A 454% revenue jump and four billion-dollar contracts drove Nebius shares to $227.70, though CoreWeave's own 20% rally tells a bigger story about GPU demand.
Nebius Group posted the kind of quarterly earnings that make analysts spill their coffee. The AI cloud infrastructure company saw its shares jump 18% to $227.70 on August 12 after reporting Q2 revenue of $582.3 million, a 454% year-over-year increase that topped consensus estimates of $572.75 million.
The stock’s year-to-date performance now sits at roughly 131%. For context, that means a $10K investment in Nebius at the start of the year would be worth about $23K today.
The numbers behind the rally #
Nebius’s AI cloud segment, which accounts for 98% of total revenue, grew more than 500% year-over-year. That kind of concentration in a single business line would normally make investors nervous. In this case, it’s the concentration that’s making them money.
Four landmark contracts exceeding $1 billion each anchored the quarter. Perhaps more importantly, 70% of those contracts included customer prepayments. Customers paying before they even use the service is the corporate equivalent of a restaurant with a three-month waitlist. It signals demand that outstrips supply.
Management pointed to durable demand trends and near-term capacity constraints, a combination that typically supports pricing power. When you can’t build data centers fast enough to meet orders, you’re in a pretty comfortable negotiating position.
CoreWeave actually had the bigger single-day move #
The comparison to CoreWeave is instructive, and perhaps a little unflattering for the “outperforming” narrative. CoreWeave shares climbed 20% to $108.40 on the same day, edging out Nebius’s 18% gain in pure percentage terms.
CoreWeave’s Q2 revenue came in at approximately $2.5 billion, dwarfing Nebius’s $582.3 million. The company also disclosed a revenue backlog of $104 billion, a figure so large it deserves a moment of quiet reflection. That backlog represents years of contracted future revenue from customers who need GPU compute capacity and have committed to paying for it.
CoreWeave’s management, like Nebius’s, flagged sold-out GPU capacity as a defining feature of the current market. The company raised its full-year guidance on the strength of the quarter. When a company with a $104 billion backlog says it’s raising guidance, that’s less of an earnings surprise and more of a structural statement about AI infrastructure demand.
Cloudflare, the third point of comparison, sat this particular rally out entirely. Its shares held steady at $307, which makes sense given its positioning. Cloudflare operates primarily in edge computing and content delivery, a different layer of the cloud stack than the GPU-intensive compute that Nebius and CoreWeave provide. Comparing Cloudflare to Nebius on AI cloud performance is a bit like comparing a trucking company to a semiconductor fab. They’re both in the supply chain, but they’re solving different problems.
Why GPU cloud is the trade of 2026 #
The simultaneous rallies in Nebius and CoreWeave reflect something broader than two good earnings reports. Hyperscaler spending on AI infrastructure has intensified throughout the year, and companies that can deliver GPU compute capacity are benefiting from what amounts to a supply shortage.
Nebius’s 454% revenue growth and CoreWeave’s $104 billion backlog tell the same story from different angles. The demand for AI training and inference compute is growing faster than the industry can build out capacity. That’s the kind of market dynamic that rewards early movers with outsized returns.
The prepayment structure in Nebius’s contracts is particularly notable. When customers are willing to pay upfront for cloud capacity they haven’t yet consumed, it de-risks the revenue stream and provides capital for further buildout. It’s a virtuous cycle: more prepayments fund more capacity, which attracts more customers, which generates more prepayments.
Both companies also benefit from a shift in how enterprises are thinking about AI compute. Rather than relying solely on the big three hyperscalers (AWS, Azure, Google Cloud), many organizations are diversifying their GPU cloud providers. Nebius and CoreWeave have positioned themselves as credible alternatives with purpose-built AI infrastructure.
The risk, of course, is cyclicality. GPU demand looks insatiable today, but the history of technology infrastructure is littered with buildout cycles that eventually led to overcapacity. The fiber optic boom of the late 1990s comes to mind. For now, though, near-term capacity constraints suggest that inflection point remains distant.
Investors watching this space should pay attention to two metrics going forward: backlog growth and prepayment rates. If those numbers hold or expand, it suggests the demand curve hasn’t peaked. If they start to flatten, it could signal that the market is approaching equilibrium, and equilibrium is rarely as profitable as scarcity.
Disclosure: This article was edited by Estefano Gomez. For more information on how we create and review content, see our