We went looking for the jobs in Britain's AI build-out. At one company the answer filed at Companies House is Nil.

Every data centre proposed in this country arrives with an argument attached. Investment, growth, and jobs. The jobs are usually the part said last and remembered longest, because they are the part that means something to the town it lands in.

We decided to check.

Not by asking anyone, and not by reading a press release. By going to the audited accounts that these companies are required by law to file, and taking two numbers out of the same document: the electricity they use, and the people they employ. Both are compulsory disclosures for a large company. Both are free for anyone to read.

At Ark Data Centres it worked exactly as it should. One hundred and thirty nine people. Two hundred and eighty thousand megawatt hours. You can hold both facts in one hand.

At Vantage Data Centers UK Limited the answer, filed on the seventh of this month, is that the company used five hundred and five million kilowatt hours of British electricity and employed nobody. Nil, and Nil the year before.

The people are next door, in a sister company, four hundred and fourteen of them. That is not a trick and I am not going to pretend it is one. Separating the company that owns a building from the company that staffs it is ordinary, legal and older than any of this technology. Nothing has been concealed. Every figure in this edition came from documents the companies filed themselves.

But look at what it means. When the next site is proposed and somebody stands up and says it will bring jobs, the accounts of the company holding the electricity cannot be used to test that claim. Not because anyone is hiding, because of how the boxes are arranged. And the sister company's own filings say its staff serve the wider European business, so we cannot even tell you how many of the four hundred and fourteen are here.

I have been going back and forth on how strongly to put this, and I have landed on this. The problem is not the companies. They have complied. The problem is that we have built a national argument about employment on top of a disclosure regime that was designed to measure carbon, and it does not answer the question we are actually asking.

That is a failure of the people who write the rules, and it belongs to all of them. Five AI Growth Zones have been named. The government's own delivery page says the first cohort will bring twenty eight point two billion pounds of investment and over fifteen thousand jobs, with more than five thousand of those in the North East and three thousand four hundred in Lanarkshire. An electricity discount is due to follow from April 2027.

So a number does exist. It is a forecast, published by the people asking us to accept the trade. What does not exist is any requirement on anyone to report, in a form a reader could check, how many jobs actually turned up. The promise is specific to the nearest hundred. The outturn is unmeasurable. That is true of this government, of the last one, and of the opposition of either.

I will say what I think this is really about, since a note is the place for it.

We are being asked to accept a trade. Britain gives up something genuinely scarce, firm power in the right place at the right time, and gets something back. That may well be a good trade. I am not against these buildings and this paper has never argued that we should not have them.

But you cannot judge a trade if only one side of it is measured. We can tell you to the kilowatt hour what goes in. What comes out, in the only currency this paper has ever cared about, is a number nobody is required to publish.

One last thing, because it would be dishonest to leave it out. Ines tried seven companies and could only read two, because the other five filed their accounts as pictures rather than as figures a machine can open. So this is two companies and not an industry, and you will see the desks say so repeatedly today. That is not modesty. It is the difference between what we found and what we would like to have found, and I would rather run a small true thing than a large one we cannot stand behind.

We will keep pulling this thread. The forecast is on the record. The outturn is not, and until it is, fifteen thousand is a hope with a decimal point on it.

— M.

This note is mine: the view, and the call to run it. It begins as a draft, drawn from work the AI and I have researched and argued out together, the same way every desk in this paper is made, and I answer for every line because I read every line. Those desks run on models built by Anthropic, one of the labs sitting on the very scoreboard we report, so we tell you plainly: we cover this from inside it.