The UK’s £400 billion AI estimate mostly came from Google

For nearly two years, the UK government cited an estimate that AI could add £400 billion to the country’s economy. Last week, AI Minister Kanishka Narayan told Parliament that the figure came from a report commissioned by Google, while the government’s own estimate was significantly lower: £55 billion to £140 billion by 2030, with a central forecast of £120 billion. Narayan also called the new estimate highly uncertain.

The UK’s estimate of AI’s economic contribution was cut by more than threefold

Where the two estimates came from

The £400 billion figure appeared in a publicly available 2023 report prepared by British consultancy Public First, now known as PF Global. Responding to questions from WIRED, PF Global’s head of economics, Jonathan Dupont, said the forecasts differed because the government, in the company’s view, assumes AI will be adopted more slowly.

Dupont noted that the technology is changing rapidly, so the assumptions behind models have to be revisited. He said the government’s forecast assumes a slower adoption rate than seemed possible when PF Global built its model in 2023.

Government departments and agencies used the £400 billion estimate for nearly two years under two different governments. Ministers had a lower forecast by at least November 2025, yet four months later, the government agency Skills England, which oversees apprenticeship programs, was still publishing the earlier figure.

Narayan’s response is the first confirmation WIRED has found that the government compared the two estimates and identified who commissioned the report. However, the minister did not withdraw the £400 billion figure or say that it was wrong.

Questions about the use of forecasts

Dame Chi Onwurah, a Labour MP and chair of the parliamentary Science, Innovation and Technology Committee, told WIRED that the government is entitled to talk about AI’s potential to improve public services. But, she said, inflated expectations are becoming an obstacle to digital modernisation: overstating the benefits of technology can undermine the case for change. She urged the government to assess AI’s capabilities realistically and be open with the public about potential savings.

PF Global said that more centralised AI adoption and integrating it into the civil service’s workflows could speed up the technology’s spread. The company said it had not changed its estimate of AI’s overall potential, but that with slower adoption, its central forecast for 2030 was now much closer to the government’s. PF Global expects growth to accelerate in the years that follow.

Lord Holmes of Richmond, a Conservative member of the House of Lords, asked ministers back in January 2025 where the £400 billion estimate came from, but received no answer. He told WIRED that the government had repeated the figure many times, other organisations had followed suit, and now the estimate had to be substantially revised. In his view, the government should have explained this sooner and more clearly. Holmes added that reliable data matters because decisions about spending billions of pounds of public money depend on it.

The dispute over £120 billion and £400 billion has become part of a broader debate about where the line falls between government support for technology and overly optimistic promises. Former technology minister Peter Kyle also cited the £400 billion estimate while promoting AI’s benefits, while many MPs outside the government raised questions about the technology’s risks.

Sumedha Deshmukh of the Ada Lovelace Institute called the revision to the government’s forecasts a positive step, but said she was concerned that it came nearly two years after ministers began citing research paid for by Google. Seemingly precise estimates, she said, often rest on assumptions and interests, and companies frequently fund or influence research.

Deshmukh urged the government to check figures before using them to shape policy, fund independent research and measure whether promises are being met. Before relying on forecasts of economic growth or productivity gains from AI, departments should find out where an estimate came from, what assumptions it rests on and who stands to benefit.

WIRED requested comment from the Cabinet Office, the department responsible for AI in government, and from Google. PF Global said it continues to work for Google, but that all its research is independent and the company develops its own findings and assumptions.

Dataist’s take: The gap between the forecasts shows how much estimates of AI’s economic impact depend on assumptions about the pace of adoption. In this case, the government used a figure from research commissioned by Google for nearly two years, even though it later had a substantially lower forecast; the earlier figure was never officially withdrawn.