Who Captures the Value of Artificial Intelligence?

I’ve been thinking about where the money will go in an AI economy if most of the hyperscalers are based in the US or China. My initial intuition was that it could become unsustainable for countries that do not control their own frontier AI - huge amounts of money could flow overseas, while the countries hosting the leading AI companies capture both the profits and the tax revenue. Where does that leave the rest of the world?

Imagine a future where AI can perform a large proportion of the work currently done by people. Employment could fall substantially and governments would need to replace lost tax revenue. They may also need to pay for an extensive income support scheme or even a universal basic income.

For the United States (and maybe China), it seems straightforward. The government could simply tax the profits of AI companies, which presumably are going to be enormous eventually.

But what about the rest of the world?

If British businesses and public services are paying American companies to use their AI, some of the economic rent is flowing overseas. We cannot simply tax OpenAI’s worldwide profits as though it were a British company.

There could be enormous savings in the public sector. Take a simplified example of the NHS. Its annual spending is roughly £230 billion. If AI eventually produced something approaching a 30% productivity gain, the theoretical value of that improvement would be around £69 billion a year.

Even an intensive NHS-wide deployment of advanced generative AI is only likely to cost in the low single-digit billions. If Anthropic received £2 billion of that spending, that would be £2 billion of revenue flowing to an American company in return for technology potentially creating tens of billions of pounds of value in Britain.

Even if Anthropic eventually made a very high 30% profit margin on that business, the profit attributable to the NHS contract would be around £600 million. At a US corporation tax rate of 21%, that would equate to roughly £126 million of US federal corporation tax, before allowing for the much more complicated reality of international tax arrangements.

So Britain might spend £2 billion overseas and indirectly contribute something like £126 million to the US Treasury, while receiving perhaps tens of billions of pounds of additional productive capacity at home.

The same principle applies across the economy. A British engineering firm might pay an American AI company £1 million a year but use it to generate £10 million of additional output. Most of that £10 million does not automatically disappear to America. It could be expected to appear in Britain as profits, wages, investment and, other taxable economic activity.

The rewards of effectively utilising AI seem like they might outweigh the benefits of owning AI to a greater extent than I had intuitively assumed.

Obviously, domestic AI companies at or near the frontier would be a huge advantage. We’ve seen from recent trade restrictions following the launch of Fable that depending on overseas AI could have significant drawbacks.

But we do not necessarily need to produce the world’s dominant AI models to benefit from them. We need British organisations to become exceptionally good at using them.

For a country like the UK, the priority might be to adopt AI rapidly and effectively across both the economy, rather than focus on building out sovereign capability. At the same time, we should continue integrating AI in areas where we already have national strengths, such as financial services, biosciences and aerospace.