Andy Burnham won’t give the Green Party its pause on data centres. But Ofgem will sell them one instead, at up to £712,500 a megawatt, refundable to anyone rich enough not to need it back.
Data centre doomerism threats
On 9 September a Labour MP asked Andy Burnham to back a moratorium on new data centres. “At this point, I won’t go as far as a moratorium,” the Prime Minister told the Commons. The same day Verdant, a think tank launched in March and linked to the Green Party, published the economic case for one. A week later Ofgem’s consultation on its data centre commitment fee closed, and the regulator will decide by the end of the year whether Britain wants Dublin’s hangover without ever having been to the party.
Those first three words are carrying the Government’s entire data centre policy. As drafted, Ofgem’s fee is a moratorium on anyone who can’t lodge tens of millions of pounds before a spade goes in, with a de facto exemption for balance sheets big enough not to notice.
The Verdant briefing, The Jobs Mirage of Data Centres, is the work of James Meadway, who runs the think tank with Deborah Doane. He was John McDonnell’s economic adviser in the Corbyn years, and in 2014 he was writing for Counterfire, the 2010 breakaway from the Socialist Workers Party. He has since left Labour for the Greens, and in May he stood for Tower Hamlets council, finishing fourth in Bromley North, a ward that elects two. Former Green MP, Caroline Lucas, also sits on Verdant’s board.
Read the CV as a warning label. He is at least candid about what he wants: a pause on consent for large data centres until an establishment-led assessment framework is in place, with demonstrated electricity and water supply as a binding condition of consent, and a public consultation on how much computing capacity Britain needs and where.
His summary adds “a social values test” for the technology. The briefing promises four recommendations and prints three. The fourth failed the social values test.
He never endorses Ofgem’s fee, and he has no need to because his evidence is Ofgem’s. He takes the regulator’s queue, 73 gigawatts of data centre applications against a national peak demand of about 45, multiplies it by Ofgem’s own benchmark of £9.5m of capital per megawatt, and announces £693bn of investment, around 23 per cent of GDP, that could never be built. Nobody said it would be. Five of the briefing’s footnotes point to Ofgem. Ofgem should ask itself what the consultation is actually for.
On jobs he has a point. The industry’s figure of 40,000 new jobs was always a stretch, and nobody in my trade underwrites a data centre on its headcount. Verdant’s own sample finds about 1.2 permanent jobs per megawatt. A kettle doesn’t employ anyone either, and nobody wants it banned.
Jobs per megawatts
The yardstick Verdant offers instead is worse. Ranked by jobs per megawatt, schools and universities score 4,534, hospitals 3,626, and steelworks 291. So a school is a better use of the grid than a steelworks by nearly sixteen to one. The best use of a megawatt, on this measure, is to switch it off and hire people to fan the servers.
The metric captures labour intensity, which in any other argument goes by the name of low productivity, and even the Government called it deeply misleading. We don’t judge a reservoir by its staff.
Which brings us back to the fee, because Meadway’s pause on planning consent, and Ofgem’s deposit, both end up in the same place. Nothing happens.
The deposit
Ofgem wants large data centres (the threshold is 40MW) to secure between 2.5 and 7.5 per cent of that £9.5m benchmark. That’s about £237,500 to £712,500 a megawatt.
In practice, that’s an unconditional payment from the day Ofgem accept a grid offer until the day the power is switched on. For a 500MW campus that is up to £356m.
It comes back at energisation. It is forfeited if the project leaves the queue, and also if it fails to progress according to the rules, a far wider trapdoor than the press release suggests. The level sits in NESO’s licence, where it can be changed after you’ve committed.
The benchmark gives the game away. The fee is pegged to the cost of the whole campus, when the thing being rationed is a connection. On Ofgem’s own table, the security that covers the network’s exposure averages 0.2 per cent of project capital. The fee’s floor is twelve times that. Market pricing puts the works for a 100MW connection at about £25m. So Ofgem’s floor roughly equals what the connection costs to build, while the ceiling is nearly three times as high.
Then read what counts as security: a letter of credit or performance bond from a Qualified Bank, a performance bond from a Qualified Company, or cold hard cash. The first two are gated by credit rating, so an unrated development company gets its letter of credit only by handing the bank the cash. Yet the cheap route exists only for balance sheets that never needed it, which makes the security regime itself the wealth test. Sadly for Meadway, the only people that can afford his pause are the tech giants he distrusts the most.
It reaches retrospectively, too. Existing offer holders get three months to post security once the rules land, two if their agreement is unsigned, and walking away after that costs the full fee. And cutting your reserved capacity is charged the same way. Hand back half of a 200MW position, precisely the behaviour this reform exists to produce, and you pay up to £71m on the megawatts you surrender. Even the wheel clampers wait until you’ve parked.
Catherine Tate’s Nan would have answered the consultation in four words. One of them unprintable. And hers would have been the most accurate response Ofgem could have received. But the joke really is on us, because half the powered land market is a Catherine Tate character. Take Lauren, who has 300MW reserved on a twelve-acre dog-legged paddock. Told that her first of its kind on-site hydrogen fuel cell wind turbine has no approval, she wants to know if her face looks bovvered. Or Derek, who has consent for a scheme no data centre operator would build, and when asked for a letter of intent from an off-taker, replies how very dare you.
And Nan herself, who has done none of the work, stands between the buyer and the power, and wants half the upside for the privilege of being taken along for the ride. Ofgem didn’t dream this fee up on its own. It spent two years in meetings with Nan. Then sent the bill to everyone who just looks like her.
What Europe charges
On 1 April Germany switched to a queue that ranks projects by maturity. Applicants of 100MW or more pay a €50,000 application fee and, on accepting an offer, a deposit of €1,500 a megawatt, credited against their construction contribution and forfeited only if the project fails for reasons within their control. Meanwhile, Ofgem’s floor is roughly 180 times the German deposit and its ceiling more than 500 times.
France’s fast track asks hyperscalers for a 10 per cent bank guarantee, staged payments of the real connection costs, and a building permit within 16 months or the capacity goes back.
Ireland never priced its queue, just simply closed it. And Dublin got a moratorium in all but name. That is the menu facing every regulator in Europe: curate the queue, price it, or ration it. Ofgem named its programme Curate and then designed an instrument that looks like pricing but works like rationing. That takes a certain talent.
The concession
The underlying problem is real. Contracted demand in the queue went from 41GW to 125GW in under a year, with data centres claiming 73GW across 315 projects. In my opinion, most of that queue is fiction and will never be built, and I say that as someone in the market. It is sadly littered with parties that suffer from never having done a billion pound development with a grid connection before.
This summer I reviewed a large site on the strength of its data centre potential. Was the grid offer still live? No confirmed answer. Was there an energisation date? Nope. Had the capacity in the brochure ever been formally quoted? Barely even estimated. Had there been any correspondence with the network operator in two years? No. My last question ran to one line: is anything on the power secured? The marketing process closed the following week regardless.
That is the problem of speculation. But note where it lives: in the land market, priced by respectable agents and filtered by exactly one mechanism, a buyer who checks. Eleanor Warburton, the Ofgem director behind the fee, is right that consumers shouldn’t carry the risk of other people’s fantasies. The question is whether her price is too high for the people that aren’t fantasising.
A test of wealth
To its credit, the consultation also proposes milestones testing financial capability, commercial maturity, and procurement, which is the German instinct. The money is another matter. Of 120 responses to Ofgem’s call for input, 18 backed a flat upfront fee and 67 preferred staged increases or incremental release, designs that tie cash to progress. Ofgem picked the room’s least favourite for “administrative simplicity”, while its own scoring concedes that posted security weakens a project’s incentive to withdraw. So, a fee meant to flush projects out of the queue encourages them, on its designer’s own admission, to stay in it. Madness.
Milestones test whether you’re real. A flat deposit per megawatt tests whether you’re rich. Ofgem barely disputes this. Question 10 asked respondents to confirm its assumption that speculative projects “will not be able to secure such a fee”, and twelve pages later the document concedes that tests based on credit ratings “favour larger or more established developers”. It is rare to watch a regulator mark its own homework and fail it.
Taking a powered site from contract to shovel ready costs under £200,000 a megawatt, so the deposit for joining the queue costs more than the work, and at the top of the range more than three and a half times as much. The only counterparties who can tie up £356m without flinching are the hyperscalers and sovereign wealth funds.
They arrive last, signing leases on sites somebody else found, optioned, consented, and powered. The casualty is development equity: the specialists who carry British farmland to a live connection, holding options that give them nothing a bank will lend against. Ofgem wants their cash sitting idle in its queue. It won’t, which is exactly what the consultation intended.
Sell to whom?
Paragraph 4.25 of the consultation says the fee will bear on developers whose strategy is “powered land banking”, and expects it to push them to build through to energisation or “sell the project on to a developer who will minimise their exposure to the fee”. Sell to whom? To the only buyers the fee leaves standing.
Slaughter and May, a firm not given to hysteria, predicted in August that Ofgem’s curation would become “an outright cull”. It will be a cull of the domestic industry, handing the field to foreign balance sheets, courtesy of a government whose AI minister talks about sovereignty.
Nor should ministers take those balance sheets on faith. Nathan Tankus argues the big tech borrowers are lightly geared cash machines; Ed Zitron argues too many of their commitments sit off the balance sheet. Either way Ofgem loses. If Tankus is right, £356m deters nobody it was aimed at and prices out Britain’s own developers.
If Zitron is right, the fee hands the queue to fragile money, because a deep pocket isn’t always a sound one. The week Ofgem published, Situational Awareness, the hedge fund named after the AI boom’s founding manifesto, took margin calls on a book geared about four to one and sold the bulk of its public holdings to Citadel at a discount. A wealth test either fails to filter or filters for the wrong thing.
And the market already runs a commitment test, priced daily. The same European parcel is worth three to seven million euros as dirt and north of sixty million with a secured, dated connection, though nothing else about the land has changed. Unverified power gets dirt value without a regulator lifting a finger.
The test Ofgem wants already exists. It’s called the land price. Something no politician, policymaker or regulator seems to understand.
The tenant trap
Worse than the fee is the tenant milestone. Ofgem’s provisional position requires a leased or sold data centre to produce a binding compute contract, covering at least 20 per cent of capacity for a term of at least a year, by the sixth milestone. On its own that is a modest ask. The trap sits earlier. The first tenant document falls due six months after the connection agreement is signed, which on Ofgem’s own timetable is before a planning application need even be made. Nothing in the drafting says a contract conditional on energisation counts as binding, and conditional is the only kind this market signs.
A hyperscaler building for itself clears the test with a Certificate of Intended Self Usage, a document about itself, signed by itself, while anyone else must produce a counterparty. Fall behind and you lose your place and your deposit.
Meadway’s briefing cites the procurement test as a way of telling whether a project is real. Making developers buy long-lead items like transformers to prove their sincerity turns paper speculation into steel speculation, which a man supposedly worried about scarce resources might have bothered to notice.
Here is how a British site gets built. You buy or option the land, win the consent, secure the connection, and then go to market, because a hyperscaler signs when a site is real. Powered land lenders want sight of the tenant and the power staged before money flows.
The tenant is how the developer exits, and when it arrives enabled or ready to build it will happily carry real grid costs: a £30m substation, or £40m where reinforcement is needed. What it won’t do is sign at the front of a queue.
Ofgem’s consultation records respondents warning of exactly this circularity. It notes the objection and proceeds regardless. Now add Verdant’s second recommendation, proven power and water as a condition of consent, and the loop closes: no consent without power, no power without a tenant, no tenant without consent. A deposit taxes the early bird. The tenant rule, with Verdant’s amendment, shoots it, worm or no worm.
One foot on each pedal
This is the same state that designated data centres critical national infrastructure, published an AI Opportunities Action Plan, invented AI Growth Zones, opened the NSIP regime to data centres in January, and used the Planning and Infrastructure Act to strip out the statutory duty to consult before applying. Every one of those reforms says faster. Then the regulator arrives with a screeching handbrake.
In July the Prime Minister abolished DSIT, the department that ran the growth zones, and at the end of that month officials still couldn’t say who would run them instead. OpenAI paused its Stargate UK plans in April, citing energy costs and regulation. Scotland has gone past weighing a freeze: in September Holyrood voted to take no decisions on data centres over 50MW for up to a year, a moratorium in all but name, and the Greens who asked for one abstained. Classic. And in July, New York froze new data centres for a year. Investors read all of it. And money, unlike a data centre, doesn’t need planning permission to leave.
Southall
Meadway’s briefing blames data centres for crowding housing off the grid and cites his own borough: Tower Hamlets. It warned last year that housebuilding at scale could stall for a decade for want of electricity. That constraint is real. But its cause is a grid nobody built, and neither a pause nor a deposit requirement gets a substation built in Poplar.
Southall shows the order of events. PropViews’ own Nick Cuff told the story this summer of the old Quaker Oats site, where Honey Monster Puffs were made for eighty years. Consent existed for 1,997 homes, on condition that the developer first built a loss-making creative hub, presumably for all the artists who can afford Southall, and on the applicant’s own appraisal the scheme sat at minus 46 per cent profit before Covid.
The factory came down in 2022 and the site sat empty until CyrusOne arrived with a hyperscale data centre and £17.8m for Ealing. The comforting reading is that server halls are eating London’s housing.
The accurate one is that housing was already dead, killed by the cumulative weight of individually reasonable conditions. The data centre was just the last viable use standing on the corpse.
Before Ofgem decides
Britain didn’t destroy housebuilding with one bad law. It did it by degrees, a review here and an affordable housing levy there, a condition on a condition, each defensible in a planning committee no working man or woman can attend.
All this however, is lethal in the round: a bureaucratic lasagne, layered with good intentions. The data centre sector is now being served the same dish, face down on the floor.
The fix is simple. Tie security to the cost of the works, as France does. Release it in stages as milestones are met, which is what 67 of the 120 respondents asked for.
Count a tenant contract conditional on energisation as binding and move the first tenant document to after consent. Forfeit only where failure is within the developer’s control, as Germany does. None of that lets a speculator squat on capacity, and all of it keeps the people who find and build Britain’s sites in the queue.
The Prime Minister says he won’t go as far as a moratorium, at this point. He doesn’t need to.
Britain has built the Nan-ny state’s final boss: it does none of the actual work, puts itself between the site and the power, and wants £712,500 a megawatt posted for the privilege. Ofgem calls it curation. Everyone in powered land calls it grounds for nanger management.
