UK cuts its 2030 AI impact forecast from £400bn to £120bn

When a startup deck or corporate strategy says that “AI will add hundreds of billions to the economy,” almost nobody asks where the number came from. The UK has just shown why that question is worth asking.

What happened

The UK government now expects AI to add between £55bn and £140bn to the country's economy by 2030. Its central estimate is £120bn. That is roughly one-third of the £400bn figure cited in the government's AI Opportunities Action Plan.

AI minister Kanishka Narayan presented both figures in a written response to Parliament. He also confirmed an important detail: the report underpinning the £400bn figure had been commissioned by Google.

In other words, a key benchmark for the national AI strategy rested on research funded by one of the AI market's largest players. The government is now relying on its own, far more modest estimate.

Why this matters beyond the UK

Imagine choosing a cloud provider and reading a performance comparison commissioned by that same provider. You would not necessarily dismiss it as false, but you would account for who commissioned it and verify it against your own data. Macroeconomic forecasts deserve the same treatment.

A sponsor does not necessarily manipulate the calculations. But it influences the questions, assumptions and scenarios. An optimistic assumption such as “AI will be adopted rapidly everywhere” can easily triple the final figure without creating any formal mathematical error.

A gap of more than threefold shows how sensitive such estimates are to their starting assumptions. Even the new range is broad: the £140bn upper bound is two and a half times the £55bn lower bound. This is honest uncertainty, not a precise prediction.

There is also a practical impact. Figures in government plans shape grants, tax incentives, procurement and regulation. An inflated benchmark can steer resources in one direction, while a revision changes the expectations of markets and investors.

What developers, product teams and founders should do

First, do not put someone else's macro forecast into a pitch without checking its source. If an investor discovers that your “£400bn market” came from a vendor-sponsored report, confidence in your other numbers will fall.

Second, build estimates from the bottom up. How many teams will realistically buy your product, what budgets do they have, and what share of their workload do you address? This model may be less exciting, but it is defensible.

Third, use a range rather than a single point estimate. The UK government itself provides an example: £55–140bn, with a midpoint of £120bn, says more than one polished number.

Fourth, ask three questions whenever you read a report. Who commissioned it? What assumptions does it make about the speed of adoption? What happens to the result if those assumptions are cut in half?

Limitations

The available material does not explain the methodology behind the new calculation: the assumptions underlying the £120bn figure and how they differ from the previous ones remain unknown. It is therefore impossible to claim that the new estimate is more accurate. It is merely independent of the vendor sponsor and more cautious.

The main lesson is not about the UK or Google. Every polished number about AI's future is a hypothesis with an author and a sponsor. If you are building a product, budget or pitch around it, identify both first.