On November 12, 2026, Nvidia officially announced a leadership transition: after 33 years at the helm, founder Jensen Huang resigned as president, chief executive officer, and member of the board of directors, effective immediately and without any advisory title. Chief Financial Officer Colette Kress stepped in as interim CEO, while Harvey Jones—a founding-era director who has served on the board since the company’s inception—became the first independent chairman in Nvidia’s history.
When the news broke, the most baffling detail was the whiplash in the timeline. Just a week earlier, public debate was consumed by tabloid and espionage rumors linking Huang to Alice Huang, a shareholder in the Singaporean compute firm Megaspeed. Yet on the very day of his resignation, a joint investigation by six CNN reporters confirmed that neither the U.S. intelligence community nor the Department of Justice had concluded Alice Huang was a Chinese spy. If the national-security allegations had not been substantiated, why did a board that had issued a statement backing Huang just four days earlier turn on him completely within three days?
The stock market’s reaction that day was equally counterintuitive. After plunging at the open on the loss of its visionary leader, Nvidia was quickly bought up by retail investors and closed down just 4.34%, while rivals AMD and Broadcom both surged more than 5.5%. The real carnage hit downstream players that normally depend on Nvidia’s chip supply: neocloud provider CoreWeave and server assembler Super Micro Computer. Putting the board’s three-day reversal together with the sell-off in downstream customers reveals what this storm actually broke.
Start with the first puzzle: why could the board defend him against external espionage rumors, yet not survive internal workplace relationships? When Huang sent an all-hands email on November 8 admitting to an extramarital affair with Alice Huang, her status was that of a shareholder at an external customer company. As long as the issue involved an outsider, the board could frame Huang as the target of foreign intelligence entrapment, hire former Acting Manhattan U.S. Attorney Joon H. Kim to lead an independent investigation, and stand in the same trench as the founder to mount an external defense. By Wednesday, however, media reports revealed that Huang had carried on a long-term relationship with Kimberly Powell, the Vice President of Healthcare who reported directly to him, followed the next day by revelations of a third extramarital affair with another female employee inside the company.
Under U.S. public company governance rules, an undisclosed romantic relationship between a CEO and a direct report is an untouchable third rail: because of the power asymmetry between supervisor and subordinate, the company cannot clear itself of suspicions of favoritism in promotions, raises, and budget allocations. What alarmed directors even more was Huang’s cover-up. He had lied repeatedly to deny any affairs during internal inquiries since late last year, and last autumn he even went behind the board’s back to privately hire the crisis PR firm Heller to quietly intervene in media inquiries directed at the company. It was not until General Counsel Tim Teter confronted him in person at his home on November 6 of this year that he confessed to the relationship with Alice Huang—while still concealing the two internal relationships with subordinates. That directly triggered the McDonald’s precedent that U.S. public company boards fear most: when McDonald’s former CEO was caught in a relationship with an employee, the board took his word that it was an isolated incident and allowed him to depart on friendly terms with a massive severance package. After subsequent investigations uncovered three additional undisclosed relationships with subordinates, pension fund shareholders sued the entire McDonald’s board in Delaware Chancery Court for breach of fiduciary duty, forcing the company to turn around and sue its former CEO to claw back over $100 million. With that precedent in mind, once two internal subordinate relationships surfaced just five days after Huang confessed to the external affair, Nvidia’s independent directors knew the stakes: even after decades of friendship, if they did not sever ties with him cleanly that very night, they would be the ones sitting in the defendants’ dock in a shareholder derivative lawsuit the next morning.
The other ticking clock that forced the board to act overnight was the third-quarter earnings report scheduled for November 17. Under the Sarbanes-Oxley Act (SOX), enacted in the wake of the Enron accounting scandal, every quarterly financial report filed by a public company requires the CEO and CFO to sign personal certifications—carrying severe personal legal liability—attesting to regulators that the company’s internal controls over finances and personnel are operating effectively and that management has not concealed any conflicts of interest or fraud. At a typical tech giant with a Chief Operating Officer and layers of delegated committees, an executive’s personal misconduct might not taint the business ledger. But Huang built Nvidia into a single-person centralized machine with no intermediate buffer layers: the company has no second-in-command, roughly 60 executives report directly to him alone, even Executive Vice Presidents lack independent budget authority, all employees send weekly emails listing their Top 5 tasks for his review, he personally reviews compensation packages down to rank-and-file employees, and his son-in-law Nico Caprez serves as Vice President of Global AI Infrastructure Growth. In a system where every personnel and capital approval runs through one person, Jensen Huang’s personal candor was the sole hub of the company’s internal controls.
By Wednesday night, the board and CFO Colette Kress faced a deadlock: if they kept Huang on after catching him concealing internal conflicts of interest, no one would dare sign the statutory internal-control certifications for the Q3 report five days later; if they refused to sign and announced a delay in the Q3 filing, the market would immediately suspect accounting irregularities, triggering a stock crash. The only way out was to remove Huang from management and the board before earnings, allowing Colette Kress—who was untainted by the scandal—to serve as interim CEO and legally sign off on the Q3 report on her own. That is also why Nvidia made a point of stating in its transition announcement that Q3 results would be released as scheduled on November 17.
At this point, many readers may wonder: Nvidia is a $4 trillion empire that Jensen Huang founded and ran for more than three decades—how could the board simply replace him overnight, and why couldn’t he overrule the directors like other Silicon Valley founders? That happens to be Huang’s structural vulnerability. At Meta or Google, founders like Mark Zuckerberg hold super-voting shares with ten votes per share, allowing them to control a majority of voting power even with a modest equity stake so the board can never fire them. Nvidia, however, operates under a strict one-share, one-vote structure, and Huang holds only about 3.6% of the common stock. Under the company’s bylaws, the 10 independent directors who sit alongside him on the board can remove him from office at any time with a simple majority vote.
With the vote count completely out of his hands, the terms of his employment structure closed off his remaining exit routes. None of Nvidia’s executives have severance agreements, and Huang still held $108 million in unvested equity awards granted by the company in tranches. Had he refused to cooperate and waited for the board to formally terminate him the next day for cause—citing code-of-conduct violations and concealed conflicts of interest—not only would those unvested shares have been forfeited, but the company’s clawback provisions would also have been triggered, exposing the historical bonuses and stock awards he had collected over past years to recovery. Consequently, during a late-night video board meeting on Wednesday, an exhausted Huang repeatedly muttered “whatever” and “I don’t care anymore,” ultimately choosing to resign unconditionally, voluntarily forfeiting his unvested shares in exchange for a clean break, and retaining the law firm Paul, Weiss to handle the ongoing investigation. Even though more than 1,600 employee messages flooded the company’s internal suggestion box pleading for his return, the outcome was irreversible in the face of securities-law certification red lines and the directors’ own personal liability.
Once it becomes clear why the board had to replace him, a second question naturally follows: nothing is wrong with Nvidia’s chips or its engineering teams, so why did Wall Street immediately dump Nvidia’s downstream partners the moment it learned the CFO was taking over? Understanding that reaction requires looking at the other major project Huang pursued over the past two years alongside selling chips. In the second half of the large-model arms race, hyperscale giants like Microsoft, Google, and Amazon all began pouring billions into custom in-house silicon to reduce their dependence on Nvidia. To keep revenue doubling, Huang had to cultivate hundreds of thinly capitalized small miners—such as emerging neocloud rental platforms like CoreWeave and scores of AI startups—to step up and buy his expensive GPUs. Yet these smaller players lacked the cash to buy billions of dollars’ worth of GPUs and faced punitive interest rates from commercial banks because they lacked traditional collateral. To help cash-strapped miners afford his shovels, Huang turned Nvidia itself into their lender—making Nvidia, as venture capitalist Tom Tunguz’s ledger tally put it, the central bank of the AI industry.
Huang’s first mechanism for injecting liquidity into downstream buyers was writing direct cash checks to take equity stakes in customers or buy out startup technologies. Over the past two years, Nvidia deployed hard cash to invest in more than 200 private startups and over a dozen public companies, spending roughly $60.8 billion in total, while laying out another $42.0 billion to acquire the open-source platform Hugging Face and buy out technology licenses and core teams from companies like Groq and Poolside. The very first thing those startups did after receiving Nvidia’s investment capital or buyout proceeds was turn around and place orders for Nvidia chips. By the second quarter of this year, this playbook reached an extreme: the cash Nvidia spent in a single quarter on equity investments and acquisitions equaled 112% of the operating cash flow it actually collected from selling chips after everyday operating expenses—meaning it recycled every dollar of net cash earned from GPU sales back into the ecosystem and still had to chip in extra out of pocket.
The second mechanism was extending ultra-long payment terms to cash-strapped customers. Historically, the rule for buying Nvidia chips was that invoices had to be settled within 90 days. Later, to help capital-short neocloud customers stand up their data centers first, Nvidia allowed select customers to take delivery of GPUs upfront and stretch payment terms to as long as one full year. Within just six months, uncollected customer accounts receivable on Nvidia’s balance sheet ballooned to $63.1 billion, with 70% concentrated among its top five customers.
Direct investments and trade credit alone were still not enough: when small compute providers tried to lease data center shells, sign power purchase agreements, or borrow from banks, landlords and lenders were afraid they would go bankrupt mid-lease and refused to sign long-term contracts. So Huang deployed a third mechanism, having Nvidia sign a massive wave of off-balance-sheet backstop commitments that do not appear as formal liabilities on the balance sheet and are disclosed only in the footnotes to the financial statements. Much like a parent co-signing a child’s apartment lease, these guarantees cost Nvidia nothing upfront during normal times, but if the customer defaults, Nvidia has to step in with hard cash to cover the hole. According to an in-depth investigation by Hunterbrook and the footnotes to the company’s Q2 filing, combining $279.0 billion in upstream capacity purchase commitments locked in with suppliers like TSMC and nearly $90 billion in equity and unfunded investment commitments brings Nvidia’s total off-balance-sheet and investment commitments to a staggering $530.5 billion. Most striking of all is a full $164.5 billion in lease guarantees and compute take-or-pay agreements that directly shoulder risk for downstream customers. That figure includes a 20-year, $105.0 billion lease guarantee backing OpenAI’s Texas data center alone, alongside tens of billions of dollars in take-or-pay contracts with neoclouds like CoreWeave promising that if their GPUs sit idle without tenants, Nvidia will spend its own cash to buy back the unused compute so they do not lose money.
Above all three financial mechanisms stood the most critical backstop of all: Jensen Huang’s personal credit with the market. Several startup founders who received funding recalled that in 30-person Nvidia investment due-diligence meetings, Huang was often the only one asking questions, with deals frequently greenlit on airport tarmacs as soon as his private jet touched down. Whenever Wall Street bondholders doubted whether these neoclouds could service their debt, Huang would even step in personally to jawbone the market: in mid-October of this year, when top neocloud ally CoreWeave saw its bond default risk spike and its stock plunge nearly 10% intraday, Huang went live on Bloomberg Television that same day and declared, “I am very optimistic.” Convinced that Huang would never let the company fail, the market pulled CoreWeave’s stock out of the red on that single remark, sending it to a 6% gain by the close. Prior to that, Huang had personally convened six Wall Street titans—including Blackstone, BlackRock, and Goldman Sachs—to structure a $500 billion private credit pool designed to let compute companies pledge their Nvidia GPUs like real estate to secure low-interest loans, even exploring having Nvidia itself put up a 25% first-loss tranche to backstop the residual value of second-hand GPUs.
Now, the founder who greenlit checks on personal authority is suddenly gone, and the interim leader taking the helm—Colette Kress—happens to be a veteran CFO who has kept Nvidia’s books for 13 years and brings more than two decades of senior finance experience at major tech companies. Media reports had previously surfaced a telling detail: when Huang insisted internally on signing that hundred-billion-dollar data center lease guarantee for OpenAI, it was Colette Kress who led the finance team in voting against the deal inside the company, arguing the risk was too high, only to be overruled by Huang. Today, with Huang out and an independent chairman and outside counsel examining the books, Seaport Research semiconductor analyst Jay Goldberg pointed out the core implication: the $500 billion GPU-backed loan pool Huang was putting together with Goldman Sachs and Blackstone has not yet been formally signed, and the CFO is highly likely to pause it for review upon taking over—nor is she about to keep handing out one-year IOUs based on personal relationships or spending more than 100% of operating cash flow to subsidize money-losing customers.
Understanding that shift explains why Nvidia’s Thursday stock action was so tug-of-war: for Nvidia’s own shareholders, lending less money to cash-poor cousins and signing fewer hundred-billion-dollar lease guarantees is actually good for the lender’s own wallet. Nvidia has no trouble selling its GPUs; Microsoft and Amazon are still waiting in line. Previously, Wall Street’s biggest worry was that Huang was overextending the company by swapping hard cash for drawers full of IOUs and lease guarantees that could drag Nvidia under if downstream players imploded. With a conservative CFO now in charge, Nvidia’s own cash flow is actually safer—which is why retail investors felt confident pouring $111 million into the dip within the first half hour of trading, narrowing Nvidia’s closing loss to 4.34%. At the same time, archrivals AMD and Broadcom both jumped more than 5.5% to finish as the top two gainers in the S&P 500 that day: whereas Huang had previously locked customers onto Nvidia’s ship through equity stakes, extended credit, and compute take-or-pay deals, any loosening of that credit bundling gives major buyers more breathing room to purchase alternative chips from AMD or partner with Broadcom on custom in-house silicon.
The companies that truly had the ladder pulled out from under them were the downstream compute providers heavily reliant on Huang’s credit lifeline. Leading neocloud CoreWeave, which closed down 6.86%, is the textbook example. CoreWeave had just reported its third-quarter earnings two days before Huang’s resignation, exposing a fragile business model: it borrows money to buy massive quantities of power-hungry, heat-intensive Nvidia GPUs and rents them out to customers. Although it generated more than $3 billion in Q3 revenue, after deducting steep GPU depreciation, data center electricity bills, and employee salaries, the hard-earned operating profit from its core business came to just $207 million. Yet to buy those GPUs, it took on $46.9 billion in high-interest debt, and its net interest expense to lenders in that single quarter alone hit $846 million. In other words, its operating profit did not even cover a quarter of its interest bill—its interest burden was 4.1 times its operating income, resulting in a single-quarter net loss of more than $700 million. Worse still, CoreWeave’s management had just admitted on its earnings call that with prevailing market interest rates so high, the company is shut out of issuing unsecured corporate bonds at the parent-company level on its own credit before 2027, leaving it dependent on customer prepayments and contract-backed loans to stay afloat. At that exact juncture, the sudden departure of Jensen Huang—the man who could calm the bond market with a single TV appearance and organize a GPU-backed credit pool—prompted credit investors to vote with their feet immediately.
Server assembler Super Micro Computer, which fell even harder with an 11.30% plunge, was hit on the same day by a double blow of working-capital pressure and a federal probe. Super Micro does the heavy lifting of the supply chain, fronting its own cash to buy chips and components, assembling them into server racks, and selling them to customers like CoreWeave. Its latest quarterly report showed its gross margin had fallen to 12.7%, and due to inventory buildup and unpaid customer bills, it suffered a net operating cash outflow of $434 million in a single quarter—meaning a full quarter of selling servers left it burning more than $400 million in cash. If Nvidia’s CFO tightens upstream payment terms while downstream neoclouds face tougher borrowing conditions and cut orders, Super Micro gets squeezed from both ends in the middle. Adding insult to injury, Bloomberg reported on Thursday that Manhattan federal prosecutors are investigating whether entities linked to Alice Huang smuggled export-controlled chips into China through Super Micro servers, while Singaporean authorities publicly criticized the company that same day for limited cooperation in their server transshipment inquiry. To be sure, Thursday’s U.S. equity market was also buffeted by macro headwinds from a global Microsoft cloud outage and rising Treasury yields; a tightening of AI compute credit remains Wall Street’s pricing thesis for now, and Nvidia’s internal independent investigation has not yet issued its final conclusions.
Back in Washington, the political reality of this storm is far more pragmatic than a spy thriller. Although the FBI spent a year investigating how Megaspeed—the Singaporean firm controlled by Alice Huang—helped Chinese buyers transship billions of dollars in chips, and Secretary of State Marco Rubio wrote a memo this spring and briefed the President aboard Air Force One during an overseas trip in May, officials close to the Acting Director of National Intelligence told the press bluntly that labeling her a Chinese spy was overstated, and the CIA is unconvinced she knowingly passed intelligence to Beijing. One former U.S. Commerce Department official summed up the reality to reporters in a single pointed sentence: “It doesn’t take an intelligence operation to achieve that outcome.” The underlying commercial logic is straightforward: Chinese tech giants desperately needed high-end compute, and Jensen Huang desperately needed to sell more GPUs to hit his growth numbers. Their commercial interests naturally aligned, and as long as regulatory rules left loopholes open, businessmen pushed the trade as far as it would go.
Even though the intelligence community has poured cold water on the spy narrative, Washington’s political blame game is just getting started. In the past, Huang could board the President’s plane to lobby in person; now that the strongman is off the stage, every faction in Washington is rushing to distance itself. On the day of his resignation, Senate and House committees issued four summonses and subpoenas at once—not only summoning Nvidia General Counsel Tim Teter and Commerce Secretary Howard Lutnick, but also scheduling Huang’s closed-door House testimony for December 1, the exact day he had originally been slated to deliver a keynote address in Washington as the godfather of the AI industry. The next day, Treasury Secretary Scott Bessent publicly pointed the finger at the Commerce Secretary, charging that Treasury had pushed months ago to put Megaspeed on the export-control blacklist, only for Commerce to sit on its hands.
Even more consequential than politicians trading blame is the defection of Nvidia’s erstwhile allies in Silicon Valley. According to congressional lobbying records disclosed by Politico, OpenAI—having just secured that hundred-billion-dollar lease guarantee from Nvidia—privately informed congressional offices on the day of Huang’s departure that it now supports bundling three bills cracking down on AI chip exports to China, including the Chip Security Act, into the must-pass year-end National Defense Authorization Act (NDAA), completely reversing the wait-and-see posture it held in October. OpenAI’s commercial calculus is coldly pragmatic: if Congress uses legislation to permanently shut down the gray-market chip quotas flowing through Southeast Asia and other overseas channels, it will not only slow down Chinese competitors, but the scarce GPU capacity freed up as a result will stay right at home in the United States—ready to be fed first into OpenAI’s own data centers.
Jensen Huang may no longer be at the center of the storm, but the $530.5 billion compute credit machine he built is now facing a live stress test. Up next, Colette Kress will deliver Nvidia’s Q3 earnings report as interim chief on November 17 and hold an all-hands meeting the following day, setting the course for how the company will rein in its sprawling off-balance-sheet guarantees and one-year payment terms. Right behind that, the interest rate quotes from the six Wall Street institutions on GPU-backed loans will serve as the next key barometer. To track the industry tremors of the post-Huang era, rather than watching the daily ticks of Nvidia’s own stock price, watch the interest costs that downstream neoclouds have to pay to borrow money and buy GPUs.