What Jensen Huang’s Exit Puts at Risk Is Nvidia the Lender, Not Nvidia the Chipmaker


server room aisle with metal equipment racks

Nvidia’s board made three promises when it announced Jensen Huang’s resignation on November 12: the company’s “operations, customer commitments and product roadmap” would “continue without interruption.” That line, from board chair Harvey C. Jones in the press release Nvidia filed with the SEC, covers the chip business well. It says nothing about the other business Nvidia has been running for the past year: lending its balance sheet to the companies that buy its chips.

That second business is where Huang’s exit changes something real for AI buyers. The roadmap is mostly engineering that is already done. The backstops are judgment calls, and the person now making them is the CFO.

What happened, and what is still only alleged

The facts Nvidia has confirmed are narrow. The 8-K says that on November 11 the board accepted Huang’s resignation as president, CEO and director, effective immediately. Colette Kress, CFO since 2013, is interim CEO and keeps the CFO job. Jones, a director since 1993 and a former Synopsys CEO, becomes the company’s first board chair. The board will consider internal and external candidates for a permanent CEO. Neither document gives a reason for the resignation or mentions severance, a consulting role or a founder title.

The background is reported, not established. The board opened an independent investigation on November 8 after reports linked Huang to a former executive of Megaspeed, a Singapore company, whom U.S. officials reportedly suspected of working for Chinese intelligence. Huang acknowledged being unfaithful to his wife and denied compromising national security or Nvidia secrets, according to CNBC’s running coverage, which also notes that “the U.S. government has not presented proof of espionage.” Bernama, carrying Anadolu’s report, says the independent directors hired former acting U.S. Attorney Joon H. Kim and Wachtell, Lipton, Rosen & Katz on November 10.

The market’s first verdict was harsh and then less so. Per CNBC, the stock opened down more than 13% on November 12 and closed down 4.3%, after losing more than $740 billion in market value between the November 6 close and the November 11 close. CoreWeave fell roughly 9% to 12% and Nebius 17% to 19%, though Nebius also reported earnings that week. AMD gained nearly 17% from Monday.

None of that tells you whether your Blackwell or Vera Rubin order ships on time. Janus Henderson’s Jeremiah Buckley told CNBC, “We think the product roadmap is solidified through 2028,” and that matches how chip programs work: tape-outs, foundry slots and rack designs for the next two generations were committed long before this week. The part of Nvidia that runs on one person’s appetite for risk is the financing.

The $164.5 billion Nvidia stands behind

Nvidia’s 10-Q for the quarter ended July 26, 2026 has a section titled “Additional Commitments and Guarantees.” It opens with this: “We have entered into arrangements to assist select customers with securing the land, power, shell, and data center capacity needed to support their growth.” The next sentence is the one to keep in mind this week: those commitments “are dependent on the performance of our customers and partners.”

Arrangement (10-Q, quarter ended July 26, 2026) Amount What it does
AI cloud agreements $36B, phased FY2028 to FY2032+ Nvidia commits to buy cloud capacity from AI clouds that bought its hardware
Data center leases not yet commenced, for third parties $20B Roughly 15-year leases starting FY2028 to FY2029, which Nvidia expects to reassign
Land, power and shell guarantees for AI clouds $3.5B maximum gross exposure Covers partners’ lease payments if they default; partners hold $712M in escrow
SB Energy guarantees for OpenAI’s Ohio campus Capped at $105B Credit support on ~4.25 GW of 20-year leases, signed August 2026

XenoSpectrum’s tally groups those four lines at $164.5 billion, about 72% of Nvidia’s $229 billion in shareholders’ equity, and labels the grouping its own classification. That is a gross figure: it does not net out OpenAI’s agreement to reimburse Nvidia or the escrow, and the $105 billion guarantee only pays out after SB Energy fails to re-lease or sell the site, as The Next Web laid out in August. Nobody expects Nvidia to write a $164.5 billion check.

The line that matters most to people who rent GPUs is the smallest-sounding one. Under the AI cloud agreements, the 10-Q says, Nvidia commits “to cloud service agreements, which the AI clouds can unilaterally stop providing to us and sell to third-party customers at more advantageous rates.” Read that from the neocloud’s side: Nvidia is a buyer of last resort for capacity nobody else wants, at a price the provider would gladly abandon for a better one. That is a floor under neocloud utilization. It lets a provider with a thin balance sheet borrow against GPUs because the downside case has a well-capitalized buyer in it. (This site looked at the Ohio side of the same strategy in July, when the number being discussed was $250 billion.)

Why a CFO in the chair changes the arithmetic

Every one of those arrangements was approved under a CEO who treated ecosystem growth as the job. CNBC’s coverage, citing The Information, reported that Kress opposed the Ohio backstop. Nvidia has not confirmed that. If it is accurate, the person now acting as CEO, who still holds the CFO job, argued against the largest item on that list before it was signed.

Signed contracts are not the risk. The board’s “customer commitments” language plainly covers them, and walking away from a disclosed guarantee would be a legal and reputational disaster nobody on that board wants. The risk is the next contract. A neocloud that planned 2027 capacity on the assumption of another Nvidia rent-back agreement, or a lender that priced a GPU-backed loan with that assumption baked in, now has to ask who at Nvidia will sign it and on what terms. KeyBanc’s John Vinh called the change “a material increase in strategic, organizational, and execution risk,” and Stifel said it would watch “retention of senior leadership, and channel and export-compliance posture.” Neither phrase is about silicon.

I have sat on the buying side of supplier leadership changes more than once in IT operations at a large telecom, and the pattern is consistent. The press release promises continuity, and on the product side it is usually right: releases backed by a signed order arrive. What moves is the discretionary money. Extended payment terms, co-funded lab equipment and the informal promise to hold inventory until you are ready all go back to a committee, and some never come out. So when a key supplier changes leaders, I sort every arrangement into two piles, what is on paper and what was a relationship, because a new leader honors the first and renegotiates the second.

The CoreWeave numbers show how far that dependency runs. Its 10-Q for the quarter ended September 30 reports $132.3 billion in remaining performance obligations against $4.0 billion in stockholders’ equity and about $41.6 billion in recourse debt, with two unnamed customers accounting for 24% and 30% of quarterly revenue. A company with that shape lives on lenders’ confidence, and Nvidia’s capacity commitments are part of what those lenders are underwriting. A September 2025 order form obliges Nvidia to buy CoreWeave’s unsold capacity through April 2032, per the same XenoSpectrum breakdown.

There is a counterweight. Supply is still tight. Futurum’s Daniel Newman put it bluntly to CNBC: “If everyone turned tomorrow and went to AMD, AMD couldn’t supply.” Nvidia does not need to subsidize demand for its chips the way it might have if demand were soft, which is exactly the argument a finance-led interim CEO could use to stop writing new backstops without losing a sale.

What to check if your GPUs come from a neocloud

If you buy GPU hours directly from AWS, Azure or Google Cloud, this week changes little for you. If your training runs or inference fleet sit with a specialist provider, a few questions are worth asking before your next renewal, not after:

  • Does any of our capacity sit under an Nvidia cloud service agreement? Providers do not have to disclose it and Nvidia does not name counterparties, but your account team usually knows whether a cluster was financed with an Nvidia commitment behind it.
  • How long is our term compared to the provider’s financing? A three-year reserved contract with a provider whose next tranche of capacity depends on a new Nvidia agreement carries more risk than one built on capacity already delivered.
  • What happens to prepaid amounts if the provider is acquired or restructured? Smaller providers with heavy debt are the likeliest to be bought or restructured. Check assignment and termination clauses now, while you still have leverage.
  • Is there a second provider qualified for the workload? For inference in particular, a self-hosted or alternate route priced out in advance turns a supplier problem into a routing decision.

The first real signal arrives on Tuesday. Nvidia reports fiscal third-quarter results on November 17 at about 1:20 p.m. PT, with the call at 2 p.m. PT, according to the press release. The 10-Q that follows will update the commitments table. If the $36 billion in AI cloud agreements grows, Kress is continuing Huang’s strategy. If it holds flat while revenue keeps climbing, Nvidia has quietly stopped being the AI industry’s lender, and neocloud pricing will start to reflect that. Last quarter’s earnings were about losing China and growing anyway; this quarter’s most important number may be in a footnote.

Ty Sutherland

Ty Sutherland is the Chief Editor of AI Rising Trends. Living in what he believes to be the most transformative era in history, Ty is deeply captivated by the boundless potential of emerging technologies like the metaverse and artificial intelligence. He envisions a future where these innovations seamlessly enhance every facet of human existence. With a fervent desire to champion the adoption of AI for humanity's collective betterment, Ty emphasizes the urgency of integrating AI into our professional and personal spheres, cautioning against the risk of obsolescence for those who lag behind. "Airising Trends" stands as a testament to his mission, dedicated to spotlighting the latest in AI advancements and offering guidance on harnessing these tools to elevate one's life.

Recent Posts