Compute infrastructure and the AI Opportunities Action Plan

Compute infrastructure and the AI Opportunities Action Plan

The widely anticipated AI Opportunities Action Plan, published 13 January 2025, has the bold ambition to place Britain at the centre of the AI revolution.

techUK welcomed the ambitions of the AI Action Plan - and follow-up expression of interest - to boost the UK's AI capabilities, delivering positive effects across our society and economy. The Plan was also the first time this government has signalled how it expects to approach the issue of computing - or 'compute' - after the early announcement to pause funding for the previously announced Exascale programme. The Plan signalled that the government would produce a further compute strategy by the Spring to fill notable gaps, including how the UK expects to lever existing capacity, navigate semiconductor supply planning, and whether it will develop a roadmap for exascale. In this insight we explore how to make the current government's AI plan succeed as this policy develops - and why success requires moving at pace.

The journey so far

The AI Opportunities Action Plan states that, to build a strong foundation for the AI revolution, Britain needs sufficient, secure and sustainable AI compute. Compute in this context refers to high-performance computing power (also referred as large-scale compute) to train AI models and to run 'inference' (applying AI to draw conclusions from data).

It is worth noting that the AI Action Plan is not the first time UK Government has tried to define its approach to high-performance computing power. In 2018, for example, UK Research and Innovation (UKRI) developed the first national research and innovation infrastructure roadmap - noted as the first time UK Government made a commitment to build an exascale computer.

Furthermore, in 2022, the independent Future of Compute Review developed ten comprehensive recommendations that called for investment into skills, international partnerships, software development and more. This review is significant for two reasons: First, it correctly viewed compute infrastructure holistically, attempting to address long term, systemic challenges facing businesses and researchers; Second, it underscored explicitly the significant shortages of publicly owned compute with recommendations to achieve an exascale system and the need for compute infrastructure tailored specifically for AI in the UK. This paper led to the commitment to AI Research Resource (AIRR) with powerful supercomputers located in Bristol (Isambard AI) and Cambridge (Dawn). At the time, the government also committed the building of exascale compute, which was later paused.

The connection between AI and compute was underscored in the more recent AI Opportunities Action Plan which states that, to build a strong foundation for the AI revolution, Britain needs sufficient, secure and sustainable AI compute. Or, in other words, there can be no AI revolution without the compute infrastructure on which this is built.

What does the AI Action Plan say about Compute?

Whilst the AI Opportunities Action Plan ties compute and AI together explicitly, it is more reserved in its commitment to build publicly owned compute to meet the demand of AI than previous government policy.

It accurately recognises that compute infrastructure should be provided though different modalities, including both public and private access, including through the cloud, and international and sovereign capability. This is not necessarily a step change from the Future of Compute Review, which recognises the necessity for different modalities of access, but is likely to take a different position on the extent to which it should foot the bill through publicly owned compute. It is worth acknowledging that, given the changes in the economic and fiscal situation in the UK, Government is likely to focus on financing compute where it is, in their opinion, absolutely necessary to unlock growth across the UK.

There are a total of six recommendations related to compute within the AI Opportunities Action plan. These include a commitment to set out, within six months, a long-term plan for the UK's AI infrastructure needs backed by a 10-year investment commitment. While techUK welcomes this commitment, it is not the first time such a plan has been created, only to be shelved (Afterall, the multi-year Future of Compute Review lasted just two years before being replaced by this new approach.)

The lack of action and inconsistency is a source of frustration for industry. Indeed, it is understandable why some within the tech industry are puzzled by further drawn-out timescales to develop a plan when there was one in place. Even if the timelines are met for the new investment plan, it will still be an entire year since there has been policy on compute - and a year from when the building of the publicly owned exascale machine was paused. Building compute infrastructure takes time. If the private sector is expected to shoulder the largest share of the financial burden of the compute needs of the country the government must recognise that. These delays mean the UK risks missing this crucial opportunity to lead not just in AI innovation, but technological innovation more broadly.

What about the ambition for publicly owned compute? There is a call to increase the UK's compute capacity by at least 20 times by 2030, but there is no mention at all of the planned exascale programme. And while the UK is, set to double the capacity of the national AI Research Resource this does not seem to be based on a deeper understanding of the UK's compute needs including for activities outside of AI. This is particularly important when you consider that compute is intertwined not just with AI, but also integral for the UK's quantum ambitions. techUK believes that the future Compute Strategy must be grounded in and reflect the compute needs of all the priority sectors of the economy and the technological applications that will be used by them to drive growth and release productivity gains.

Finally, with the recognition that the UK cannot provide all compute through publicly owned compute, it also advocated for international compute partnerships with like-minded countries to increase the types of compute capability available to researchers and catalyse research collaborations.

What is needed - a holistic approach to compute

The proposed ten-year investment plan must consider emerging compute technologies, investment in software, skills, and wider high-performance computing capabilities to complement AI. The Government should not miss the opportunity to further emphasise these areas of compute policy that can fall under the radar in the 'hype' of AI. This is particularly important in research, where not all pioneering compute intensive research requires or is suited to the infrastructure preferred for AI. The UK should not ignore the research communities that have made the UK world leader in scientific and medical research when prioritising AI infrastructure. However, in a review dedicated exclusively to AI compute, it is difficult to understand how the needs of this community will be addressed.

Furthermore, the plan should explore the interplay between compute and other emerging technologies, including Quantum. Quantum computing is well suited to help tackle some of the leading challenges compute is trying to solve, from weather forecasting to drug discovery, molecular dynamics, and more. In many use cases, hybrid application development will be the key for tackling problems at scale, which provide the benefits of both classical computation as well as quantum computing technologies. This integration between classical compute and quantum is already being explored by pioneering work at the UKRI's Hartree Centre, but this is not being co-ordinated at a national level, and without government ambition. In contrast, the modular design of Europe's JUPITER system is preparing for what the future of compute will look like beyond AI. This highlights the danger of viewing compute only as a foundational tool for AI, limiting experimentation in other areas. As the European leader in quantum, it is surprising that the UK is not pushing forward in hybrid quantum systems.

A focus on access

With constrained budgets, the UK must leverage current resources to their maximum potential. This means removing significant and cumbersome barriers to access to compute resources, so that compute is accessible to all users across academia and industry. One initiative underway to democratise compute access is the AIRRFED demonstrator project, funded under the UK ExCALIBUR programme. It aims to address several challenges in the realm of federated AI research, particularly regarding access, interoperability, and resource management, through developing a modernised software stack. For example, it simplifies access to large-scale computing for both industry and academia through a single sign-on system, making infrastructure appear as private cloud and paving the way for much needed AI services for the public and private sector.

This also raises the need for cloud to be recognised as a suitable way to access both public and private compute. Cloud access has the potential to make compute more accessible to a broad range of users, with workflows able to run between commercial cloud, private cloud and on-premise compute. Indeed, many cloud providers already have the scalable infrastructure and expertise needed to rapidly expand the UK's AI compute capacity. Working in partnership with commercial cloud providers to improve access, for example through aggregating compute demand across research institutions and startups to achieve economies of scale, along with targeted programs to help organisations better utilise cloud resources, would remove barriers to compute adoption. Furthermore, the government could explore using credits for private compute via the cloud as a way to supplement access to publicly owned compute infrastructure which is already overstretched.

The underpinning software designed specifically for high-performance compute environments will also be essential to realising the benefits of investment. The UK has strengths, including the ExCALIBUR programme launched in 2019 to develop exascale software and algorithms.1 This £47.5m investment from UKRI demonstrates that compute is not just buying semiconductors or building data centres, it is about building the mechanisms to enable access and next generation research.

Recognition of supply chain vulnerabilities

There is no mention of the vulnerabilities of the semiconductor supply chain as it relates to growing compute in the UK. Semiconductors are critical to high-performance compute as the "chips" on which all compute is run. Yet semiconductor production is the end product of a complex, specialised, global and capital-intensive sector, that requires several businesses working within their own specialisms. This global supply chain has significant susceptibility to external shocks: the semiconductor supply shortage in 2020 was estimated to have reduced global GDP growth by 1%. Alongside economic insecurity, supply chain disruption can also threaten critical national infrastructure, such as power and transport networks, making the semiconductor supply chain essential to UK resilience. There needs to be greater co-ordination to reflect and respond to geopolitical instability and its potential effect on access to the chips needed for compute.

A long tern plan that includes exascale

As for exascale, the shelving of the next generation of compute has left several questions unanswered. techUK believes that any long-term roadmap, such as the proposed ten-year funding plan on compute should include a strategy for achieving exascale capabilities. However, whether it will be included remains to be seen.

It is time to act, and at pace. techUK and our members stand ready to work with the government to deliver this plan and ensure the UK seizes the opportunities presented by AI technology, placing the necessary compute infrastructure at the centre. We strongly recommend early industry involvement in compute strategy, rather than simply reviewing a proposal, to fully understand what is possible with the resources available.  We look forward to showcasing the benefits of compute both with, and beyond, AI technologies, while encouraging UK Government to finally set out its stance on whether the future of compute means exascale.

techUK - Seizing the AI Opportunity

The UK is a global leader in AI innovation, development and adoption.

The economic growth and productivity gain that AI can unlock is vast, but to fully harness this transformative opportunity, immediate action is required. Our aim is to ensure the UK seizes the opportunities presented by AI technology and continues to be a world leader in AI development.

Get involved: techUK runs a busy calendar of activities including events, reports, and insights to demonstrate some of the most significant AI opportunities for the UK. Our AI Hub is where you will find details of all upcoming activity. We also send a monthly AI newsletter which you can subscribe to here.

Upcoming AI events

Latest news and insights

Subscribe to our AI newsletter

AI updates

Sign-up to our monthly newsletter to get the latest updates and opportunities from our AI and Data Analytics Programme straight to your inbox.

Contact the team

Visit our AI Hub - the home of all our AI content: