The sudden departure of Jensen Huang not only ended an era for Nvidia but also shook the financial foundation on which the entire AI ecosystem operates.

In the past three years, Nvidia, under Huang's leadership, played the role of "AI central bank," weaving a circular financing network exceeding $500 billion through equity investments, credit guarantees, and leaseback agreements, continuously providing blood transfusions to this cash-burning industry. Now, this founder has stepped down due to an extramarital affair scandal, and the successor is none other than his chief financial officer, Colette Kress, whom he publicly belittled as a "frightened deer" and dismissed as someone who "only knows finance."

According to the Wall Street Journal, it was Kress who proactively reduced Nvidia's guarantee commitment to OpenAI's Ohio data center from the initially negotiated $250 billion down to about $105 billion; she also repeatedly questioned how to price the guarantees and how to calculate investment returns during the rapid expansion of financing guarantee projects.

From the market's perspective, this new leader makes Nvidia's balance sheet safer but plunges all AI companies that are counting on Nvidia for their next checks into deeper uncertainty. Goldman Sachs analyst James Schneider pointed out that existing commitments are likely to be honored, but "without Huang's vision, connections, and convening power, the next round of similarly scaled multi-party ecosystem investments will be harder to realize."

A circular financing network exceeding $500 billion

To understand the impact of Huang's departure, one must first grasp how large the financing machine he built truly is.

According to Nvidia's 10-Q filing on July 26, the company's contingent liabilities and commitments have reached $530.5 billion. This figure does not yet include newly signed contracts after July, nor the up to $125 billion, or 25%, guarantee that Nvidia committed to provide for Huang's announced AI computing financing platform targeted at a scale of $500 billion.

The mechanism remains almost unchanged: Nvidia buys equity or provides guarantees, and beneficiaries commit to purchasing Nvidia hardware, with funds flowing back in the form of GPU orders. Safe Superintelligence, which has yet to launch any products, has already secured massive computing power commitments from the Vera Rubin architecture; Meta has increased orders for Vera Rubin, Spectrum-X, and BlueField-4; Nebius signed a 5GW deployment agreement with IREN—money flows out in equity form and returns in revenue form.

The Ohio data center project is the most extreme presentation of this logic. According to the Wall Street Journal, Nvidia ultimately provided OpenAI with a $105 billion lease credit guarantee, with an option to add 3.8GW of computing power. According to Nvidia's CFO, each generation of infrastructure deployed in the Ohio campus corresponds to about 1.5 million GPUs, which equates to Nvidia's revenue of about $150 billion to $200 billion—in other words, Nvidia is using a $105 billion guarantee to exchange for up to $200 billion in chip sales, repeating this cycle over a period of 20 years.

Bloomberg quoted warnings from Panmure Liberum, stating that "Nvidia Bank" serves as both the moat for this chip company and a source of credit risk—limited free cash flow is being massively committed to an industry that is still working to achieve commercial monetization.

Jensen Huang himself pressed the brakes, while Kress had long been the behind-the-scenes brake puller

Huang himself had already sensed the hidden dangers of this expansion before his resignation.

According to the Wall Street Journal, in August, he summoned top financial executives to his office and candidly stated that Nvidia was in "trouble": the company was negotiating guarantees with many firms unable to pay for chips, with some transactions being led by his son-in-law, Nico Caprez, who was promoted to vice president in January. Caprez, along with another vice president, Raj Mirpuri, had promised potential partners in the spring that Nvidia could complete credit support arrangements within weeks, provided they agreed to share a certain percentage of revenue, causing dozens of companies to line up.

In this meeting, Huang even interrupted Kress's report on the spot, calling her a "frightened deer" and labeling her as merely a "reporting CFO," not someone capable of partaking in strategic decision-making. Meanwhile, he described the real crisis the company faces: major cloud service providers are accelerating their shift to alternative chips, reducing their dependency on Nvidia, while the large number of wafers and memory that Nvidia has committed to purchase are coming "like an avalanche." According to Nvidia's financial reports, the related supply commitments surged from about $58 billion to about $279 billion within just three months.

Just before Huang issued the "avalanche warning," Kress had quietly initiated the braking action.

According to the Wall Street Journal, it was she who compressed Nvidia's guarantee commitment to OpenAI's Ohio project from the initial negotiation of $250 billion down to the final $105 billion, and continuously questioned the pricing logic and return calculations of the guarantee projects internally. In late August, Nvidia announced it would pause some revenue-sharing credit support transactions, requiring all new deals to be reviewed by the executive team and the board. In October, Huang met with CEOs of several major clients in Berlin, using the story of a Japanese ramen chef to convey a clear message to the executives of the cloud computing companies present: "You are not deciding the value of my product."

$500 billion financing platform: high-profile debut, zero deals

In August, Huang, dressed in his iconic leather jacket, announced this financing platform amid a group of executives from six top global asset management firms, attempting to turn GPUs into a category of assets available for financial institutions to invest in—similar to the securitization of real estate or toll roads, transforming rapidly depreciating hardware into collateralizable infrastructure assets.

However, according to Bloomberg, the platform has yet to finalize any deals; the six firms have only signed preliminary memoranda of understanding with Nvidia without forming final agreements.

One obstacle is the level of guarantees. Currently, Nvidia's residual value support is capped at 25% of the transaction size, while competitor Broadcom has provided approximately 86% of the guarantee for a $35 billion debt transaction with Anthropic in May this year and is set to provide about 72.5% for another $63 billion transaction. Faced with such a disparity, some clients and Wall Street partners have pressured Nvidia to raise its guarantee percentage. According to Bloomberg, the platform may not announce its first deal until January next year—provided that Huang's departure does not affect the process.

Sunaina Sinha Haldea, head of global private capital consulting at Raymond James, told CNBC that the management change provides all parties with a natural "pause button" to reassess transaction terms. She does not believe capital will withdraw, but warns that "they may demand higher returns for this uncertainty." She also expects lenders' inquiries to spread across the entire AI infrastructure transaction market: who is truly bearing the risks in these structures?

Kress takes over: a tightened grip, a complicated situation

From Huang's despised "reporting CFO" to Nvidia's first female interim CEO, Kress's rise to power is dramatic. Individuals familiar with Nvidia's investment decisions indicated that Kress had previously raised objections to deals heavily pushed by Huang and is not expected to advance new transactions at a pace close to Huang's speed.

Goldman Sachs analyst James Schneider noted after downgrading Nvidia's target price that Nvidia will fulfill all announced commitments, but the next round of large-scale multi-party ecosystem investments will be harder to advance due to the lack of Huang's personal resources and influence. Goldman Sachs particularly pointed out that Nvidia's stock had previously been trading at a discount, partly due to market concerns: if Nvidia or its competitors cease to provide financing, the growth of AI infrastructure will sharply slow down.

UBS analyst Tim Arcuri listed "financing" as the most significant impact of Huang's departure and estimated that a large portion of Nvidia's quarterly free cash flow of $50 billion to $60 billion is being flowed back to clients, suppliers, and emerging cloud computing firms; even a brief pause could siphon more than $50 billion from the AI ecosystem—timing coinciding with rising interest rates and tightening liquidity in the credit market.

William Blair analyst Sebastien Naji believes that existing contracts will persist following the CEO transition, but what is under threat is the next round of transactions and Nvidia's ability to defend its market share in 2027 and beyond. With Huang's departure or being constrained by investigations, competitors like Broadcom and AMD may gain greater market space.

Established and Uncertain: Which Transactions Will Be Impacted

Currently, the outside world believes that investments signed with legally binding contracts are unlikely to be affected. Nvidia's planned $5 billion investment in Safe Superintelligence and around $2.5 billion in the Thinking Machines Lab founded by Mira Murati both had contracts signed months ago and are expected to proceed as planned.

The transactions that will be more impacted are those still in negotiation. Nvidia plans to invest $10 billion in the upcoming IPO of Anthropic, but whether this will continue remains unclear; Nvidia previously provided a $105 billion lease credit guarantee to Anthropic. The agreement to negotiate a credit guarantee of over $5 billion for Nscale had not been signed by the end of October. Australian cloud computing company Firmus is seeking at least $2 billion in private financing after retracting its IPO, with Nvidia having previously participated in negotiations, but the result is undetermined. The humanoid robotics company Figure AI is reportedly discussing an additional $1 billion investment with Nvidia, but the latest developments are unclear.

Charu Chanana, chief investment strategist at Saxo, pointed out that the real risk is not the collapse of individual partnerships, but that amidst high debt levels, circular financing, and widespread skepticism regarding returns, Huang's departure may trigger a repricing of the overall risk premium of the AI ecosystem—"This is not yet a financing crisis, but it could serve as a catalyst for a broader risk repricing."

Kress will host Nvidia's third-quarter earnings conference call on November 17, where analysts will undoubtedly question her views on the $500 billion commitment and whether Nvidia will continue to play the role of "AI central bank." This will be the most direct test of whether the circular financing machine can continue to operate after Huang's departure.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到[email protected]，本平台相关工作人员将会进行核查。