There's a building in Loughton, Essex, United Kingdom, with room for 23,040 of Nvidia's high-end GB300 chips. It would be the biggest AI datacenter in the country.

It was meant to switch on early next year. Bu now it can't. They want to switch it on but apparently, there's no power.

The grid company, UK Power Networks, told the owner it can't supply enough electricity until the early to mid-2030s.

Imagine that. You build one of the biggest compute factories in the world, and the country it’s in says nah we can’t get power to it… for eight years.

Kind of says something about the state of the U.K. as well.

Anyway, Microsoft wasn’t very interested in paying for a very large paperweight, so as you’d expect, they’ve walked away from the £2bn project.

The owner of this AI factory is Nscale, a British “neocloud” that builds AI data centers and rents out the computing power.

It spun out of Arkon Energy, a crypto miner, in 2024, and now it’s looking to list in New York with a valuation around $35 billion.

The question is, with Microsoft walking away from a mega AI factory, when Nscale lists, is it an IPO to get on board with, or one to avoid at all costs?

Nscale managed $140.6 million in revenue in the first half of 2026 and yet managed to lose $1.02 billion in the process.

Yuck. Sounds like 2021 all over.

That's more than $7 lost for every $1 of sales.

Its contracts add up to $103.4 billion, yet only $2.6 billion is tied to machines already running.

Its flagship site in Norway is six months late, with 6.4 of a planned 30 megawatts live.

Its filing even states management had “substantial doubt” it could continue operations as forecast funding included reliance on uncommitted debt and equity financing.

In other words, it was counting on money to survive that it hadn't yet raised.

Then, its biggest listed rival, CoreWeave (Nasdaq: CRWV), lost a quarter of its value just last week.

The fears around CoreWeave are that it’s a massive debt bomb about to explode and expose the entire “neocloud” market as a house of cards, and everything will come crashing down, “going to zero”.

In addition, research from Gartner says more than 80% of neoclouds won't survive as independent businesses by 2030.

Plenty of good reasons to avoid the sector.

But I think that would be a mistake. Here’s why…

Nscale's own filing knew power was always going to be a problem. The whole sector knows power is a problem.

“As demand for AI compute has accelerated, access to reliable, large-scale contiguous power has emerged as the primary gating factor for AI infrastructure deployment.”

The Essex project just proved them right.

And while Microsoft might be out, reports are now that Anthropic is lining up Nscale to buy $45 billion worth of compute from them.

Microsoft walks away here, and Anthropic walks in the door with an even bigger order over there.

This is the volatile (and fickle) nature of the current compute market wrapped up all inside of a week.

Worth noting, Anthropic already had a $45 billion deal with Nscale, from August to rent compute from their Monarch campus. This is said to be another on top of that, a second Monarch building

And apparently Anthropic has dibs on a third.

Microsoft hasn't said why it left, and it hasn't stopped buying.

The big question the bears keep trumpeting on about is that demand is peaking. But the truth is, demand is still accelerating.

Anthropic can’t get enough compute. SpaceXAI can’t get enough compute. OpenAI can’t get enough compute. Microsoft might have walked away from Essex, but it will just get it elsewhere, it’s not like that demand disappears.

Meta, Google, Amazon, they’re all still jostling to get more compute than the market and the neoclouds can supply.

My view is that Nscale is the real deal, and as long as it can build, someone will buy what it builds.

And the beaten-down neoclouds, from CoreWeave to IREN to Nebius, look like bargains to me too.

For most of them to have peeled off a quarter of value in a week because bears strangled the market, is just a great long-term buy for me. They could easily flip back and double, or triple, in a matter of weeks from here.

Of course, power is still a massive story here to figure out, but when you see an opportunity like the neoclouds have given this week, and with Nscale’s IPO coming at us pretty soon, it’s a segment I like, and I think should be a part of an investor’s portfolio.

More soon,
Lewis Maguire