This kind of growth headline is the stuff that makes people lose their minds—and that’s exactly why I don’t fully trust it.
A 684% jump in sales sounds like proof that the AI wave is “real,” that anyone touching data centers is basically printing money. And sure, maybe they are. But when a number is that explosive, my first question isn’t “how high can it go?” It’s “what was the starting point, and what does the company have to do to keep this going?”
Here’s the clean version of what’s being reported: Nebius Group, a cloud computing provider based in Amsterdam, says its first quarter sales hit $399 million, up 684%. The driver is demand for its data centers. It’s also expanding hard in the US with a new gigawatt-scale “AI factory” in Missouri and a second site in Pennsylvania. And it picked up the core team from Clarifai to boost what it can do.
That’s a lot of movement at once. Revenue spike, major buildouts, and an acqui-hire style talent grab. When a company does all of that in the same breath, it’s usually because the market is rewarding speed more than it’s rewarding stability.
I get why people cheer this. AI needs compute. Compute needs data centers. Data centers need power, land, chips, network gear, and a ton of money. If Nebius is in the right place at the right time, this is exactly what a breakout looks like.
But I don’t think the story here is “AI demand is huge.” We already know that. The real story is that the business model is turning into an arms race where everyone is forced to build big before they fully know what steady demand looks like.
Imagine you’re a startup building an AI product. Your biggest fear isn’t just competitors—it’s getting cut off from compute when you finally get traction. So you sign up with whoever can promise capacity and fast access. If Nebius can say “we’re building more, we can handle your growth,” that’s a strong pitch. And it can pull revenue forward fast.
Now imagine you’re a bigger company. You don’t want to be the executive who “missed AI,” so you lock in spending too. Not because you have a clear plan, but because you want the option to move. That kind of spending creates huge demand… until budgets tighten or the internal project gets canceled.
That’s the part that makes me cautious. Not because I think the demand is fake, but because I think a meaningful chunk of it is fear-based and status-based. And spending driven by fear is loyal right up until it isn’t.
The expansion details matter here. A gigawatt-scale facility is not a casual bet. These are long, expensive commitments. You don’t just “pivot” out of that. If the AI boom stays hot, Nebius looks smart and early. If it cools even a little, you can end up with a very real problem: too much capacity, too much fixed cost, and customers suddenly shopping for cheaper deals because they’re no longer desperate.
And the winners and losers won’t just be shareholders. If you’re a city that welcomed a big data center project, you might get jobs and new tax money. But you’re also tying your local power and land to a single trend. If that trend changes, you don’t get to easily repurpose a giant facility built for one type of customer with one type of need.
On the talent side, acquiring the core team from Clarifai is also telling. This isn’t just “we’re building warehouses for chips.” It’s “we want brains too.” That could be smart—cloud alone can be a brutal business if you’re only competing on cost. If Nebius is trying to offer more than raw compute, that’s a real attempt to move up the value chain.
But it can also create confusion. When companies chase both massive infrastructure buildouts and higher-level product work at the same time, they sometimes end up doing neither as well as they think. Big infrastructure demands boring excellence: uptime, security, support, predictable delivery. Product teams demand speed, experimentation, and constant change. Mixing those cultures is harder than press releases admit.
I’m also not going to pretend the 684% number tells the whole story. Percent growth can be a magic trick when the base is small. The $399 million figure is the more useful anchor—but even that doesn’t tell us how durable the sales are, how much comes from a small number of customers, or how sticky those contracts are. Public reporting rarely gives you the full texture people need to judge risk.
Still, I don’t want to dismiss it. There’s a real chance this is the beginning of a new “power era,” where compute and electricity become the quiet bottlenecks behind everything. If that’s true, companies that can build and operate data centers well will have leverage. They’ll shape which products get built quickly and which ones die in a waiting room.
But if we get this wrong, we end up with a weird economy: too much money locked into giant facilities, pressure on power grids, and a bunch of companies forced to keep spending just to justify what they already built.
So here’s what I actually want to know, and what I’d argue we should debate instead of just cheering the headline: are we watching real long-term demand being met responsibly, or are we watching a fast land-grab that will punish whoever is last to blink?