NEW YORK, March 5, 2026, 14:33 (ET)
- CoreWeave has landed a multi-year deal to handle Perplexity’s AI inference workloads.
- Nvidia’s GB200 NVL72 clusters are at the core of the deal, though financial details remain under wraps.
- Shares slipped Thursday, with investors zeroed in on funding and spending concerns.
CoreWeave has struck a multi-year partnership with AI search startup Perplexity, putting forward dedicated Nvidia GB200 NVL72 clusters aimed at handling inference—the task of running trained AI models to generate real-time answers for users. Financial terms weren’t made public. “AI applications running in production require more than just access to raw infrastructure,” said Max Hjelm, CoreWeave’s senior vice president of revenue. Perplexity’s chief business officer, Dmitry Shevelenko, described CoreWeave as “an essential partner” supporting its infrastructure plans. CoreWeave
CoreWeave is under pressure to show it won’t leave fresh data centers underused—this partnership comes right in that spotlight. Shares slipped roughly 6.6% Thursday.
CoreWeave’s rapid outlays have left some investors uneasy, despite the company insisting that most of the spending lines up with confirmed customer demand. Speaking with Reuters in late February, CEO Michael Intrator said CoreWeave opted “to build faster” in order to ramp up infrastructure. D.A. Davidson’s Alexander Platt, for his part, called the aggressive expansion “a positive signal”—as long as new capacity actually gets deployed. Reuters
Perplexity is sticking with a multi-cloud strategy, and its deal with CoreWeave aims to bolster fresh offerings as demand climbs. CoreWeave, for its part, plans to roll out Perplexity Enterprise Max for its own employees.
Perplexity is turning to CoreWeave’s Nvidia Grace Blackwell clusters for inference, Axios reported, in a move signaling that speed could become critical as consumer AI tools ramp up for bigger, nonstop workloads.
CoreWeave finds itself in a strange spot—not a hyperscaler on the scale of Amazon or Google, yet its spending habits tell a different story. The company is pouring money into top-tier chips and power, angling for the same AI contracts as the giants. Elsewhere, smaller “neocloud” competitors such as Nebius are after those customers too, pitching quicker access to hard-to-get GPUs as their edge.
The Perplexity deal leaves the stock’s main overhang unresolved: can demand hold up, and can deployments hit deadlines, to support a capital-intensive buildout exposed to financing costs? A pullback in customer spending or any delays with data centers would kick revenue recognition down the road—though the company’s outlays wouldn’t let up.
CoreWeave has relied on long-term contracts and fresh financing to ramp up, pitching investors on the idea that its outlays are backed by signed customer commitments. Now comes the test: in the next few quarters, the challenge is whether those announced deals actually deliver consistent revenue and margins—rather than just more hardware purchases.