Nvidia's $6.3 Billion Backstop Puts CoreWeave's Balance Sheet on the Bull Case's Front Line

Jensen Huang says the selloff is temporary. Dan Ives says it is a buying opportunity. Neither claim answers the question that now sits at the center of the Nvidia story: whether the customers building AI infrastructure can actually pay for it.

Nvidia closed Friday at €190.52, down 1.7% on the day — a modest move that nonetheless captures the market's mood. Speaking in Berlin, Huang argued that AI will reshape industries and countries, that the buildout is only beginning, and that large hardware investments will eventually dispel current worries. Short-term headlines, he said, do not change the underlying fundamentals.

Ives, of Yorkville Ives & Co., takes the same side with a harder number attached. He reaffirmed a $300 price target and called the pullback a buying opportunity, pointing to supply — not demand — as the binding constraint on AI chip sales. The company itself expects roughly 70% revenue growth in fiscal 2028 and has guided to about $108 billion in revenue for the current quarter, with a margin of ±2%.

A Backstop With Conditions Attached

Where the bullish case runs into friction is CoreWeave. Nvidia holds a stake in the AI cloud operator and has separately agreed to take up unsold compute capacity through April 2032, a commitment worth up to $6.3 billion. The arrangement applies only to capacity CoreWeave fails to sell elsewhere, so it is not a payment due in full — but it does tie Nvidia's fortunes to a partner carrying roughly $50 billion in debt and lease obligations.

That debt has become expensive. The yield on CoreWeave's 2030 notes climbed to almost 16% this week, with Reuters reporting a rise to 15% on bonds issued in June. The widening spread reflects doubts about credit-financed AI projects whose returns arrive years after the borrowing.

Huang is unmoved. On Thursday he said CoreWeave's financial position is becoming more stable and that Nvidia assistance is not needed — an assessment, not a guarantee.

Fixed Rents as a Counterweight

The bull case has more than optimism behind it. Wedbush analysts see no evidence of broad AI demand weakness in the recent nervousness, citing firm GPU rental prices even for older systems and supply chains that remain tight. Amazon, meanwhile, plans to deploy another 2 million Nvidia GPUs across 2027–2028 — a demand anchor, though still a planned investment rather than a completed one.

What would settle the debate is visible relief in the relationship between capacity, revenue and capital needs. A calmer share price alone would not prove it.

The Bear Case Needs No Collapse

A slowdown does not require AI demand to break. It only requires financing costs and capital needs to outrun returns, at which point operators can defer projects or cap expansion even while interest in additional compute persists. Reuters has already reported signs of delays in CoreWeave's data center projects and slower growth in capital spending at major cloud providers. CoreWeave's own guidance of $35–39 billion in capex for 2026 underscores how much funding the buildout demands.

In that scenario, Nvidia's purchase commitment carries more weight than a statement of confidence — it is an obligation. No specific loss follows from it, but investors should not read it as proof that the infrastructure buildout already pays.

Earnings Season Takes Over

The next hard evidence arrives with quarterly results from the largest technology companies. Meta reports Wednesday, Microsoft on October 28 and Amazon on October 29. Commentary on capital plans, utilization of existing infrastructure and returns on that spending will matter more to Nvidia holders than any headline about chip demand.

Two paths remain open: a temporary crisis of confidence against still-durable demand, or an economically forced slowdown in the buildout. Huang's optimism points to the first. The coming reports will show whether actual investment practice agrees.