Tony Blair Institute: UK needs bit barns to lead in AI deployment, not training

3 months ago 2

Britain should not try to compete with America and China in the race to build cutting-edge AI models and focus instead on widespread AI adoption, but even this will require a boost local compute capacity.

This is the message from the Tony Blair Institute for Global Change, which warns that the UK needs to "take infrastructure seriously" to remain competitive in the AI era, and fix systemic problems that cannot be resolved overnight.

This will require time, investment, and for the government to make building datacenters a national strategic priority, says the former PM's non-profit organization.

In its report, Blair's think tank says Brtain lacks the resources to keep up with the US, China and the Gulf States, which are all pumping hundreds of billions of dollars into vast, energy-hungry datacenters for AI training. The UK doesn't have the money, the available land or the energy resources to follow suit, it adds.

Instead, the focus must be on deploying and widely adopting AI, "demonstrating to the world how to effectively apply it across sectors including health, education, government, defence and science." This is where the economic gains will be found, the report says, in seeking ways to bolster productivity, improving public services and driving innovation across the economy. The jury is out on if or how this can be achieved.

Following this strategy will still require considerable investment in AI infrastructure, a point the UK government already acknowledged in its AI Opportunities Action Plan unveiled at the start of the year, plus the Compute Roadmap detailed recently.

The Tony Blair Institute says these steps are not enough, and claims "the situation is now dire" as the UK has placed AI at the centre of its growth and security goals, yet lacks sufficient infrastructure to carry it off. At the current build pace, the country is unlikely to meet its 2030 target of 6 GW of AI-ready capacity on UK soil, it says, and blames planning and permitting delays, constraints with the national grid and soaring industrial energy costs for holding up progress.

This is backed up by a separate report from fDi Intelligence (part of the Financial Times) which claims the UK could face up to a 5 GW shortfall in the datacenter compute capacity it needs. This is based on an analysis commissioned by the Department for Science, Innovation & Technology (DSIT) and assumes supply will nearly double from 1.8GW today to 3.3GW, while there is expected to be demand for between 5.1 GW and 8.5 GW.

Britain will struggle to meet the 6 GW target, as most of the current 1.8G W of bit barns are concentrated around London and not optimized for AI.

The Tony Blair Institute recommends the government follow a strategy of "accelerated diversification," which means quickly building out resilient infrastructure. This will require it to make changes to attract new investment in a way that reduces risk, spreads capability regionally to improve resilience and supports a strong domestic ecosystem – easier said than done.

Other recommendations are that the government must ensure the National Energy System Operator (NESO) integrates bit barn demand (as estimated by DSIT) into national plans and builds to support this in "dynamic updates."

However, the AI Energy Council, formed last year, was tasked with ensuring this, but it is operating under a traditional British government cloak of secrecy, declining to reveal what has been discussed or any decisions made during the two meetings it has so far held.

The report instead recommends a team of AI and datacenter experts be formed within NESO to support demand planning and accelerate AI integration in the energy system.

Unsurprisingly, it also recommends amending the planning process to issue decisions within an eight-month period, and using ministerial call-in powers for high-investment datacenter projects and grid investments.

These are actions the government has already embarked upon, designating these sprawling server farms as critical national infrastructure (CNI), which allows developers to override any local opposition to datacenters being built in an area. Categorizing them as Nationally Significant Infrastructure Projects (NSIPs) allows developers to apply to the central Planning Inspectorate (PINS) for planning permission, bypassing the local authority altogether.

Other noteworthy recommendations are that the government should adopt a strategy to develop a series of new gigawatt nuclear power station projects, as well as reforming the way nuclear is regulated in the UK both to expedite building while also reducing costs.

The Tony Blair Institute report also suggests some actions similar to those already taken by the Trump administration, such as changing the rules to allow co-location of AI datacenters with energy generation sources, plus identifying government land that could be used for bit barns and offering this to private developers.

All of these suggestions sound like a big ask for any UK government, which are characterized by sclerotic indecision and glacial decision making at the best of times. The current administration is currently struggling with chronic budget constraints and the urgent need to invest in a range of other areas as well.

The Tony Blair Institute concludes with a warning that if the UK fails to build, it will fall behind. If it gets it right, it has a real opportunity to lead in AI, by gaining the expertise in how to apply it and build the right infrastructure to support it.

Despite the calls to action, there are growing concerns that all this investment in AI might turn out to be inflating a hype bubble. A report from McKinsey & Company identifed widespread unease because nobody is really sure what the level of AI demand is going to be in future, while other research found that generative AI has had no significant impact on earnings or recorded hours in any occupation so far, despite the billions poured into building and training the models.

The accuracy of agentic AI recently fell under the spotlight when Gartner forecast that 40 percent of projects will be ditched by the end of 2027 due to rising costs, unclear business value or insufficient risk controls.

Last year, Baidu chief Robin Li described the AI sector as being in an "inevitable bubble," similar to the dot-com bubble at the end of the 1990s, while a report from Lenovo found that only 4 of 33 AI proof-of-concept projects it surveyed actually progressed into production, equating to an 88 percent failure rate.

Still, the UK government and the Tony Blair Institute remain unperturbed by the generative AI naysayers. ®

Read Entire Article