‘Ambition and delivery are not aligned’ – Starmer’s AI Action Plan, one year on
Exactly a year ago, Keir Starmer, with the help of Matt Clifford, promised to “mainline AI into the veins” of the UK’s economy.
The so-called AI Opportunities Action Plan, a 50-point blueprint unveiled in January 2025, was meant to turn that rhetoric into results, placing the UK on equal footings with the likes of China or the US, as a global AI superpower.
A year later, the money seems to have arrived but the more concerning question is whether the plumbing is keeping up.
New data from the Startup Coalition’s AI index shows the UK’s top AI firms have raised over £20bn in private capital, commanding a cumulative valuation north of £45bn.
Business services alone account for about 400 AI startups, pulling in £8.3bn, while financial services and healthtech companies have driven unprecedented growth, proving Britain’s growing strength in sector-specific AI.
Vinous AIi, deputy executive director of the Startup Coalition, dubbed it as evidence of a maturing ecosystem.
“The UK is already a leader in the application layer”, he said. “With such a clear competitive advantage, the government needs to double down on where we are already winning”.
Private capital lands ahead of the state
For all the billions pouring in from venture funds and hyperscalers, public backing remains strikingly thin.
Of the £20.2bn raised by leading AI firms, just £456.5m, which is barely 2.2 per cent, came from government.
But other European countries like France have lent heavily into state co-investment, using public capital to de-risk scale-up technologies domestically.
Britain’s approach has relied instead on market momentum, a gamble that private confidence alone will do the heavy lifting.
The gap shows up early, with university spin-outs, a traditional strength of UK innovation, still giving away an average equity stake of 16.1 per cent to their parent institutions.
That is a significant improvement on a decade ago, but remains high by international standards, which risks deterring later-stage investors.
The Startup Coalition is now urging ministers to consider a tiered equity cap to stop promising research dying on the vine.
Tax, too, is emerging as a quiet brake on scale, and the Coalition has proposed a distribution-based corporation tax regime, under which firms pay zero corporation tax on reinvested profits and are taxed only when cash is distributed.
Infrastructure – the real bottleneck
If funding forms one fault line, compute is the definite other, and potentially the most high risk of the two.
Data centre applications hit record highs in 2025, with over 60 new proposals filed across England and Wales, CityAM revealed.
Investors are circling powered land, old hotels and former coal mines alike, in a gold-rush scramble to secure compute capacity for AI workloads.
Yet grid access remains painfully slow. Developers routinely face connection delays of eight to ten years, particularly around London and the South East.
“Ambition and delivery are not yet aligned,” says Ben Pritchard, chief executive of data centre power supplier AVK. “Growth has been held back largely by constraints around power availability.”
The government’s flagship solution, in the form of so-called AI Growth Zones, is moving, albeit slowly.
Of the four zones announced so far, only one has begun ground preparation, with most unlikely to deliver meaningful capacity before 2026.
That lag matters in an industry where investment decisions are global and timelines are unforgiving.
Stuart Abbott, UK managing director at VAST Data, warns against what he calls an “AI sugar rush”.
“If the UK wants this to be durable rather than a one-year spike, it has to treat AI infrastructure like economic infrastructure,” he says. “That means energy, storage, data pipelines and skills”.
Regulation grey zone
On regulation, the government has largely stuck to its principle-led approach.
The Information Commissioner’s Office (ICO) has committed to a statutory AI code of practice, while Ofcom and the ICO continue to issue guidance rather than blanket rules.
It is a deliberate contrast to the EU’s heavier-handed AI Act.
For now, business largely welcomes the flexibility, but copyright issues remain unresolved.
Creative industries are pushing back hard against proposals to loosen data-scraping rules for model training, while ministers insist no final decisions have been taken.
Measured purely in pounds invested, the AI Opportunities Action Plan has succeeded, and Britain remains Europe’s most attractive AI market, buoyed by world-class research, deep capital pools and a growing application ecosystem.
Yet grid reform, tax incentives, spin-out rules and compute access remain as significant constraints.