In January, the UK government unveiled its AI Opportunities Action Plan. The Prime Minister, Sir Keir Starmer, predicted that the UK would become an AI superpower.

The UK once was an AI leader, but no longer. The government’s 50-point plan for reversing the decline sounds plausible at a high level and aims at commendable goals if they could be delivered. Unfortunately, I don’t think they can.

Many of the plan’s recommendations are ill-conceived. For instance, spending over-stretched public money on computing power – what the plan calls “sovereign AI compute” – is the wrong priority when there is no national competitive advantage in doing so.

Then there is the idea of converting information held in our public services into unique datasets that can feed AI’s voracious appetite for training material. The decades-long NHS efforts show how difficult it is to integrate disparate data sources.

The recommendation to attract AI high-flyers through “an internal headhunting capability on a par with top AI firms” is unlikely to be credible. Furthermore, will funding just 15 sector AI specialists make a meaningful difference when all industries have enormous AI opportunities?

The government plans to lead the way with its own use of AI for public services, envisaging a nimble, technically expert, integrated approach across public services. We cannot wait for that improbable outcome.

A financially realistic plan must recognise that there will be conflicting priorities vying for scarce public money. This week it’s increasing defense spending, last week it was the NHS, and the week before it was house building.

Taking a more pragmatic approach

The plan must be cheap, impactful, and easy. Investing in AI research at our top universities and encouraging industry collaboration is an obvious solution.

Oxford, Cambridge, UCL and Imperial all feature reasonably high in AI innovation rankings (though US and Chinese universities dominate). There is a strong argument for putting greater AI investment into these four institutions. (Spreading investment around the country and to more institutions would be desirable, but doing so would limit the impact.) These AI departments are already strong and have existing money from government, the universities’ own funds, and donors - for example Larry Ellison’s pledge to invest £130m in Oxford.

What form should this new investment take?

First, we need to pay researchers more. A senior professor in the US earns more than double their UK equivalent. Correcting this will break pay structures and irritate other academics, but if we can’t withstand that criticism, how will we make other tough decisions? Having elite professors also brings their labs, their funding, their commercial relationships, their prestige, and the opportunity for their university and the UK to participate in commercialising their innovations.

Second, the government should go further than its current plan to make it easy for skilled AI researchers to immigrate by introducing a tailored immigration category for AI researchers with faster approvals for spouses and children, shortening the indefinite leave to remain requirement for these academics, and offering seamless transition to an AI Entrepreneur visa.

Third, a lack of adequate computing power limits UK university AI research. Relatively small amounts would be needed compared with the national computing power the government envisages. In any case, Chinese AI firm DeepSeek recently showed how to use clever software engineering to reduce the cost of training large language models compared to the brute force, money-no-object approach of OpenAI and others. This is just the kind of thing our universities should be good at, and it requires only millions, not billions, of pounds of computing power.

Elite AI university departments, not underdogs, attract private investment. Big tech firms like Microsoft, Amazon, Google and Meta fill their development labs with the best post-graduates and post-doctoral students and want to build relationships with them, for example by funding PhDs. They build their research centres near the best universities.

Startups emerge near their founders’ universities, taking advantage of the local AI ecosystems of talent, funding, infrastructure and collaborative networks. The universities can expand their support of innovation and entrepreneurs emerging from their labs.

Overall then, the government needs to do what is cheap, easy and impactful and invest in the elite AI university departments from which overall AI growth will come. It is not chance that the businesses that the government is so proud of – DeepMind, ARM, and Wayve – were all founded by British academics.

Sometimes the answer is not that complicated.

Andrew Rashbass is the Founder and Chairman of ScultureAI and is the former CEO of Euromoney Institutional Investor PLC, Reuters, and The Economist Group.