The Goldilocks Moment That Wasn't

Nscale just pulled off the largest Series C in European history. £1.6 billion ($2 billion) at a £14.6 billion valuation. Former Meta COO Sheryl Sandberg and ex-Deputy PM Nick Clegg on the board. Jensen Huang calling them a "national champion."

Problems is, according to a damning Guardian investigation published last week, their flagship UK supercomputer site is still being used as a scaffolding yard.

I built something like this in 2009. Well, not quite this — I was trying to convince estate agents to buy SMS marketing packages, not governments to believe in phantom data centres. But the pattern's identical: big promises, bigger valuations, and reality lagging somewhere behind a construction site in Essex.

When £2.5bn Becomes an "Intention"

Here's where it gets properly British. The government, when pressed about Nscale's £2.5 billion commitment to build "one of the most powerful AI computing centres ever built" in Loughton, quietly admitted it wasn't "a formal contract, rather an intention to commit capital."

An intention.

I've seen this positioning before. It's like when we used to tell clients our SMS platform would "revolutionise customer engagement" when we were three blokes in a WeWork with a Twilio API. The difference is we weren't asking for billions or appointing former cabinet ministers to our board.

The site visit tells the real story. Guardian journalists found the promised supercomputer location still occupied by scaffolding equipment. The facility that's supposed to be operational by end of 2026 — hosting tens of thousands of Nvidia GPUs delivering 90 megawatts of compute — remains very much a storage yard.

The CoreWeave Problem

It's not just Nscale. The investigation revealed that CoreWeave's £1 billion "investment" to build "two new data centres" actually involved renting space in facilities built in 2002 and 2015. Rather than building anything new, they'd simply leased existing capacity from Google and Fujitsu.

That's not investment. That's tenancy with a press release.

CoreWeave's getting hammered in court too — a securities class action alleging they concealed operational failures while overstating scaling capabilities. Their Q4 2025 results were brutal: £452 million loss versus £49 million expected. Share price down 40% over six months.

This is what happens when positioning doesn't match product-market reality.

The Sandberg Factor

Sandberg's appointment is fascinating. She told the New York Times she sees similarities between Nscale founder Josh Payne and Zuckerberg — both young entrepreneurs she's backing for transformational growth.

But there's a crucial difference. When Sandberg joined Facebook in 2008, they had 70 million active users and revenue approaching $300 million. The product worked. The scaling challenge was operational, not foundational.

Nscale's challenge isn't scaling — it's building. Their flagship project exists primarily in PowerPoints and government announcements. That's not a scaling problem; that's a delivery problem.

The Infrastructure Reality Check

The AI infrastructure race is real. Gartner projects £2.52 trillion in global AI spending through 2027. But infrastructure isn't software — you can't iterate your way out of a planning permission.

DCDC Real Estate director Jonathan Pow told the Guardian that CoreWeave's promised 1GW renewable energy project in Scotland is "total pie-in-the-sky." The site currently has 24MW capacity — less than 3% of what's promised.

This is the gap between Silicon Valley thinking and British planning reality. You can't agile your way through Epping Forest District Council.

The Early Investor Goldmine

Here's the bit that'll make you wince. Nscale shares were allocated at 1p in October. After the latest funding round at £14.6 billion valuation, that's a 350,000% return for early investors.

Three hundred and fifty thousand percent.

I've seen this movie. Bitcoin mining companies pivoting to AI infrastructure, spectacular valuations, and early investors cashing out before the concrete's even poured. It's crypto 2017 with better branding.

Government by Press Release

The Department for Science, Innovation and Technology confirmed it's "not playing an active role in auditing these commitments." Investment figures come from the companies themselves. No verification mechanism exists.

This is how you get phantom infrastructure. Starmer wants Britain to be "an AI maker, not an AI taker," but making requires building, not announcing.

The Verdict

Nscale's funding is genuine. The investor list is heavyweight. The board appointments are serious. But infrastructure is ultimately about steel and silicon, not strategy and storytelling.

Clegg brings regulatory expertise. Sandberg brings operational scaling experience. But neither can turn a scaffolding yard into a supercomputer by Christmas.

The UK's AI ambitions deserve better than phantom commitments and intention capitalism. Real infrastructure takes time, planning, and unglamorous persistence — qualities that don't translate well to funding decks or government photo ops.

Until Nscale can show actual servers in actual buildings processing actual workloads, they're just another very expensive intention. And at £14.6 billion, we're paying premium prices for budget delivery.