

The broader challenge for enterprises is balancing speed with control. AI adoption is often driven by the need to improve efficiency and competitiveness, but rapid deployment can expose weaknesses in data governance and system design.

Organizations are therefore taking a more cautious approach, prioritizing auditability, security, and reliability alongside performance. This includes:

• Centralizing data to reduce fragmentation and improve consistency
• Embedding governance and permissions into every layer of the stack
• Continuously monitoring AI outputs to detect errors and bias
• Designing systems that can withstand failures without disrupting operations

The emergence of agentic AI is likely to intensify these requirements. As systems become more autonomous, enterprises will need to ensure that decisions remain transparent and aligned with business rules.

In practice, this means building AI systems that are not only powerful but also predictable and controllable. The ability to trace decisions, enforce policies, and recover from failures will be critical as AI moves deeper into core business processes.

For many organizations, the path forward will involve incremental deployment — starting with tightly scoped use cases, validating outcomes, and gradually expanding capabilities. Building trust in AI systems is as important as building the systems themselves.

The transition to enterprise AI is no longer a question of capability, but of governance. Those who can manage the risks effectively are likely to gain a significant advantage as AI becomes embedded across the enterprise.

A European push to build secure military satellite communications (satcom) is accelerating, with a new standards-led waveform program designed to work across allied networks rather than fragment them.

The European Protected Waveform (EPW) program, launched in 2023, is moving into a new phase, marking a key step in Europe’s push for secure, interoperable military satellite communications.

Developed under the European Defence Fund with an estimated cost of €29.9 million, the program focuses on creating waveform standards that can operate across geostationary Earth orbit (GEO) and non-geostationary orbit (NGSO) systems while supporting both sovereign European missions and coalition operations with NATO partners.

“One of the examples that we are thinking about is the European Protected Waveform. It’s a prime example of when you think about open architectures, new waveforms and new capabilities, combining all of these elements into one end-to-end system,” said Koen Willems, Vice President, EU Programs & Government Relations at ST Engineering iDirect Europe.

“The key element is open architecture. That is where you define the standards, and then you can put sovereign capabilities on that architecture—whether it’s UK, European, or NATO,” Willems said.

Last November, the EPW initiative completed over-the-air testing in Germany, validating its ability to operate across both geostationary and low Earth orbit (LEO) satellite systems while maintaining resilience against jamming, cyber threats, and unauthorized access.

Toward 2030 capability

The EPW program is aligned with Europe’s broader satellite communications roadmap, including the Infrastructure for Resilience, Interconnectivity and Security by Satellite (IRIS²) constellation, with initial capability targeted around 2028 and full capability by 2030.

IRIS² is the European Union’s multi-orbit satellite constellation designed to provide secure communications for governments and defense while also delivering commercial connectivity services.

Willems said the initiative involves a consortium of companies and nations working to deliver a fully operational capability by the end of the decade.

The program is designed to support both sovereign European operations and coalition missions, reflecting the dual priorities of strategic autonomy and interoperability.

He said the primary driver behind this shift is the reality of joint operations.

“The driver in Europe was joint communications; it’s rare that one country goes into operation. You have joint training, joint operations, so you need interoperability, and that is something you get through standards,” he said.

Rather than forcing countries to adopt identical systems, open architectures allow national capabilities to coexist within a shared framework. This approach enables sovereign control while maintaining compatibility across alliances.

He said the overlap between European Union and NATO membership creates a natural pathway for alignment, even as funding and governance structures remain distinct.

“We know that the majority of our nations in Europe are also NATO nations; that is where the conversation can happen,” Willems said.

Software-defined shift

At the Defence In Space Conference (DISC) 2025 in London, Willems outlined how European defense communications are shifting away from platform-specific systems toward standards-based architectures.

He said technical transformation is equally critical to enabling interoperability at scale.

“You needed to deploy several modem systems from different vendors, one was connecting to GEO capability, another to partner systems, and another to multi-orbit, and that is a logistical nightmare,” he said.

Historically, military satellite communications relied on multiple hardware systems to connect to different networks, creating operational complexity in deployed environments.

That model is now being replaced by software-defined systems that support multiple waveforms and satellite constellations on a single platform. Software-defined capabilities also enable multi-purpose deployment across aerial, naval and land platforms, reducing the logistical burden while improving flexibility.

“That logistical nightmare is something that we can overcome today. The magic word here is software-defined,” Willems said.

These systems allow operators to adapt communications dynamically, switching between networks and configurations without requiring additional hardware deployments.

Multi-orbit resilience

A central feature of the emerging architecture is multi-orbit connectivity, which integrates GEO, medium Earth orbit (MEO) and LEO satellites into a unified network.

This capability is increasingly important in contested environments where communications infrastructure may be disrupted or targeted.

“Multi-orbit is there to expand your operations. It makes sure that you are resilient, so you can switch between satellites when you are jammed or when there is interference,” Willems said.

Multi-orbit systems enable seamless switching between orbital layers and provide redundancy and continuity, ensuring that critical communications can be maintained even under electronic warfare conditions.

He said this approach also supports broader operational agility, allowing networks to scale and adapt to different mission requirements.

Cloud and automation

ST Engineering iDirect Europe, based in Sint-Niklaas, Belgium, serves as the company’s EU Satcom Centre of Excellence, focusing on ground-segment technologies tailored to European defense requirements. The unit operates under EU and Belgian regulatory frameworks and has more than three decades of experience in satellite communications systems, supporting both commercial and government programs.

Willems pointed to virtualization and automation as key enablers of next-generation defense communications. Traditional systems rely on dedicated hardware for processing and network management, limiting flexibility and scalability.

“Virtualization means putting processing not on a one-to-one device, but on servers or in a private cloud,” he said.

This shift allows capabilities such as network management, signal processing, and modulation to be centralized or distributed as needed, improving efficiency and responsiveness. Automation further enhances these systems by enabling orchestration across multiple assets and networks.

“Automation is about orchestrating capabilities across different satellite assets and networks, and planning globally,” Willems said.

Together, these technologies form the backbone of a more agile, software-driven communications architecture that can evolve alongside operational requirements.

Transition and integration

The transition toward future architectures requires careful planning of legacy systems, particularly as defense operators migrate from existing platforms to next-generation capabilities.

“You also need to think about the transition towards the newer platform. You need to make sure that technology can migrate together with your entire network toward future capabilities,” Willems said.

This migration challenge is particularly relevant for military users who rely on long-lifecycle systems, in which incremental upgrades must be integrated without disrupting ongoing operations.

Open architecture provides a pathway for this evolution by allowing new capabilities to be introduced alongside existing infrastructure.

“We’re not only thinking about modulation standards, but also about integration. We need to make sure that you can piece all these elements together into one architecture,” he said.

This broader approach enables defense organizations to combine mobility, automation, and network management into a unified system, reducing operational complexity while improving user experience.

These developments are aligned with Europe’s long-term satellite strategy and future 5G-based architectures.

“We’re discussing not only DVB (Digital Video Broadcasting), but also 5G and next-generation 5G. We need to make sure that these future releases are fully secure not only for commercial operations, but also for military operations,” Willems said.

The combination of open architecture, software-defined systems, and multi-orbit connectivity will define how military communications networks are designed and deployed. The ability to integrate sovereign and allied capabilities within a single framework will be critical as defense operations become more interconnected and technologically complex.

Eutelsat Network Solutions, a Paris-based global satellite operator, is advancing a multi-orbit strategy that integrates geostationary Earth orbit (GEO) and low Earth orbit (LEO) systems, as defense and government users shift toward more resilient, layered communications architectures in contested environments.

The move reflects a broader industry transition away from reliance on single-layer satellite networks toward integrated systems that combine multiple orbital and terrestrial assets to maintain connectivity under disruption.

For defense users, the shift is not only technical but operational. Modern missions increasingly depend on continuous data flows across dispersed assets, from command centers to forward-deployed units and autonomous systems. That requirement is forcing communications providers to deliver networks that can maintain performance even when individual nodes or links are degraded.

Ian Canning, president and chief executive officer of Eutelsat, said this evolution is changing how customers define reliability, moving beyond traditional uptime metrics toward mission assurance, where connectivity must be sustained under dynamic and contested conditions.

“Eutelsat is the only satellite operator in the world that is delivering both GEO and LEO services, and those networks can be fused together, not just with our networks but with competitor networks as well, to provide resilient communications,” Canning said.

The company’s approach combines GEO, LEO and terrestrial networks into a unified architecture that can dynamically route traffic across multiple pathways, reducing vulnerability to jamming, interference and infrastructure failure.

Multi-network resilience is increasingly structured around primary, alternative, contingency and emergency (PACE) frameworks, which require communications systems to operate seamlessly across different layers.

“Providing global, secure communications across multiple platforms that can support a real PACE plan is absolutely critical for customers operating in complex and contested environments,” Canning said.

He said no single network layer can meet future operational requirements, because modern missions demand redundancy, flexibility and the ability to adapt rapidly to changing conditions.

That has led to growing interest in orchestration platforms that can manage multiple networks simultaneously, automatically selecting the most appropriate pathway based on performance, availability and mission priority. Such capabilities are expected to become increasingly important as network environments grow more complex.

Scaling for unmanned operations

A key driver behind the shift is the rapid growth of unmanned and autonomous systems, which are significantly increasing demand for satellite connectivity.

“The man-to-unmanned ratio is shifting very quickly, and that is going to require scalable, resilient communications that can support a much larger number of endpoints than traditional systems were designed for,” Canning said.

He added that communications demand is no longer driven primarily by human users, but increasingly by distributed platforms such as drones and remote sensing systems that require persistent, high-capacity connectivity.

This evolution is placing new pressure on network design, with greater emphasis on scalability, bandwidth management and prioritisation of traffic across different mission types.

Canning said operators must also consider how to efficiently allocate limited spectrum and capacity as the number of connected devices increases, particularly in scenarios where large volumes of data must be transmitted in real time.

The ability to manage these demands while maintaining secure and reliable connectivity is becoming a key differentiator for satellite operators.

Timing, interference and signal integrity

Alongside scaling challenges, satellite operators are addressing growing technical demands around timing, interference and signal assurance in contested environments.

“We were asked to stabilize timing within the LEO network, because timing is critical for many applications, particularly in contested environments where interference and disruption are real risks,” Canning said.

He said this requirement has driven the development of additional capabilities that are now being extended across Eutelsat’s broader service portfolio, improving reliability under adverse conditions.

Operators are also increasingly expected to provide tools for monitoring and managing network performance in real time, enabling users to respond more effectively to disruptions.

This includes the ability to detect interference, reroute traffic and maintain synchronization across networks, all of which are critical for applications ranging from communications to navigation and data processing.

Electronic warfare impact

Recent conflicts have highlighted the vulnerability of communications systems to electronic warfare, reinforcing the need for multi-layer architectures.

“The tactical SATCOM environment has changed significantly over recent years, and the lessons from Ukraine have shown very clearly how electronic warfare can disrupt communications and force operators to rethink resilience,” Canning said.

He said operators have had to adapt quickly to ensure services continue to function despite jamming and interference, thereby accelerating demand for systems that can switch between networks and maintain continuity.

In practice, that means building systems that are not only resilient by design but also adaptable in real time, allowing users to shift between different communication paths without interrupting operations.

The experience has underscored the importance of integrating multiple layers of connectivity to provide redundancy and flexibility in contested scenarios.

Assured communications and control

Beyond resilience, customers are increasingly seeking greater control over how communications services are delivered and managed.

“We describe our services as secure, agile, resilient and assured, and that includes providing transparent service-level agreements that customers can rely on, alongside capabilities such as reserving capacity, prioritising access and applying geofencing,” Canning said.

He said users expect not only connectivity but also performance guarantees, as well as the ability to allocate resources according to mission priorities.

Canning spoke at the Defence In Space Conference (DISC) 2025 in London, where participants from the defence and space sectors discussed how satellite communications are evolving to meet changing operational requirements.

The session focused on the integration of space and terrestrial networks, the impact of electronic warfare and the growing role of autonomous systems in shaping communications demand.

Eutelsat is a satellite operator that became the world’s first fully integrated GEO-LEO provider following its 2023 merger with OneWeb. The combined company operates more than 35 geostationary satellites and a LEO constellation of over 600 satellites, enabling it to deliver both high-throughput broadcast services and low-latency broadband connectivity globally.

Historically, Eutelsat has been a major provider of video services, broadcasting nearly 7,000 television channels for more than 7,000 broadcasters and reaching around 274 million homes worldwide.

The group has expanded its focus in recent years to include connectivity services for maritime, aviation, remote regions and government users, alongside continued investment in new technologies such as Eutelsat Quantum, a fully reconfigurable software-defined satellite.

As demand for resilient, integrated communications continues to grow, multi-orbit strategies are expected to play an increasingly central role in both commercial and defense applications, positioning operators such as Eutelsat at the forefront of the evolving satellite communications landscape.

Looking ahead, the convergence of satellite and terrestrial networks is expected to accelerate further, driven by advances in software-defined infrastructure, network virtualization and automation.

These developments are likely to enable more flexible and responsive communications architectures capable of supporting a wide range of mission requirements.

A new wave of artificial intelligence (AI)-driven software development is forcing enterprises to rethink how and when security is applied. Code generation is shifting from human engineers to autonomous agents operating at scale.

AI coding assistants are generating increasing volumes of production code. Traditional application security models, designed for slower, human-led workflows, are struggling to keep pace. This creates a growing gap between development speed and the ability to detect and mitigate vulnerabilities.

“We shift the security controls, the security tests, to the earliest possible point in time, which is when the agent generates the code. Without such additional help, the agents do not always produce functional code, and they do not produce secure code,” said Henrik Plate, head of security research at Endor Labs.

“Rather than checking security after development, we perform those checks when the agent generates the code. That is where we want to intervene,” he said. “If you wait until later, the volume of code is already too large to review manually.”

This shift toward earlier intervention reflects a broader structural change in software engineering. AI systems are no longer simply assisting developers; they are generating large portions of application logic. As a result, security must operate at machine speed and integrate directly into the code generation process.

The risks associated with AI-generated code stem in part from training data. Large language models are trained on a mix of secure and insecure examples drawn from open-source repositories and developer forums.

“There is good source code, and it’s great that the models have been trained on the good and secure code, but there is also a lot of insecure code,” Plate said. “When it generates code, you don’t really know whether it will be inspired by the secure code or the insecure code.”

This uncertainty introduces systemic risk. Organizations increasingly rely on AI-generated outputs without fully understanding their limitations. In many cases, vulnerabilities are embedded at the point of creation and propagate through software systems before being detected.

From coding to architecture

Endor Labs was founded in 2021 and is headquartered in Palo Alto. It provides an AI-native application security platform that secures both open-source dependencies and AI-generated code. The company combines program analysis, vulnerability intelligence, and automation to identify risks across complex software supply chains.

Its platform spans multiple layers of application security, including software composition analysis to inspect open-source dependencies, static application security testing for first-party code, container scanning, and secret detection. These capabilities are embedded into CI/CD (continuous integration and continuous delivery/deployment) workflows, so security checks run continuously as code is written and updated.

Plate spoke to TechJournal.uk in an interview on the sidelines of DevOpsLive, part of Tech Show London, on March 5. The discussion focused on how AI is reshaping software development practices and security strategies.

He said the company is also extending these capabilities directly into AI agents through integration mechanisms such as model context protocol (MCP) servers, allowing agents to access security intelligence while generating code rather than after the fact.

The rise of AI agents is fundamentally changing how developers build software, shifting the focus from writing code to defining system architecture.

“Rather than writing code line by line, we will develop by giving instructions, by expressing an architecture, explaining the design of the software that we want to be implemented by the agent,” Plate said. “You leave the implementation of that whole system to the agent.”

“You take a step back. You have a much bigger picture from the start, and try to describe a whole system, and you leave the implementation of that whole system to the agent,” he said. “It requires a certain seniority to describe a system in all its complexity.”

This transition is redefining engineering roles. Developers are increasingly expected to operate as system architects, while AI agents handle execution. At the same time, the volume of generated code is expanding rapidly, making manual review impractical.

In this environment, automated security tools are becoming essential not only for detecting vulnerabilities but also for guiding AI systems toward more secure outputs during development.

He said this requires a deep analysis of how code behaves, including tracking data flow through applications and across dependencies. By linking vulnerability databases, such as the National Vulnerability Database (NVD), with code-level insights, the platform can identify the exact functions and execution paths where risks occur. This reduces false positives and improves prioritization.

He added that scale is becoming a defining challenge, with large enterprise customers operating thousands of repositories and rapidly expanding codebases driven by AI-assisted development.

New attack vectors

One of the most immediate threats emerging from AI-driven development is the exploitation of hallucinated components. These are non-existent packages or dependencies suggested by models.

“The model would propose a package that does not actually exist,” Plate said. “Attackers realized this and started to actually create those packages that didn’t exist before in order to deliver malware.”

“You saw patterns in the names that were proposed in the hallucinations,” he said. “The attackers jumped on this occasion and started to create those packages.”

These incidents illustrate how vulnerabilities can arise not only from flawed code but from the interaction between AI systems and open ecosystems. As developers increasingly accept AI-generated suggestions, the risk of introducing malicious dependencies grows significantly.

The rapid emergence of new AI-related ecosystems is expanding the attack surface and creating additional opportunities for malicious actors.

“As soon as there is a new ecosystem, attackers jump more and more quickly on that opportunity in order to also deploy malicious components,” Plate said.

This pattern has been observed across emerging technologies, including new registries and integration frameworks that allow users to extend AI agents with additional capabilities. In many cases, attackers move faster than security teams can respond.

To counter this, Endor Labs has developed agentic workflows that scan newly published software packages at high frequency. He said public registries, such as NPM (Node Package Manager) and PyPI (Python Package Index), see around 40,000 new packages daily. Automated systems are required to analyze them within minutes to detect malicious behavior before widespread adoption.

The changing user base also contributes to rising risk. While traditional open-source attacks primarily targeted developers, newer ecosystems are being adopted by a broader audience with less security expertise.

Talent reshaped

The rise of AI-driven development is reshaping the software engineering labor market, particularly for entry-level roles.

Plate said the increasing reliance on AI agents is raising the bar for developers, who must now focus on system-level design rather than implementation details.

“It requires a certain seniority in order to describe a system in all its complexity, in relatively simple words, so it can be implemented by an agent,” he said.

As companies adjust to this new model, many are slowing hiring and reassessing the skills required for engineering roles. Junior developers face greater challenges entering the workforce as demand shifts toward more experienced talent.

At the same time, AI is lowering barriers to software creation, enabling a wider range of users to build applications without deep technical expertise.

“It’s a huge enabler as well,” Plate said. “Many more people can create software, and so that is a good thing. We just need to make sure that this is all secure and not subject to vulnerabilities.”

Looking ahead, the balance between speed and security will define the next phase of software development. As AI continues to accelerate code production, enterprises will need to embed security directly into development workflows. This ensures innovation does not come at the cost of resilience.

Enterprises are increasingly misjudging cybersecurity priorities as artificial intelligence (AI) accelerates data flows, focusing on blocking intrusions at the perimeter rather than limiting how far attackers can move once inside corporate systems.

As corporate data spreads across cloud platforms, SaaS applications, endpoints, and legacy infrastructure, the critical risk is no longer initial access but how much sensitive information can ultimately be exposed, moved, and exploited after a breach.

“If an attacker gets in, how much damage can they actually do? That’s really the question,” Stuart Wilson, director of sales engineering at Forcepoint, told TechJournal.uk in an interview.

“A lot of organizations have created their own security problems over time,” he said. “They’ve deployed technology in silos, and they don’t have a consistent way of identifying or protecting data.”

The shift toward data exposure risk reflects a broader structural weakness in enterprise security. Many organizations lack consistent classification, access controls, and visibility across fragmented systems, increasing the likelihood of large-scale breaches.

In practice, sensitive information is often duplicated across environments without clear ownership, while access rights accumulate over time without regular review. As a result, a single compromised account can expose far more data than security teams anticipate, particularly in hybrid cloud environments where policies are not consistently enforced.

Wilson said this lack of control is compounded by the speed at which organizations have adopted cloud and SaaS platforms over the past decade, often without redesigning security frameworks to match the new operating model.

Security teams often focus on external threats such as phishing or malware, but internal gaps—such as excessive permissions or poorly managed data flows—can amplify the impact of any intrusion.

Data-first strategy

Wilson said these challenges are driving a transition toward data-centric security models, where protection follows information rather than infrastructure.

Forcepoint, a US-based cybersecurity company specializing in data protection, is positioning its platform around this approach as organizations scale artificial intelligence (AI) adoption.

“We’ve been focused on data security for a long time—it’s part of our heritage,” he said. “Our approach is about data security everywhere—across endpoints, networks, cloud and SaaS.”

The company integrates capabilities across endpoints, networks, cloud environments, and SaaS platforms, combining data security posture management (DSPM) and data detection and response (DDR).

DSPM provides continuous discovery and classification of sensitive data, helping organizations understand where critical information resides and how it is exposed. DDR adds a response layer, enabling security teams to detect suspicious activity and take action before data is exfiltrated.

Together, these capabilities are designed to shift security operations from reactive incident response to proactive risk reduction, particularly in environments where data is constantly moving across systems.

“It’s not just about visibility—it’s about control and response,” Wilson said.

Platform expansion

Forcepoint said recent platform updates focus on automating policy creation, improving incident response and strengthening protection across AI-driven workflows.

These include:

• AI-assisted policy generation using natural-language inputs
• Automated detection of anomalous data access patterns
• Faster incident response through integrated analytics

The company said these capabilities aim to reduce response times and improve consistency as organizations manage growing volumes of sensitive data.

Forcepoint’s recent announcements also highlight a broader push to embed AI more deeply into security workflows. The platform’s AI layer is designed to assist analysts by prioritizing alerts, summarizing incidents, and recommending remediation actions, thereby reducing the operational burden on security teams as data and alert volumes increase.

Regulatory pressure

The regulatory landscape is also shaping enterprise security strategies, particularly for companies operating across multiple jurisdictions.

At the UK Cybersecurity Expo, part of Tech Show London, held on March 5 in London, Wilson discussed how global compliance requirements are converging around stricter data governance standards.

“If you operate across multiple regions, you’re going to have to align with the strictest regulations anyway,” he said.

The trend mirrors earlier shifts driven by the General Data Protection Regulation, which forced companies worldwide to adopt higher data protection standards.

Forcepoint's Chief Data Strategy Officer, Ronan Murphy, said in a byline article that the European Union’s AI Act is expected to extend this approach by imposing risk-based requirements on how AI systems are developed and deployed.

The EU AI Act introduces obligations around data quality, transparency, and accountability, particularly for high-risk applications, reinforcing the need for stronger data governance.

Murphy noted that organizations deploying AI systems will need to demonstrate how data is sourced, processed, and protected throughout the application lifecycle. This includes maintaining audit trails, ensuring data integrity and implementing controls to prevent misuse or unintended exposure.

Organizations that fail to meet these standards may face operational restrictions or financial penalties, making early compliance a strategic priority.

Wilson said companies should not wait for domestic regulation to take effect before strengthening their data controls.

“We saw that with General Data Protection Regulation (GDPR), and we’re likely to see the same pattern with AI,” he said.

AI and data trust

Beyond compliance, Wilson said the effectiveness of AI depends on the reliability of the underlying data.

“AI is only as good as the data you put into it,” he said. “If you don’t trust your data, you can’t trust your AI.”

Many organizations remain at an early stage of what he described as a data maturity journey, lacking clear oversight of data ownership, classification, and usage.

“This is really a data maturity challenge,” he said.

Without strong governance, AI systems risk amplifying existing weaknesses, including data leakage and compliance failures.

In particular, generative AI tools introduce new pathways for sensitive data to be accessed, processed and potentially exposed, especially when employees interact with external models or integrate AI into business workflows without clear policies.

Wilson said organizations need to treat AI not just as a productivity tool but as a new layer of data risk that must be governed with the same rigor as traditional systems.

Forcepoint is incorporating AI into its own platform to address these challenges, particularly in automating security operations.

“We’re using AI to help automate policy creation and response,” Wilson said.

The company said its strategy is to enable real-time protection as enterprise environments become more distributed and complex.

As organizations expand digital infrastructure and deploy AI at scale, Wilson said the focus must shift from preventing breaches to controlling their impact.

That requires a clear understanding of data, consistent governance and the ability to respond quickly when incidents occur.

The race to embed artificial intelligence (AI) into manufacturing is accelerating, but the benefits remain unevenly distributed. As governments and large corporations push ahead with industrial AI, smaller factories risk being left behind due to structural constraints in talent, capital, and production models.

China has moved aggressively to formalize this transition. On January 7, 2026, the Ministry of Industry and Information Technology released a sweeping “AI + Manufacturing” action plan, marking a shift from policy guidance to execution. The initiative aims to upgrade more than 50,000 enterprises by 2028, embedding AI across production, logistics, design, maintenance, and management.

The plan targets the development of 3–5 general-purpose industrial models, 100 high-quality datasets, and more than 500 application scenarios, while promoting embodied AI, industrial intelligent agents, and humanoid robotics.

Shinichiro Nakamura, president of Daiwa Steel Tube Industries and one to ONE Holdings Pte. Ltd., said China’s rapid push to integrate AI into manufacturing is forcing Japan and other Asian producers to accelerate their strategies, with pressure building across the region.

“We recognize that this pressure is higher for large corporations and tends to be relatively lower for smaller-scale entities,” Nakamura said. “Furthermore, against the backdrop of recent geopolitical factors, the impact of tariffs and regulations is also perceived to be growing significantly.”

He said the fundamental reasons for the large AI gap between factories of large corporations and SMEs lie in the quality of existing resources and the required performance levels, adding that there is a clear difference in the quality of talent that large corporations and SMEs can attract.

“Regarding required performance, large corporations typically engage in mass production of a few product types, while SMEs handle small-batch production of many product types,” he said. “The former environment facilitates automation with robots, requires less human involvement, and allows for easier investment, whereas the latter is the complete opposite. This gap is expected to be more severe during the initial AI implementation phase, easing somewhat afterward. However, fundamentally closing this gap is not easy and will require time.”

The divergence reflects deeper operational realities: large manufacturers benefit from standardized processes and scale that make automation more viable, while smaller factories face variable production environments that complicate AI deployment.

Geopolitical pressures, including tariffs and regulatory shifts, are accelerating the need for industrial transformation, particularly among globally exposed manufacturers.

AI adoption gap

Nakamura shared his views in an email interview with TechJournal.uk, focusing on China’s industrial AI strategy and the direction of manufacturing in Asia.

He said that, despite the scale of China’s ambitions, the reality of AI adoption remains uneven, particularly among small- and mid-sized manufacturers.

“For SMEs in manufacturing, AI adoption is progressing in terms of general-purpose applications through tools like Gemini and NotebookLM included in Google Workspace, or Microsoft’s Copilot and ChatGPT,” he said.

“However, attempts to transform the entire workflow of their manufacturing processes through AI adoption are still rare and are considered to be in the developmental stage.”

While awareness of AI tools is growing, deeper integration into production systems—such as predictive maintenance, automated quality control, and end-to-end optimization—remains limited outside large enterprises.

Government policies are likely to reinforce this divide, favoring companies with scale and resources.

“Large corporations, with their significant social impact, abundant resources, and ability to deliver results and outcomes, are likely to lead the way,” Nakamura said. “Conversely, micro-enterprises, with their limited social impact and scarce resources, face a high risk of being left behind.”

Policy and paths

Japan is also accelerating its Fourth Industrial Revolution agenda, aiming to quadruple investment in AI and semiconductors to boost productivity and offset a shrinking workforce. The push aligns industrial policy, capital deployment, and skills development to modernize manufacturing at scale.

Central to this effort is the 2025 AI Promotion Act, a pro-innovation framework that prioritizes AI development and adoption across industry and the public sector. Overseen by the prime minister–led AI Strategic Headquarters, it emphasizes research funding, talent cultivation, and shared computing infrastructure.

The strategy builds on “Society 5.0,” integrating data-driven AI across manufacturing, healthcare, and mobility to address labor shortages. It favors agile, “soft law” governance with voluntary risk management and international coordination under G7 Hiroshima AI Process guidelines.

Against this policy backdrop, manufacturers face two distinct strategic paths: either accelerating to capture first-mover advantage or adopting a more incremental approach.

“We recognize two realistic paths: either taking risks to proactively pursue first-mover advantages, or undertaking the bare minimum based on government support and partnerships with large companies,” he said. “This decision depends on each company’s strategic intent.”

Stronger ecosystem support will be essential to enable more companies to pursue the first path.

“At the same time, we recognize that establishing mechanisms not only by government and public administration but also by financial institutions to support the former through subsidies and other means is extremely useful,” he said.

Such mechanisms could include targeted financing, shared industrial data platforms, and standardized AI tools tailored to specific manufacturing sectors.

Leadership profile

Nakamura is a third-generation industrialist and the grandson of Tomeichi Nakamura, who founded Nakamura Pipe Manufacturing Factory (now Daiwa Steel Tube Industries) in 1932. The company has grown into a global manufacturer of galvanized steel tubes, supported by proprietary technologies such as the in-line galvanizing “Daiwa Z Process.”

Under Shin Nakamura’s leadership, the business has expanded beyond traditional manufacturing. Having joined the company in the 1990s and becoming president in 2003, he led a restructuring of operations by adopting advanced ICT systems, modern management practices, and global expansion initiatives. Through Singapore-based o2Oh, he now oversees a network of companies spanning Japan, Vietnam, the United States, and India, combining industrial operations with digital innovation.

The group includes:

• Daiwa Steel Tube Industries — a manufacturer of galvanized steel tubes and developer of in-line galvanizing processes

• IndustrialML — develops real-time data platforms for smart factories

• Superior Technologies (California) — focuses on advanced galvanizing systems and monitoring equipment

• Daiwa Lance International (Vietnam) — manufactures lance pipes and supplies customers in more than 50 countries

Together, these companies integrate AI, machine learning, and industrial Internet of Things (IoT) technologies into production environments.

His strategy focuses on bridging the gap between legacy manufacturing and digital transformation, using real factory environments to test and scale new technologies.

Execution challenge

Looking ahead, the fourth Industrial Revolution will be defined not only by technological breakthroughs but also by how effectively they are implemented across diverse industrial environments.

His group’s approach combines traditional manufacturing expertise with investments in AI-driven systems and digital infrastructure. By developing and testing solutions within its own factories, and extending them across its global network, the company aims to create scalable models for industrial transformation.

The challenge lies in integration rather than invention—aligning advanced technologies with the realities of factory operations, workforce capabilities, and economic constraints.

In that context, the success of industrial AI may depend less on headline innovation and more on disciplined execution across thousands of factories worldwide, particularly among those currently at risk of being left behind.

The UK is doubling down on AI and quantum with a £2.5 billion push to scale both technologies at pace. The plan pairs a new Sovereign AI Fund with a major expansion of quantum infrastructure and procurement.

The coordinated announcement, made on March 17, signals a more interventionist approach to technology policy, with ministers aiming to anchor high-growth companies domestically while building national capability in next-generation computing systems.

Chancellor of the Exchequer Rachel Reeves said the government faces a strategic choice in shaping emerging technologies, warning against leaving development solely to market forces or foreign players.

She said the UK aims to achieve the fastest adoption of AI in the G7 while maintaining domestic control over its direction.

She said quantum computing represents a generational leap in capability, with the potential to transform industries ranging from healthcare to energy. The government intends to ensure these technologies are developed, deployed, and commercialized within the UK.

Sovereign AI push

The first strand of the announcement focuses on strengthening the UK’s position in artificial intelligence through a Sovereign AI Fund and broader ecosystem support.

The government will launch a £500 million Sovereign AI Fund in April to provide capital, compute resources, and infrastructure support to high-potential British AI firms. The initiative aims to help companies scale domestically while remaining competitive globally.

The fund is part of a broader policy shift to align AI development with national priorities, particularly in economic resilience, public services, and industrial competitiveness.

Key elements of the AI package include:

• £500 million Sovereign AI Fund to support scaling companies
• Part of a wider £2.5 billion combined AI and quantum investment package
• Target to achieve the fastest AI adoption in the G7
• Access to compute infrastructure and technical resources
• Support to anchor high-growth AI firms in the UK
• Integration with broader industrial strategy focused on digital technologies
• Alignment with regional growth and skills development initiatives
• Backing for startups and scale-ups across the UK’s AI ecosystem

The UK already has a strong foundation in AI, with a high concentration of startups and research talent.

Since 2020, more AI companies have been founded per capita in Britain than in any other European country, reinforcing its position as a leading innovation hub.

The government’s strategy seeks to convert that early-stage advantage into sustained commercial leadership, ensuring that companies not only launch in the UK but also grow and remain headquartered there.

Quantum scale-up

Alongside AI, the government has outlined an ambitious plan to establish the UK as a global leader in quantum computing, committing up to £2 billion to the sector.

Central to this effort is a first-of-its-kind procurement programme, branded “ProQure: Scaling UK Quantum Computing,” which will support the development and deployment of large-scale quantum computers in the UK by the early 2030s.

Under the ProQure programme, companies will develop prototype systems for evaluation, with the most promising progressing to larger-scale machines integrated into national infrastructure. The approach is designed to stimulate demand, unlock private investment, and accelerate the transition from research to commercial deployment.

The programme bridges the gap between research and commercialization by providing guaranteed demand for emerging technologies, enabling companies to raise capital and accelerate development.

Key components of the quantum investment include:

• Up to £1 billion for procurement of commercial-scale quantum computers
• Over £500 million for quantum computing development and applications
• More than £400 million for sensing, navigation, and supporting infrastructure
• £125 million for quantum networking and £205 million for sensing technologies
• £13.8 million for five National Quantum Research Hubs
• Additional funding for skills, commercialization, and infrastructure programs

The initiative will invite companies to develop prototype systems, which will then be evaluated before scaling into operational machines for public and commercial use.

The government expects quantum technologies to deliver significant economic benefits, including improved drug discovery, more efficient energy systems, and enhanced national security capabilities. Estimates suggest the sector could generate more than £200 billion in economic impact over the coming decades.

The UK already hosts one of the world’s most advanced quantum ecosystems, supported by more than a decade of public investment and a growing base of private-sector participation. The new funding aims to consolidate that position by accelerating large-scale deployment.

Industry response

Industry stakeholders broadly welcomed the announcement, while emphasizing the need for supporting infrastructure alongside hardware investment.

“Quantum computing is moving from promise to practical impact - unlocking new computational capabilities that can help solve complex real-world problems,” said Philip Intallura, HSBC Group Head of Quantum Technologies. “It will also reshape how we think about cybersecurity, making it vital that we invest now in quantum-safe approaches.”

He added that it is encouraging to see the UK government backing the skills, research and scale-up needed to turn world-class science into deployable capability and long-term economic growth.

Dan Holme, chief executive of Qoro Quantum, said the ProQure programme addresses a critical gap in demand but warned that hardware alone will not deliver value.

“The ProQure programme is exactly the right move, but the procurement of quantum hardware only delivers value if there is an intelligent software layer to orchestrate workloads across those machines,” he said. “Without the proper software, these early quantum machines are incredibly difficult and expensive to operate. It’d be like putting a manual gearbox into a modern F1 car—you’re not maximizing the potential of the hardware.”

Professor Callum Littlejohns, deputy director of CORNERSTONE, said the focus on scaling and deployment strengthens the strategy and that infrastructure will be key to supporting domestic firms.

“It’s right that the UK government has put scaling at the heart of its quantum leap scheme,” he said. “To nurture homegrown firms, we need the right infrastructure in place to ensure companies can access the capabilities they need to grow.”

The combined announcements reflect a broader policy direction that places advanced computing at the center of economic strategy, with the government seeking to align investment, infrastructure, and industry development under a unified framework.

As global competition intensifies across AI and quantum technologies, the UK’s approach reflects a renewed emphasis on sovereign capability, balancing open innovation with strategic control over critical technological assets.

Artificial intelligence (AI) is moving from model training to real‑world usage—and that shift is rewriting data‑center demand. The bottleneck is no longer computing; it is power, land, and time to connect.

Across Europe, AI workloads are rising faster than infrastructure can keep up. Grid capacity, permitting, and skilled labor are becoming binding constraints, widening the gap between surging demand and the pace of new supply.

“Europe will depend on how much these models are used. Inferencing is going to be the main driver for growth of AI loading in European markets,” said Philippe Diez, Senior Client Advisor at McKinsey.

Diez said data center demand is expected to grow rapidly through 2030, with annual growth of about 20% even in a worst-case scenario.

Maurice Mortell, Chairman of Digital Infrastructure Ireland, said the rapid expansion of data centers is already running into physical limits in key hubs such as Dublin, where grid constraints are becoming harder to overcome.

Diez and Mortell delivered opening-market perspectives on March 11 in Dublin, as Pure Data Centres Group (Pure DC) launched Europe’s first large-scale data center microgrid. The 110-megawatt on-site system, developed with AVK, will support hyperscale cloud and AI workloads across Europe and the Middle East, reflecting how rising AI demand is reshaping infrastructure and energy strategy.

Demand meets limits

Diez outlined how global data center demand is set to accelerate sharply through the end of the decade.

Growth is driven by continued cloud adoption and the rapid expansion of AI workloads, including both model training and inferencing.

While training workloads remain concentrated in the United States and China, Europe is expected to shift more strongly toward inferencing as enterprises deploy AI into real-world applications.

This transition will reshape infrastructure design, favoring distributed capacity closer to users rather than large centralized training clusters.

The supply of AI computing power, however, is struggling to keep pace.

Diez said the industry faces a shortfall of between 20% and one-third of projected demand, driven by delays in grid connections, administrative bottlenecks, and shortages of skilled labor.

“This gap is mostly linked to time to power, as well as the availability of people and administrative delays,” he said.

The growing importance of “time to power” is forcing operators to rethink deployment strategies.

Instead of relying solely on grid connections, companies are exploring new regions, repurposing industrial sites, and developing on-site energy generation.

“One of the main blocking points is how fast you can get power,” Diez said. “We see several strategies from operators to reduce this constraint, including moving to new locations, repurposing industrial sites, and building their own power.”

Power strategies shift

As power availability becomes the defining constraint, energy strategy is emerging as a core component of data center design.

One of the most significant shifts is the growing adoption of on-site generation and microgrid systems, allowing operators to bypass grid limitations and accelerate deployment timelines.

Fully renewable solutions remain difficult to implement at scale due to intermittency and storage constraints.

As a result, many operators are turning to hybrid models that combine renewables with dispatchable energy sources.

“Pure renewable solutions are not going to be the solution on their own,” Diez said. “You need storage and often a mix with gas-based solutions.”

“Gas-based solutions are probably the most used and chosen technology right now, at least for the next five years,” he added.

This shift is also driving new commercial models.

Rather than building and operating energy infrastructure themselves, developers are increasingly partnering with utilities and energy providers.

“Data center operators will increasingly want Energy-as-a-Service (EaaS),” Diez said. “They want someone to invest and operate power infrastructure for them.”

Ireland case study

Mortell provided a national perspective on how these trends are unfolding.

Ireland has emerged as one of Europe’s leading data center hubs over the past decade, attracting major hyperscale operators including Microsoft, Amazon, Google, and Meta.

“Ireland was very successful over the last decade in promoting itself as a destination for digital infrastructure investment,” Mortell said.

That growth was supported by strong transatlantic connectivity, a favorable business environment, and early investment by global technology companies.

Over time, clustering effects in Dublin reinforced Ireland’s position as a key node in global digital infrastructure.

However, success has created new challenges.

Rapid expansion has placed significant strain on Ireland’s electricity network, particularly in Dublin, where capacity constraints have become acute.

“Dublin City is really constrained, and that’s just the physics of it,” Mortell said.

“There are massive constraints on both the distribution and transmission networks, and that requires huge capital investment,” he added.

These pressures have triggered regulatory intervention and a broader reassessment of how digital infrastructure is planned and delivered.

Policymakers now face competing priorities, including economic growth, decarbonization targets, and the expansion of digital services.

“We have competing priorities: digitization, economic growth, and renewable energy goals, and they don’t always align,” Mortell said.

Infrastructure as utility

Looking ahead, Mortell said the industry must adapt to a new reality in which digital infrastructure is treated as critical national infrastructure rather than discretionary investment.

He said governments and industry stakeholders will need closer coordination on long-term planning, grid investment, and energy strategy to ensure infrastructure keeps pace with demand.

“Digital infrastructure is now a utility. It’s not a nice-to-have, it’s a have-to-have,” Mortell said. “It’s like a road, an airport, or a water network.”

The shift toward inferencing-led demand, combined with tightening infrastructure constraints, suggests the next phase of AI growth will depend as much on energy systems and policy frameworks as on advances in computing technology.

Engineering courses at universities worldwide are producing a steady stream of graduates, but data center operators are struggling to convert them into deployment-ready talent. As demand for artificial intelligence (AI) infrastructure accelerates, companies are rethinking how they attract, train and retain early-career professionals.

This mismatch between the number of available graduates and the skills required by industry has become a defining constraint, particularly in power generation and in critical infrastructure supporting modern data centers.

“It’s an amazing industry with a wide range of skills and opportunities, but we need to make it accessible and ensure the training is there to bring people in at a young age,” said Charlotte Berry-Selwood, Chief Delivery Officer at AVK.

“AVK Academy was basically our solution to the fact that we really couldn’t hire service engineers,” she said. “Trying to hire talent out of the industry was not possible for the scale that we needed.”

The issue reflects a gap between academic training and the hands-on, deployment-ready skills required for data center delivery.

Growing talent

Berry-Selwood spoke to TechJournal.uk in Dublin on March 11 as Pure Data Centres Group (Pure DC) launched Europe’s first large-scale data center microgrid, a 110-megawatt on-site system developed with AVK, to support hyperscale cloud and AI workloads across Europe and the Middle East.

She said companies across the sector have relied on hiring experienced professionals from a limited talent pool rather than investing in long-term workforce development.

AVK, a power solutions provider focused on data center infrastructure with about 350 employees, has begun to shift that model through its internal training programme.

The AVK Academy, launched in 2024 in Kent, combines classroom instruction with hands-on training using real-world equipment, including full-scale power systems. The programme is designed to move trainees from foundational knowledge to operational deployment roles.

The facility is equipped with full-size Rolls-Royce mtu engines and control system rigs, allowing trainees to work on the same technologies used in live data center environments. Courses are aligned with external standards, including City & Guilds certifications, ensuring participants gain recognised qualifications alongside practical experience.

Training follows an apprenticeship-style, work-learning model over roughly two to three years, with participants splitting their time between on-site deployment and formal instruction, gaining technical experience and formal qualifications through external partnerships.

Participants are linked with colleges to ensure their training translates into professional credentials, while also gaining exposure to live projects and operational environments.

Berry-Selwood said the initiative is expanding beyond service engineering into broader disciplines, including engineering, project management, and quantity surveying.

“We will be bringing young people in through a graduate-style scheme and expanding that across different functions,” she said.

She said the model creates multiple career pathways, allowing trainees to remain in service engineering or move into engineering, project delivery, or commercial roles as they gain experience.

At present, the intake remains small. Berry-Selwood said AVK is taking on about three to four people a year through the academy, but wants to increase that as its mentoring capacity and training structure expand.

She said this approach reflects a broader shift in workforce development in a sector that has scaled rapidly but lacks the institutional depth of older industries.

“The expectation that you’re going to invest in people for a number of years before they are productive needs to be understood by businesses,” she said.

Maturity gap

Berry-Selwood described the data center sector as relatively young compared with other industrial segments, noting that it has not yet fully developed the structured graduate pipelines seen in more established fields.

“We’ve existed for about 30 years as an industry, but we still lack maturity in some areas,” she said. “The talent isn’t going to come from a magical pool that we haven’t found yet. It’s only going to come from homegrown investment.”

To support early-career professionals beyond initial training, AVK has backed external initiatives such as GeN+1, a community platform for networking and career development.

As an impact sponsor, the company is working with GeN+1 to expand outreach, promote career awareness, and bridge the gap between education and employment. The initiative focuses on helping young professionals build networks, access mentorship, and better understand progression pathways within the data center sector.

“GeN+1 is for young people who have come into the industry to give them a community space and a networking opportunity,” she said. “Networking is so important as you start out your career.”

The company is also involved in broader outreach efforts to raise awareness of data center careers among students and graduates, including collaborations with schools and industry groups.

Berry-Selwood said improving visibility is critical, as many candidates are unaware of roles across engineering, operations, and commercial functions.

She added that the industry must engage earlier with schools and universities to explain the breadth of opportunities available, including roles that combine technical, operational, and business skill sets, as well as opportunities to work across European markets.

Diversity edge

Beyond talent pipelines, AVK is also focusing on improving diversity within its workforce, particularly in an industry that remains heavily male-dominated.

The company currently has around 10% female representation, above the industry average but still significantly below parity.

“We know statistically that the more women you have on the board, the more profitable your company is likely to be,” Berry-Selwood said. “We are committed to making meaningful change with clear targets that we can actually meet and close the loop on.”

She said AVK is developing a corporate social responsibility framework with measurable diversity targets, aiming to move beyond symbolic commitments and deliver tangible outcomes.

Scaling pressure

Alongside workforce challenges, the rapid expansion of AI-driven data centers is placing new demands on infrastructure delivery, particularly energy systems.

Berry-Selwood said the pace of development is forcing companies to design and build simultaneously, increasing execution complexity.

“The speed to market is incredibly fast, so you’re trying to design and build on the fly, and you are going to run into issues,” she said.

She added that microgrid deployment introduces an additional layer of uncertainty, as projects often lack established templates.

“With microgrids, there isn’t a well-trodden path. We’re having to come up with solutions in real time,” she said.

Despite these challenges, she said early deployments provide valuable learning that can be scaled across future projects, positioning companies to support the next phase of AI infrastructure growth.

As demand for compute capacity rises, she said the industry’s ability to build both infrastructure and human capability will determine how quickly it can scale.

She said long-term investment in people, alongside innovation in energy and delivery models, will be essential to sustaining that growth trajectory.

AVK, long known as a supplier of standby generators and backup power systems for data centers and other critical facilities, has repositioned itself as a strategic power‑systems provider as the rapid growth of artificial intelligence (AI) infrastructure collides with electricity‑grid constraints across Europe.

The shift reflects a broader industry-wide rethink as hyperscale facilities seek new ways to secure reliable power while advancing decarbonization. One example is a new project in Dublin that shows how developers are redesigning data‑center energy infrastructure when public electricity networks cannot expand quickly enough to meet demand.

Pure Data Centres Group (Pure DC), a hyperscale cloud and AI data‑center developer active across Europe and the Middle East, launched Europe’s first large‑scale data‑center microgrid at its Dublin campus on March 11. The 110‑megawatt on‑site system, developed with AVK, allows the campus to generate electricity during early development stages before full grid capacity becomes available.

“AVK as a business completely transformed its model. We moved away from being just a standby generator provider into a strategic power-systems solution provider,” said Ben Pritchard, chief executive officer of AVK.

“The biggest challenge was moving into the unknown. None of us truly knew what the journey would look like. No template ever existed for a data-center microgrid like this,” he said.

Pritchard said the transformation was driven largely by a structural shift in the digital infrastructure sector, where access to electricity has become one of the most significant constraints on new data center development.

“Seven years ago, grid constraints weren’t even a feature. Data centers were built in a very traditional way with two grid connections, a backup power source and UPS (Uninterruptible Power Supply) systems,” he said.

Today, however, many new facilities must design alternative power architectures from the outset, as securing grid connections is becoming increasingly difficult.

Reinventing power architecture

Pritchard spoke during a media tour on March 11 of Pure DC’s new data-center campus in Dublin, where developers have deployed a large on-site microgrid designed to supply power independently of the national grid.

Located in the fiber‑rich Ballycoolin area, the campus is designed to deliver 54 megawatts across three data centers: DUB01 (14MW), DUB02 (24MW), and DUB03 (16MW). DUB01 has been operational since 2024 and supports high‑density workloads with advanced cooling and power infrastructure.

In 2023, Pure DC also acquired an additional 25 acres west of the site for a potential 90-megawatt expansion, subject to permitting and power availability.

Pritchard said building such systems requires combining multiple technologies rather than relying on a single generation method.

“It was very apparent that one design wasn’t going to facilitate this. It wasn’t just going to be a group of gas engines. We had to move into an integrated engineering architecture with layers of technology,” he said.

The architecture incorporates several power layers designed to achieve “five-nines” reliability or a standard of 99.999% availability typically required by hyperscale data centers.

These include:

• gas-powered generation systems that provide a primary electricity supply

• battery energy-storage systems (BESS) that balance load fluctuations and improve efficiency

• liquid-fuel backup generators capable of maintaining operations in the event of a gas supply failure.

In the early stages of a data center’s life, computing load ramps up slowly, meaning engines would otherwise run inefficiently at low output.

Battery storage can absorb excess generation or discharge electricity when demand rises, smoothing fluctuations and improving overall system performance. It also allows the facility to maintain stable operations while gradually scaling capacity as customers deploy new computing workloads.

Pritchard said designing such systems required AVK to bring in expertise from several adjacent industries, including gas‑engine engineering, energy‑storage integration and high‑voltage control systems.

The company also expanded partnerships with technology suppliers, including Rolls‑Royce and Finnish engine manufacturer Wärtsilä, whose larger dual‑fuel engines are used in the permanent microgrid design planned for the Dublin campus.

Pritchard said the dual‑fuel configuration allows the engines to run primarily on natural gas but switch automatically to hydrotreated vegetable oil (HVO) if the gas supply is interrupted, ensuring uninterrupted data‑center operations.

He said the complexity reflects the operational requirements of data centers, which must maintain uninterrupted service even as computing workloads ramp up unpredictably.

“When you design a system for a manufacturing plant, you usually know exactly how the equipment will behave. With a data center, you know the final capacity you need, but the ramp rate of the load is completely undefined,” he said.

“This is actually the blueprint for microgrids. Everything that we applied to this scheme will be applied to larger schemes and future microgrids.”

While microgrids are sometimes described as temporary solutions until grid connections become available, Pritchard said the systems can play a more active role in energy infrastructure.

“People often talk about microgrids as a bridging solution, but it’s much more than that,” he said. “What we’ve built removes demand from the grid and, once connected, allows the infrastructure to participate in grid activity.”

Biomethane decarbonization test

Alongside the power-system design, the project has also tested new approaches to lowering emissions from gas-based generation.

Pure DC on March 12 announced that its Dublin campus achieved what it described as Europe’s first successful biomethane proof of concept for data-center operations.

During 2025, the facility matched all of its operational natural-gas consumption with certified renewable biomethane attributes.

The biomethane was accounted for through a combination of Irish Renewable Gas Guarantees of Origin and European biomethane certificates that track renewable-gas production and consumption across the market.

The accounting framework aligns with established European reporting standards, including the EU Emissions Trading Scheme (ETS) and widely used corporate climate-reporting methodologies.

The initiative also reflects Ireland’s broader policy strategy to expand renewable-gas production while reducing emissions from large energy users.

Ireland’s national biomethane strategy aims to scale domestic renewable-gas supply while stimulating demand from sectors such as transport and industrial infrastructure.

Pure DC said the proof-of-concept project demonstrates that renewable gas could act as a transitional mechanism for decarbonizing power systems where grid-based renewable electricity remains constrained.

Designing sustainability from the start

Maria Jose Rivas-Duarte, director of sustainability and ESG at Pure DC, said the project’s environmental strategy was embedded in the design process from the earliest stages.

“For us, sustainability is not a bolt-on afterthought. We think about this from the very onset of design,” she said.

She said the company evaluates several environmental dimensions simultaneously when designing new data centers, including energy use, water consumption, construction emissions and waste management.

The approach also considers the full lifecycle carbon impact of both the facility and the supporting power infrastructure.

“When we decided to go for an off-grid solution through a microgrid, decarbonization became critical. We knew we were going to use gas and have emissions,” Rivas-Duarte said.

To address that challenge, the company worked with energy suppliers and regulators to source renewable biomethane attributes linked to gas injected into the European network.

“In 2025, we delivered in a live data center 100% of operational gas backed by biomethane,” she said.

The supply mix included both domestic and imported renewable gas, reflecting the early stage of Ireland’s biomethane production sector. During the proof‑of‑concept phase, roughly 40% of the biomethane supply originated from Irish production, while the remainder was sourced through European renewable‑gas markets.

Rivas‑Duarte said the blended approach helped demonstrate that renewable‑gas accounting frameworks could work for large digital-infrastructure facilities as domestic supply gradually expands.

She said renewable gas should be viewed as a transitional pathway rather than a permanent solution.

“This approach demonstrates that resilience and decarbonization do not need to be decoupled. We can deliver power and decarbonization at the same time,” she said.

Looking ahead, Pure DC plans to explore long-term biomethane purchase agreements, expand renewable-energy integration and develop partnerships that allow excess heat from data-center operations to be exported to district-heating systems.

The company has set a target of achieving net-zero operational emissions by 2040.

Pritchard said the broader industry will increasingly need to integrate energy infrastructure directly into digital infrastructure projects.

He added that in the energy transition, mass electrification and rising grid demand do not always align, and that data centers deploying microgrids and dispatchable power could become part of the solution.

Pure Data Centres Group (Pure DC), a hyperscale cloud and artificial‑intelligence (AI) data‑center developer active across Europe and the Middle East, has chosen to rely on natural gas as a temporary power source for the microgrid at its new Dublin data‑center campus while waiting for full electricity‑grid capacity.

The campus operates a 110‑megawatt on‑site microgrid using natural‑gas generators supported by battery storage. The engines are designed to run on dual fuels, allowing the facility to switch to hydrotreated vegetable oil (HVO) as a lower‑carbon backup fuel when needed.

“We are not only relying on the gas network because it’s a third‑party utility network. We have to provide our own backup fuel on site, so the engines are dual‑fuel,” said Ian Whitfield, Chief Technology Officer, Pure DC. “In this case, we use HVO, a sustainable alternative to diesel.”

“We very much see this as a step on our journey. It’s been a long and very difficult journey, and we’re not quite finished yet, but we will go live later this year,” Whitfield said. “It’s part of a greater solution and a blueprint that hopefully others can follow.”

The campus in the Ballycoolin area of Dublin, Ireland, is designed primarily for several wholesale data‑center customers rather than hundreds of retail colocation tenants.

Pure DC is currently negotiating with several potential hyperscale clients and expects to close some large wholesale agreements within about a month. Each contract could span 10 to 15 years or longer.

Reliability engineering

Whitfield said hyperscale facilities must deliver extremely high uptime because they host essential digital services ranging from communications platforms to financial systems and healthcare infrastructure.

“We have to deliver continuous critical services. In our world, we deliver what we term a five‑nines (99.999%) infrastructure. What does that mean? Effectively, you’ve got just over 300 seconds of the year that you don’t have to provide availability for,” he said.

“To all intents and purposes, these facilities are designed to be up for 100% of the year continuously.”

Whitfield said the reliability target requires multiple layers of redundancy across the generation and electrical distribution systems.

“There are additional engines because we know that these engines require maintenance after so many thousands of hours,” he said. “There are duplicate power feeds from the energy center into the building, so that if we have an outage on one, or we need to do maintenance on one, we can run the facility on the other.”

Whitfield said the control system is duplicated so engineers can maintain it, perform updates, and handle network switch failures while still providing continuous services.

Private power grid

Whitfield presented the project on March 11 during a media tour in Dublin. Executives explained how the site replicates the functions of a traditional electricity grid inside the campus.

Whitfield said a traditional data center typically draws electricity from the public grid and keeps generators as backup. Pure DC replicated that arrangement on campus by operating its own on‑site power generation plant.

The campus includes three data‑center buildings housing multiple data halls designed for high‑density AI computing.

Whitfield said the energy infrastructure sits behind the buildings and currently provides about 70 megawatts of primary generation capacity.

“What you can see is the first two capacity blocks of our three‑capacity‑block energy centers, so effectively around about 70 megawatts of prime generation that is sat there, plus a lot of other infrastructure that goes to make that work effectively,” he said.

The current phase of the campus supports roughly 54 megawatts of computing capacity. That capacity could expand significantly as additional buildings are developed.

The power generators used in the microgrid are supplied by Wärtsilä, a Finnish company that manufactures and services power sources and other equipment for marine and energy markets.

China’s Chengdu Tecloman Energy Storage Technology Co Ltd supplies the Battery Energy Storage System (BESS) supporting the facility’s power management. The Pure DC Dublin facility also deploys solar panels to create an additional power supply.

Grid constraints

Whitfield said the microgrid strategy emerged after the company encountered difficulties securing a traditional electricity‑grid connection for the campus.

“We did have an application to the electricity grid. Back in the day, traditional data centers would have a utility connection and backup generators,” he said.

But the project faced the same grid‑capacity constraints that have slowed data‑center development across parts of Ireland in recent years.

Whitfield said that, like many other developers in Ireland at the time, Pure DC had its grid application deposit returned and had to rethink its approach. The company ultimately chose to pursue the microgrid plan instead.

Engineers then evaluated whether the national gas network could support a large on‑site generation facility powering the campus.

“The modeling that we did with Gas Networks Ireland in the early days showed that this facility of this scale, connected into the robust infrastructure in and around Dublin, would have little impact on the availability of gas services for the country,” he said.

Water‑free cooling

Whitfield said the campus also aims to reduce environmental impacts through its cooling technology.

“There’s a lot of press over recent times about the amount of water that goes into cooling a data center. All of Pure DC’s facilities across our portfolio use zero water in operation, so we use a closed‑loop cooling system,” he said.

“You fill it once and then it operates continuously without using any more water resources, and that’s a really important point for us.”

Whitfield said many U.S. data centers historically relied on cooling towers that consume significant volumes of water.

Whitfield, a Chartered Engineer with more than 25 years of experience in the data‑center sector and a former Chief Executive of RED Engineering Design, joined Pure DC in 2021. He leads the company’s strategic data‑center engineering and energy initiatives.

He said future phases of the campus could also support wider energy initiatives beyond powering servers.

The energy‑center design allows waste heat from generation systems to be exported for potential use in district‑heating networks or nearby developments.

“We’re working very closely with partners to bring and use this system as a catalyst for decarbonized heat for the wider community in this area,” he said.

He said facilities with integrated energy centers could eventually become part of broader energy ecosystems that combine digital infrastructure, renewable fuels, and local heating systems.

Such projects could enable hyperscale campuses to evolve from large energy consumers into infrastructure platforms supporting both digital services and regional energy transitions.

Pure Data Centres Group (Pure DC), a hyperscale cloud and artificial intelligence (AI) data-center developer and operator active across Europe and the Middle East, has launched Europe’s first large-scale data-center microgrid in Dublin, Ireland, testing how operators can keep AI expansion moving when public electricity networks cannot keep pace.

The 110-megawatt on-site system was developed with AVK, a United Kingdom-based provider of power and energy infrastructure for data centers and industrial facilities. It allows the campus to generate power during early development phases before full grid capacity becomes available, enabling Pure DC to add computing capacity without waiting for a traditional electricity connection timetable.

“The biggest barrier to deploying AI infrastructure in Europe today isn’t technology, it’s power,” said Gary Wojtaszek, executive chairman and interim chief executive officer, Pure DC. “This microgrid proves that even the most constrained markets can unlock new digital capacity.”

“The future of AI infrastructure will be built where energy and compute come together, and that’s exactly what we’re building at Pure,” he said.

The Dublin project turns a power constraint into an infrastructure strategy. Instead of treating energy as a utility service that arrives after a site is planned, Pure DC is bringing generation onto the campus itself and building around that capability.

That matters because AI workloads are changing the economics of the data-center sector. As operators race to support cloud and AI customers, access to land and fiber still matters, but power is becoming the gating factor for how quickly new capacity can be delivered.

Pure DC announced on March 10 the appointment of Gary Wojtaszek as executive chairman and interim chief executive officer, a leadership move the company said would guide its next phase of AI-driven expansion across Europe and the Middle East.

Wojtaszek previously led data-center operator CyrusOne through a period of rapid growth, helping scale the business into one of the world’s largest hyperscale platforms before its $15 billion acquisition by investment firms KKR and Global Infrastructure Partners.

Dame Dawn Childs DBE, FREng, who had served as chief executive officer since May 2023, is transitioning to the role of president of Pure DC as part of the leadership change.

Own power first

Pure DC announced the microgrid on March 11 during a media tour of its new data-center campus in Dublin. The project was presented as Europe’s first large-scale deployment of its kind, developed with AVK to support phased delivery of cloud and AI infrastructure while reducing pressure on Ireland’s national grid.

The system includes three interconnected energy centers, each capable of generating up to 30 megawatts of power. The first two centers are expected to be fully operational by the end of 2026, with a third to follow later.

In practical terms, the campus can draw on dispatchable on-site generation during its early development stages and then move toward a hybrid configuration as grid capacity becomes available. That structure is intended to support operational resilience while giving the company more flexibility over how quickly it can bring halls online.

Wojtaszek said the scale of the development made the project notable even by industry standards. He said the completed site would represent about $1.5 billion in infrastructure investment, while customers were expected to install another $3 billion to $4 billion in chips inside the facility.

“I had never seen a data center that was self-powered,” he said. “When I heard that this facility had developed its own generation system, I realized this was something really unique.”

He said the design effectively gives the campus its own power grid. Electricity is generated on-site to power the facility’s computers, and the digital output then flows outward as cloud and AI services.

That framing helps explain why the project holds significance beyond a single Dublin site. For years, data-center developers largely treated access to electricity as a prerequisite controlled by grid planners and utilities. In the AI era, some operators are moving to treat energy supply as a core component of their products.

Pure DC and AVK said the microgrid is designed to complement, rather than replace, a long-term grid connection. Over time, the campus is intended to operate in a hybrid mode that combines grid-supplied electricity with on-site energy infrastructure to improve resilience and flexibility.

The design includes a battery energy storage system (BESS) to help manage load fluctuations and improve response times. The system is intended to support future renewable-energy integration as the campus evolves.

The site also includes a combined heat and power capability. Infrastructure has been put in place for heat recovery and possible future connection to district heating networks, subject to third-party demand and regulatory approval.

Water use has also drawn scrutiny for large computing campuses. Planned water-management measures include rainwater harvesting and on-site treatment to reduce reliance on mains water for engine-related processes.

The system has also been engineered to accommodate gradual changes in fuel composition, including hydrogen blending. That does not remove the carbon debate around data-center growth, but it does give the site a pathway to adapt as gas networks decarbonize.

AVK Chief Executive Ben Pritchard said the project showed how carefully designed on-site energy systems could work alongside national planning frameworks. For Pure DC, the commercial case is equally straightforward: if power delays hold back data-center delivery, operators will increasingly look for ways to control more of the energy stack themselves.

The Mayor of Fingal County Council, Councillor Tom O’Leary, also welcomed the project, saying the microgrid demonstrates how new digital infrastructure can support Ireland’s energy transition while integrating renewable power, storage, and lower‑carbon fuels such as biomethane.

Power bottleneck

Wojtaszek used the Dublin site to make a broader point about the AI infrastructure race. He said the core constraint is no longer whether the industry can design advanced chips, software or cooling systems. The more immediate challenge is whether enough electricity can be brought to the right sites, at the right speed, to support deployment.

“Data centers consume an enormous amount of electricity,” he said. “Even a relatively small facility like this could consume the equivalent power of around 100,000 homes.”

He said that the level of demand inevitably raises questions for governments, utilities, and local communities, especially in markets already struggling with grid reinforcement, permitting, or renewable integration.

“When you consume that much electricity, it places huge demands on the grid and on power generation,” he said. “That naturally creates questions from communities and governments about how the system will support that growth.”

That helps explain why the company has framed the microgrid as more than a stopgap. In Wojtaszek’s view, self-generated campuses are likely to become a more common feature of the sector, particularly in supply-constrained markets where hyperscale and AI customers still want capacity.

He said such sites could also contribute to wider grid stability because they are built with substantial backup generation. In markets that experience weather shocks or temporary supply disruptions, operators may eventually be able to play a more active role in supporting resilience.

This reflects a wider shift in how data centers are perceived. Long treated as back-end infrastructure, they are increasingly becoming strategic assets tied to energy policy, industrial competitiveness, and national AI ambitions.

Beyond Dublin

Pure DC said it now has more than one gigawatt of data-center capacity live or under development and plans further expansion across Europe and the Middle East.

In a press release, Germany, the Netherlands, and the United Kingdom were identified as key targets for replicating similar microgrid deployments.

“We’re going to continue building data centers in FLAP-D (Frankfurt, London, Amsterdam, Paris and Dublin) markets,” he said. “At the same time, we are expanding into massive AI campuses across Europe.”

He said the company had launched its first major facility in Finland and was also looking at Spain, Portugal, Eastern Europe, and the Middle East for additional large-scale development.

The numbers underline how capital-intensive the sector is becoming. Wojtaszek said a single gigawatt of data-center capacity can represent roughly $10 billion to $15 billion of investment, a figure that rises further once customers fill those sites with high-end computing hardware.

He also argued that the pace of AI adoption is giving those investment decisions new urgency. Earlier technology waves, such as the internet and mobile phones, were transformative, he said, but AI has the potential to reach more deeply into business operations and daily life.

That broader thesis helps explain why companies are willing to experiment with new infrastructure models. If AI demand continues to rise while grid capacity remains tight, projects like Dublin may come to look less unusual and more like an early template for the next phase of the European data-center market.

For now, Pure DC’s campus illustrates the industry’s new logic: when electricity becomes the bottleneck, energy strategy moves to the center of the build.

Artificial intelligence (AI) can now write code, draft legal contracts and converse with striking fluency. What remains unresolved is how corporations will ultimately use these capabilities to tackle difficult human problems or simply to cut labour costs and boost short‑term profits. That dilemma sits at the heart of a growing debate over the future of AI deployment.

Apple’s Siri was originally conceived as an example of “humanistic AI,” a design approach aimed at augmenting human capability rather than automating people out of the workforce, according to its co‑founder Tom Gruber. As AI systems grow more powerful, Gruber warns that companies risk betraying that vision if they treat the technology primarily as a tool for eliminating jobs.

“The human job is context, understanding, and communication with humans,” he said in a recent event in Hong Kong. “The AI job should be doing a sub‑task that humans can define well, and especially can evaluate well.”

Gruber warned that poorly directed AI initiatives within companies often push middle managers toward the easiest cost‑cutting targets, typically staff reductions, rather than using the technology to create new value. He described that approach as a cowardly misuse of AI’s potential.

The universal interface

Gruber has long framed AI not as a replacement for humans but as a universal communication tool capable of expanding access to knowledge.

His interest in conversational computing began decades before modern large language models emerged. With a background spanning psychology and computer science, he began building a rudimentary language system in 1983 to help non‑verbal individuals, similar to the late physicist Stephen Hawking, generate speech.

“The idea that you could speak in your own language, with your own voice and skill... means it’s a universal interface,” he explained.

Gruber shared his views during a fireside chat at Technology for Change Asia 2026, where industry and government leaders gathered in Hong Kong to explore how emerging technologies could reshape business and society. The event organised by Economist Impact was held on March 11–12.

The session examined the evolution of AI assistants from the symbolic programming approaches of the 1980s to the massive neural networks that dominate today.

By removing the need for traditional computer literacy, voice‑based AI could empower populations that might otherwise be left behind.

Tom Standage, deputy editor of The Economist and moderator of the session, pointed to examples in India where farmers using basic feature phones access agricultural advice through voice‑driven AI tools without needing to read or type.

Gruber described this approach as a “big brain, small screen” architecture, allowing powerful AI models to operate behind simple interfaces and extend services such as healthcare guidance, education and conversational support to anyone with access to a mobile network.

Beyond the ‘super tool’ threat

The discussion also addressed the risks associated with the rapid spread of powerful generative AI tools.

Gruber warned that bad actors now possess what he called “super tools” capable of generating deepfakes, fraud campaigns and large‑scale misinformation. He said attempts to embed simple ethical rules into AI systems, similar to Isaac Asimov’s fictional laws of robotics, are unlikely to work with modern language models.

Standage added that mitigating those risks requires what he described as a “defense in depth, whole-of-society approach,” arguing that regulation should focus not only on AI systems themselves but also on restricting dangerous uses of their outputs.

Gruber said technological safeguards must be reinforced by legal protections.

“If you allow rampant use of data impersonating humans at scale by anyone, the trust basis for financial services and health care will collapse,” he warned, calling for enforceable laws against digital impersonation.

The personalised future

In a joint statement released on January 12, Apple and Google announced a multi‑year collaboration under which the next generation of Apple Foundation Models will be based on Google’s Gemini models and cloud technology.

The two technology giants said the models will power future Apple Intelligence features, including a more personalised Siri expected later this year, while continuing to run on Apple devices and Private Cloud Compute in line with Apple’s privacy standards.

Standage noted Apple’s decision to integrate Google’s Gemini model into Siri, likening the upgrade to a “brain transplant.”

Gruber said he was “very excited” about the development, predicting it could transform the assistant from a simple task manager into a far more capable conversational partner.

Looking ahead, he envisioned a future in which personal AI systems maintain extensive conversational histories with their users, enabling them to anticipate needs and provide highly personalised assistance.

Toward the end of the discussion, Gruber struck an optimistic note about the broader potential of human‑centered AI.

“Now came a pretty profound opportunity in front of us,” he said. “Now is the time to really build more space and solve those big problems you want solved.”

Artificial intelligence (AI) is set to define the next phase of transformation in the automotive industry as manufacturers move beyond electrification and focus on digital capabilities embedded in vehicles, production systems and mobility services.

The shift marks a transition from the first phase of the electric‑vehicle revolution to a new period in which software, AI models, and intelligent systems increasingly determine how vehicles are designed, manufactured, and operated.

“The automotive industry today is like a football match. The first half is electrification, and the second half is digitalization driven by AI,” said Bono Ge, country manager for the UK and Ireland at BYD. “This is what we see in China. Customers have a very strong demand for intelligent features and smart functions in their cars.”

Ge said AI applications extend beyond autonomous driving and include in‑car language models that allow vehicles to distinguish whether a driver is speaking to the car or to passengers. He said the system must understand context so that if a parent tells a child in the back seat to stop, the vehicle does not interpret the instruction as a command directed at the car.

AI systems are also being applied to energy management in electric vehicles. Charging routes and strategies can be optimized through AI models that analyze driving patterns and infrastructure availability.

The growing demand for these digital capabilities reflects changing consumer expectations, particularly in markets where electric vehicles have already reached significant adoption levels.

AI development cycles

Executives shared their perspectives during a panel discussion titled “Shifting gears with AI: the future of mobility” at Sustainability Week in London on March 2. The event was organized by Economist Impact. The session was moderated by Vijay Vaitheeswaran, global energy and climate innovation editor at The Economist.

Apart from BYD’s Ge, other speakers included Manuel Schneider, head of open innovation at BMW, and Karin Svensson, chief sustainability officer at Volvo Group.

Schneider said AI is already reshaping how vehicles are developed inside automotive companies, compressing the time required to move innovations from concept to production.

“We have been working on AI for about 15 years, so this is not new for BMW,” Schneider said. “What we see is that AI will speed up the process from the very first idea to seeing it implemented in the car.”

“It affects everything from deriving a business case, validating a product with customers, engineering and crash simulations, all the way to physical AI on the shop floor,” he said. “We will see significant speed improvements across the entire development process.”

AI‑driven engineering tools are transforming internal workflows in automotive companies. Digital simulation, predictive testing, and automated design processes enable engineers to evaluate concepts faster and reduce development costs.

Across the value chain

Beyond product development, AI is influencing nearly every stage of the mobility value chain, from vehicle design and factory operations to logistics and after‑sales services.

“We see AI as central to creating business value across the entire value chain,” said Svensson. “It affects everything from design and production to customer interfaces, operations and services.”

“The need for transportation is constantly increasing, and we need to make it cleaner, safer and more efficient,” she said. “AI can absolutely help increase sustainability, but we also need to think about sustainable AI.”

For commercial‑vehicle manufacturers such as Volvo Group, the technology is particularly important for monitoring vehicle performance and minimizing downtime.

“For our customers, uptime is extremely important,” Svensson said. “AI is a tool that helps us understand how to keep vehicles running on the road and operating efficiently.”

The adoption of intelligent systems also varies across regions. Executives noted that consumer behavior and market maturity can shape the pace at which drivers adopt AI‑enabled features.

Ge said Chinese consumers tend to adopt in‑car digital features more quickly, including voice control and other intelligent systems.

“In Europe, many customers still focus more on engineering aspects of the vehicle,” he said. “When we ask some customers whether they use voice control, they say they have never used it.”

Competing AI ecosystems

Automakers are increasingly relying on multiple AI ecosystems as they integrate large language models and machine‑learning systems into vehicles and development processes.

Schneider said BMW maintains a technology‑agnostic strategy that allows the company to use different AI models depending on the application.

“We follow a technology‑agnostic strategy,” he said. “We try to work with the best models for each use case. That might be Gemini, Anthropic, OpenAI, or DeepSeek in China.”

He added that innovation is advancing rapidly across markets, particularly in China, where developers are quickly adopting AI tools in software engineering and product development.

“We definitely see many advancements coming from China that are beneficial for the whole world,” Schneider said. “It would be wrong not to study them and make use of them.”

“We see massive speed in China,” he added. “There is a very natural way of engaging with AI when developers build software there.”

The automotive industry remains deeply dependent on global supply chains. This dependence makes it sensitive to geopolitical tensions and trade disputes.

Tariffs and geopolitics

BYD has seen strong sales growth in the United Kingdom, where imported electric vehicles are not subject to the additional anti‑subsidy tariffs imposed by the European Union. In October 2024, the EU introduced definitive duties on Chinese EVs ranging from about 7.8% to 35.3%, on top of the bloc’s standard 10% import tariff.

In an interview with TechJournal.uk, Ge said companies must remain aware of geopolitical risks that could disrupt supply networks.

“Tariffs are certainly one factor, but the main issue is whether customers understand and like the technology,” he said. “Our cars deliver good value for money, so tariffs have not been the most important factor.”

“As a company, we always need to be aware of geopolitical risks,” he said. “Supply chains and transportation can be affected, so we must watch those developments closely.”

“The automotive industry is global,” Ge added. “If something happens in China or in the Middle East, it can affect manufacturing in Europe.”

BYD’s expansion strategy reflects its vertically integrated technology model, which combines vehicle manufacturing with battery production, semiconductor development, and energy‑storage systems.

Ge said the company views this integration as a key advantage as vehicles become increasingly digital platforms driven by software and AI.

The Chinese manufacturer has already built a significant presence in the UK through electric buses and energy‑storage projects, which Ge said helped establish brand recognition before its passenger‑car expansion.

Since launching its bus business in Britain in 2015, BYD now operates about 2,800 electric buses there and has also deployed roughly 2 gigawatt‑hours of battery‑energy storage connected to the power grid, with another gigawatt‑hour under development. These projects, he said, demonstrate how mobility companies are evolving into broader providers of clean‑energy and digital‑technology services.

Future vehicle services

Looking ahead, automakers expect AI to open new revenue streams by enabling services that extend beyond traditional vehicle manufacturing.

Increasingly sophisticated sensors and onboard computing systems allow vehicles to monitor drivers, passengers, and mechanical systems in real time.

“We have many sensors in cars, and the question is what we do with that data,” Schneider said. “Can we measure heart‑rate variability or blood pressure and provide medical services inside the vehicle?”

He said such concepts remain experimental but illustrate how automotive companies are exploring new partnerships and digital services.

“This is still a small pilot project, but it shows how the automotive ecosystem is opening up to new business models,” he said. “Car manufacturers are connecting to new partners and technologies.”

Battery analytics represent another emerging application for AI systems in electric vehicles.

Svensson said AI could help increase sustainability in transport, but added that companies must also consider what she called “sustainable AI.”

“We also use AI models to monitor the battery,” Ge said. “We look at the state of charge and the state of health to understand how the battery performs over time.”

As digital capabilities expand, AI will increasingly determine how vehicles interact with drivers, infrastructure, and other vehicles, turning automobiles into connected software platforms rather than purely mechanical machines.