Data, Compute, Action!
Our reaction to the AI Opportunities Action Plan
We believe that the UK can produce globally significant AI companies – if the Government gets the foundations right.
Britain already has the right ingredients to lead. The research excellence of our universities is a significant draw for AI talent. We boast the third largest AI startup ecosystem globally. And we’re home to some incredible AI-powered breakthroughs. Take DeepMind, whose AlphaFold can predict the structure of “all of life’s molecules.” While this puts us ahead of many others, seizing AI’s full potential will demand even bolder ambition.
We welcome the publication of the AI Opportunities Action Plan – led by entrepreneur and ARIA Chair Matt Clifford – with its 50 recommendations that will be taken forward by the Government. We hope that its implementation will contribute to unleashing the potential of British innovators. In this piece, we analyse some of the Action Plan’s key recommendations, and what they might mean for entrepreneurs.
20-fold Vision
The money needed to build and operate AI systems is staggering, especially when considering the hidden costs of experimentation and failure, which often exceed publicly reported figures.
Fittingly, a cornerstone of the Action Plan is sovereign AI compute, calling for a twenty-fold boost in the AI Research Resource (AIRR) capacity. The AIRR was introduced in 2023 as national supercomputing infrastructure for AI research. Sovereignty, in this case, means that the terms and conditions of compute access can be set independently from hyperscalers – i.e. major cloud computing companies, such as Amazon Web Services. This enables the Government to act strategically, catering to the needs of academia, the public sector, and resource-constrained startups.
The Action Plan takes the AIRR on a strategic shift. First, the AIRR will now operate under Missions-focused programme directors who will oversee compute allocation. In the past, some argued that investment in the AIRR lacked direction. Under the new arrangements, it’s not hard to foresee questions on whether this approach would allow for innovation outside Mission boundaries. While the Missions will be an important framing for making a compelling case, our view is that they are broad enough to not act as a constraint in practice.
Second, AIRR’s scope expands beyond just compute resources. The Plan recommends, “the AIRR should evolve into a set of Mission-oriented clusters that bring together compute, data, and talent to pursue frontier AI research and other national priorities.” This move brings the UK’s approach closer to international counterparts like the National AI Research Resource in the US and EU’s AI Factories. While the implementation details for “clusters” are pending, this could take the form of an exciting repository of data, AI models, validation methods and more, which could create a community, accessible to both a public and private user base.
Further on data, the Action Plan sets out a series of recommendations to ensure the National Data Library provides valuable datasets for scientists and companies. To make this happen, the Plan recommends running public calls to receive proposals for new datasets of high potential social and economic value. The Action Plan also suggests creating a copyright-cleared British media asset training dataset, and reforming the UK’s text-and-data mining regime so that it is at least as competitive as in the EU, while also enabling rights holders to have control over the use of content they produce.
Electric Dreams
Access to inexpensive and reliable power is critical to energy-intensive AI work. A critical component of the Plan is the introduction of AI Growth Zones to accelerate data centre development. But the current power grid’s limitations could potentially be a bottleneck.
The Plan called to address AI power demands. In response, the Government will establish a new AI Energy Council, co-chaired by the Science and Energy Secretaries. Working directly with companies, the Council will help map out and address AI’s future energy demands – feeding into the Government’s broader ambition to become a clean energy superpower, including through technologies like Small Modular Reactors. This approach could provide a sustainable and reliable energy solution independent of the existing grid infrastructure.
Last year, we published a report detailing how nuclear power is uniquely positioned to provide the clean electricity a growing AI sector will need. We set out a plan of our own to speed up the construction of new reactors, including a recommendation for reforming the siting regime to enable the deployment of Small Modular Reactors as a way to help meet future AI energy demands in a way that aligns with the UK’s climate commitments.
Britain Needs Talent
The Plan places strong emphasis on making the UK the “natural home for elite talent.” Central to this strategy is the creation of a new AI scholarship programme, investing in higher education and other routes to AI skills, and attracting top international talent through new and existing visa routes.
One way in which the immigration system could be leveraged is by improving the High Potential Individual (HPI) visa route. We’ve proposed expanding its eligibility criteria to include more of the world’s top universities. Right now, the HPI route is short of realising its own potential in attracting top entrepreneurial talent. For example, it excludes graduates of Indian Institutes of Technology (IITs) – renowned Indian universities whose alumni list leaders like Sundar Pichai, the CEO of Google, Vinod Khosla, founder of Khosla Ventures and co-founder of Sun Microsystems, and Parag Agrawal, former CEO of Twitter.
Talent in Government matters too. Having innovative thinkers embedded in Government as a norm increases the chance of something exceptional emerging from within. On this, the Action Plan proposes hiring routes for senior technical talent into the civil service with salaries benchmarked to 75% of private-sector equivalents.
Prompt Procurement
The Plan proposes a series of improvements to Government procurement processes to ensure it remains agile to a rapidly changing AI ecosystem. Changes to procurement contract terms and the implementation of the Competitive Flexible Procedures, in particular, promise to accelerate engagement with AI companies. The proposed reforms, coupled with approaches that facilitate repeated technology acquisition across different Governmental bodies, could substantially improve the public sector’s ability to adopt and implement AI solutions while fostering innovation in the private sector. As we argued, the aim should be a flexible procurement approach that is equally open to startups and big established players, avoids vendor lock-in and enables continuous innovation.
Looking Ahead
While the Plan’s endorsement from the Prime Minister shows that its political backing starts at the very top of Government, the true test will be in its execution. The combination of infrastructure development, sustained funding commitments, talent attraction, and changes to procurement will create a solid foundation for future growth – albeit will require confronting thorny future political battles.
Perhaps what’s particularly promising about the next steps is that Matt Clifford will stay on as the Prime Minister’s Adviser on AI Opportunities. Innovative ideas conceived with great ambition often risk being watered down to something rather conventional as they pass through the policymaking machine. This move signals the Government’s seriousness about carrying the original vision forward.
But the responsibility is also on the AI sector itself to demonstrate how it can deliver on the Government’s aims. The success of the Action Plan depends on sustained political support. Ministers want something to feel positive about and to champion, so articulating the opportunity is important – and this will be our task.