CoreWeave's AI cloud business is booming—demand for Nvidia GPUs is so strong that every available chip could be sold to multiple customers. Yet the stock dropped nearly 5% on Wednesday, highlighting the tension between explosive growth and the heavy capital required to sustain it.
Demand and pricing power are strong
CEO Michael Intrator said at Goldman Sachs' Communacopia + Technology Conference that the company is "struggling to meet demand every day." He added that "every GPU we have could be sold to multiple different clients," giving CoreWeave significant pricing leverage.
Truist Securities analyst Arvind Ramnani raised his price target to $165 from $155 in August after CoreWeave increased prices across its product lineup by 25% in July. Ramnani noted that "pricing power is more than offsetting rising GPU costs," with higher customer prices expected to flow into margins during the second half of 2026.
CoreWeave's second-quarter revenue backlog stood at about $104 billion, excluding more than $25 billion in commitments signed early in the third quarter. Intrator also said the economics of newer infrastructure deals are improving because customer pricing is rising faster than costs for power, memory, and other inputs.
The cost of keeping up
Despite the demand, CoreWeave cannot convert it into revenue without continuously adding GPUs, power, and data-center capacity. The company raised its 2026 capital-expenditure forecast to $35 billion–$39 billion from $31 billion–$35 billion. Second-quarter capex reached about $9.4 billion.
Bernstein analysts called it "the strongest print CRWV has delivered," but noted that while guidance increased across revenue, adjusted operating income, annualized run-rate revenue, and active power, "capex went up higher than all of them." Bernstein raised its target to $74 from $67 but kept an Underperform rating, citing potential shifts in the long-term growth trajectory.
This explains why stronger demand does not automatically translate into a higher stock price. Every customer CoreWeave wants to serve creates another requirement for chips, infrastructure, and financing. Revenue may arrive later, while much of the spending comes first.
Debt and financing costs weigh
CoreWeave is also paying heavily to finance its expansion. Its recently arranged $2.6 billion delayed-draw term loan carries an estimated interest rate of about 8.2%–9.2%, higher than many investors expected.
Freedom Capital Markets analyst Paul Meeks sees upside, noting that "shorter-term contracts are now pricing & repricing at much higher rates," which can remain profitable even with higher borrowing costs. Freedom retained a Buy rating and $151 target.
The bull case is straightforward: scarce Nvidia compute keeps customer pricing high enough for margins to improve faster than financing costs rise. The bear case is equally clear: if AI capacity becomes less scarce, customer pricing weakens, or borrowing remains expensive, CoreWeave could be left with a huge capital base that produces lower-than-expected returns.
Wednesday's decline does not suggest customers have stopped wanting GPUs—Intrator's comments point to the opposite. The harder question is what happens after CoreWeave gets those GPUs. For more on Nvidia's role in AI infrastructure, see Nvidia's AI infrastructure plays and CoreWeave's recent jump on OpenAI's Astra model.
This article is for informational purposes only and does not constitute financial advice.
