A London company valued its own research at nothing. Two years of British tax credits on that research came to 17.1 million pounds

When a company stops believing in its own future, it has to write the number down. That is what an impairment is, and BenevolentAI's 2024 accounts contain one.

Goodwill of 23.5 million pounds, carried unchanged since the start of 2023, went to nil.

Behind that single line sits a calculation worth reading slowly, because it is management's own assessment of what the business was worth if it carried on.

Under the accounting rules a company tests an asset two ways and takes the higher of the two. The first is value in use: what the thing is worth if you run it. The accounts describe how that was built, from the portfolio of drug development assets, from the collaboration agreements, and from the platform costs going forward. Then, following a strategic overhaul announced on 11 December 2024, they record that management assessed an immaterial value in use.

Immaterial. Not reduced. Not cut by a percentage. The future value of the drug pipeline, the partnerships and the machine that was built to find the medicines came out at a number too small to matter.

The second test is what somebody would pay for the company. That came out higher, so that is the figure the accounts had to use.

Read that again, because it is the whole story in one line. The price of the listed shell was worth more than the discounted value of running the business inside it.

The number was 24.5 million pounds at the end of 2024. A year earlier the same calculation had given 114.2 million. The shares closed 2024 at 31 euro cents against 1.08 euros twelve months before.

Now the part that lands in Britain.

The same accounts record a UK research and development tax credit of 7.3 million pounds for 2024, and 9.8 million for 2023. That is 17.1 million pounds across two years, recognised as money due from HM Revenue and Customs. The 2024 figure was calculated at 14.5 per cent, up from 10 per cent, and the accounts state the group is expected to meet the tax authority's criteria for the research intensive rate.

So one set of accounts holds both things at once. A company expecting to qualify for a tax rate reserved for research intensive businesses, and the same company recording that the future value of its research was immaterial.

That is not a scandal and this desk is not going to dress it up as one. The research credit exists precisely because research fails. If it only rewarded work that came off it would not be a subsidy, it would be a prize. A company can claim it honestly, spend it properly, and still end up with nothing to show. That is the scheme working as designed rather than failing.

The gap is somewhere else.

Britain publishes what the research credit costs. It does not publish what the credit buys in work. There is no figure anywhere for how many people were employed on the research those 17.1 million pounds supported, how many of them are still employed, or where the rest went. Nobody is required to say, and nobody does.

We were only able to count the people at this company because its listing collapsed, and the exemption that had kept its British accounts thin fell away with it. Two hundred and seventy-seven researchers in 2023. A hundred and seventy in 2024.

That count exists by accident. It exists for one company. And it exists because the company died.