UBS: AI Infrastructure Demand Continues to Heat Up, CoreWeave GPU Pricing Has Risen ~38% Cumulatively
nashnova research
CoreWeave raised GPU hosting prices twice for a cumulative ~38% increase; UBS says this confirms tight AI compute supply-demand, but whether pricing power converts to profit still awaits earnings proof.
How much did CoreWeave raise prices, and how?
CoreWeave hiked hourly pricing across all GPU SKUs — different GPU product configurations — by 25% in July 2026.
Two to three months later it added another ~10%, bringing the cumulative increase to roughly 38%.
This means → It was not a one-off adjustment. CoreWeave raised prices twice and the market absorbed both rounds — downstream customers need the compute badly enough to pay up.
What signal is UBS reading here?
UBS analyst Karl Keirstead notes that Nvidia GPU hosting prices are rising broadly, and revenue per gigawatt (revs/GW — how much revenue each unit of power consumed generates) is climbing too.
He sees this as the core pillar of the bull case for hyperscale cloud providers and neo-clouds — a new wave of cloud companies focused on AI compute leasing.
In plain terms = It is not just price going up. Revenue per unit of electricity is rising too — meaning AI compute demand is real, not speculative froth.
What could slow this trend down?
Siting resistance: Local communities and state/local governments are pushing back harder on AI data-center projects, but Keirstead's research suggests the industry can handle this better than the market expects.
Financing divergence: Access to capital and borrowing costs are splitting apart. Some projects may stall as a result.
This means → Not every player can keep building. Capital is getting more expensive; companies that cannot secure funding will fall behind. But better-positioned leaders are unaffected — and UBS explicitly names CoreWeave among them.
Does raising prices automatically mean more profit?
Keirstead flags a clear caveat: there is still a time lag between price increases and profit. Revenue gains take time to flow through to each company's income statement.
This reflects a sharpening divergence logic in AI infrastructure — pricing power + financing ability will be the key variables separating leaders from laggards.
In plain terms = Being able to raise prices proves you have leverage. But investors need to wait for earnings reports to confirm those price hikes actually turned into profit — rather than being eaten up by costs.
市场有风险，内容仅供研究参考，不构成投资建议。