TL;DR

The headline £400 billion that ministers attached to AI came from a 2023 study paid for by Google, AI minister Kanishka Narayan said in a written answer on 26 October, which WIRED reported on 4 November. The newer government estimate is £55 billion to £140 billion of added UK gross value added in 2030, with £120 billion as the central figure. Anyone who has put the larger number in a business case or board paper should now check where it came from.

What changed

Dr Ben Spencer MP had asked what evidence lay behind the £400 billion in the government’s AI action plan. Narayan’s reply credits a Google-commissioned report from Public First, the consultancy WIRED says is now called PF Global, and calls even the newer estimates “hugely uncertain”. It neither withdraws the £400 billion nor calls it wrong. WIRED says it is the first place it could find both figures set against each other.

The timeline is the awkward part. WIRED reports that ministers had the lower estimate from at least November 2025, while Skills England was still publishing £400 billion four months later. Lord Holmes of Richmond says he asked about its origins in January 2025 without getting a response.

PF Global defends the work, saying its research is independent and that the 2023 model assumed faster adoption than has happened. Its central estimate for 2030 is now, it says, “much closer to the Government’s number”.

Why the source matters

Science committee chair Dame Chi Onwurah named “hype” as a barrier to digital transformation of public services. The Ada Lovelace Institute’s Sumedha Deshmukh said the government should ask where a number came from, what it assumes and who benefits.

Our own coverage has repeated the claim: a March article on AI apprenticeships described £400 billion as a government projection. The government’s own release gave the figure at the time without naming its origin. In our view, it shows how far a commissioned estimate can travel once a department adopts it.

The timing is pointed, too. On 4 November the British Chambers of Commerce published its own modelling of more than 10% on GDP by 2031, built on scenarios researched by Anthropic.

Looking Forward

WIRED says it approached the Cabinet Office and Google for comment. Spencer has since asked how the central figure was derived. Narayan’s 4 November reply points to an earlier answer, which links a government method note applying adoption curves to OECD productivity estimates. That note never states £120 billion: its worked medium scenario comes to £117 billion. In our view, the practical lesson for UK leaders is simple: treat any AI growth figure as a claim with an author and a set of assumptions, and find both before repeating it. The lower number is still an estimate, and by the minister’s own description an uncertain one.