Jensen Huang flew to China on Air Force One with President Donald Trump in May 2026 to meet Xi Jinping. Three months later he stood at the Nasdaq with six of the biggest names in technology and announced a proposed $500 billion AI buildout on CNBC. Nvidia had become the first company to pass a $5 trillion market capitalization, and its chips powered much of the more than $1 trillion expected to flow into AI that year. The company he co-founded in 1993 had no second-in-command. Everything ran through one man in a black leather jacket.
That changed on November 12, 2026, when the board replaced him with a chief financial officer. Nvidia's valuation rested on a one-year cadence Huang personally enforced across TSMC, SK Hynix and Foxconn, and that cadence now has to be kept by someone whose background is capital management.
What the board did
Nvidia's newsroom announcement states that Huang resigned as President, Chief Executive Officer and a member of the Board, effective immediately. Colette Kress, Executive Vice President and Chief Financial Officer, agreed to serve as interim CEO while keeping her CFO role. The board is searching for a permanent CEO and will consider internal and external candidates. Harvey C. Jones, an entrepreneur and venture investor with over 30 years of experience who previously served as CEO and Executive Chairman of Synopsys, became the company's first-ever Chair of the Board. The 8-K filing is dated November 11, 2026.
Those are the confirmed facts. The reasons are contested. SemiAnalysis reports that the board was investigating Huang's conduct, calls that conduct issue real, and argues the resignation is tied to large-scale GPU smuggling into China, citing its own tracking of more than 1.5GW of restricted Nvidia chips that ultimately ended up in China. It also says U.S. intelligence pressure was a major factor. PitchBook and CNBC report extramarital affairs, including one with a woman U.S. intelligence believes is a Chinese spy. None of these claims has been confirmed by the company. They bear on governance, but the succession question turns on what Huang did for Nvidia every day.
The one-year cadence was the product
The chip is the visible product. The cadence is what made it valuable at scale. Counterpoint Research says that for three years Huang set the annual rhythm for the whole AI industry, pulling foundries, memory makers, packaging partners and hyperscalers onto a one-year beat. That covered TSMC's advanced nodes (4NP and N3P), CoWoS packaging, SK Hynix and Micron's HBM3E and HBM4 memory, NVLink and Spectrum-X interconnects, direct-to-chip liquid cooling, and rack assembly by Foxconn, Quanta and Wistron, all moving on the same calendar.
A synchronized calendar across five layers of supply is a governance artifact, not an engineering one. It requires one decision-maker with the standing to tell five suppliers and the hyperscalers to move together, to absorb their risk when a process node slips, and to re-cut the plan mid-flight. No memory maker can make that call for Nvidia. Huang could.
Kress is a capital operator. She has been Nvidia's CFO since September 2013, and before that spent nearly 25 years in finance at companies including Cisco and Microsoft. The board's statement emphasizes her understanding of the business, customers and financial position. That profile suits managing a balance sheet and a crisis. It is not the profile that personally pressed TSMC for capacity or reset Foxconn's rack schedule.
Counterpoint's read is that the roadmap from Blackwell to Vera Rubin is locked in and backed by a deep engineering bench, so the ecosystem should keep moving for 18 to 24 months. After that, the outcome depends on who the permanent CEO is.
296 deals and a checkbook
Huang also used Nvidia's cash to bind the industry. PitchBook counts 296 venture deals Nvidia has joined since 2023, more than a third of them in 2026, and says it has joined a deal roughly every three days this year. It has backed nearly one in three AI unicorns founded since 2023: 49 of 153, or 32 percent by count and about 60 percent by value. PitchBook also reports Nvidia has been the go-to guarantor for the data center buildout, backstopping financing for AI labs and neoclouds.
Huang put the logic plainly on an August earnings call: "I think the only regret that I have is that I didn't invest more and sooner." A balance sheet that size, backing companies that need expensive GPUs and long-term debt, works as a guarantee only while the person managing it keeps writing checks. Lenders have been pricing in Nvidia's backing, not each startup's own cash flow.
The personal layer shows how much of the gravity was Huang's. Lloyd Blankfein wrote that he was "indispensable" and added that "the graveyards are filled with indispensable men," a line attributed to Charles de Gaulle. Michael Dell called him a "visionary of the AI age." Deutsche Telekom's Tim Höttges said a photo with Huang drew more press for his company than its iPhone launch.
Silicon allocation gives way to capital discipline
For three years the scarce resource was silicon allocation. Nvidia decided who got the next generation of GPUs and in what volume, and the rest of the stack adjusted to that power, which in practice was Huang's personal authority over suppliers and customers. Remove the person who enforced it and the constraint shifts. A finance operator optimizes for margin, capital return and downside protection. That means a longer planning horizon, less willingness to underwrite supplier risk, and more scrutiny of which customers deserve Nvidia's balance sheet.
The supply chain feels this first. TSMC and SK Hynix gain pricing power against a counterparty that no longer carries the same personal pull to win concessions. My expectation is that at least one top-five memory or packaging partner publicly slips a co-developed roadmap by a quarter within 18 months. If that happens, it confirms the cadence is no longer self-enforcing.
The neoclouds and AI unicorns feel it next. Many raised money on the assumption that Nvidia's commitment was a backstop. Without Huang's personal willingness to guarantee their debt, lenders will demand their own underwriting. My expectation is that at least two Nvidia-backed neoclouds fail to close follow-on financing without explicit backstop language within 18 months, forcing Kress or her successor either to guarantee the debt or let them fail. The likelier response is that Nvidia converts part of its venture portfolio into preferred-supply agreements, trading capital for guaranteed allocation.
Margin follows. I expect Nvidia's gross margin to compress 300 to 500 basis points over two years. The cause will not be AMD or in-house silicon, though those pressures exist. It will be the loss of Huang's leverage over suppliers who accepted his pricing and scheduling. The permanent CEO will be judged on capital allocation rather than vision: how much cash goes to backstops, how many deals close on Nvidia's terms, and whether the company is buying growth or buying loyalty.
The position is falsifiable. If Nvidia's top memory and packaging partners hold the 2026 to 2027 roadmap without public slips, and neocloud financing clears without Nvidia guarantees, the cadence survived the founder's departure and this thesis fails.
Who wins, who loses
TSMC and SK Hynix are the clearest winners. They face a buyer that can still order at scale but cannot easily force a synchronized calendar across five layers, which shifts leverage toward the suppliers holding scarce capacity.
The hyperscalers lose the single counterparty who could promise a roadmap all the way down the stack. Google, Microsoft, Amazon and Meta will hedge across Nvidia, their own silicon and AMD in ways they could previously defer. That is costlier, and it reduces the risk of one supplier controlling the calendar.
The Nvidia-backed unicorns and neoclouds face the sharpest repricing. Their financing story was Nvidia's willingness to stand behind them. Diligence tightens, and backstop language becomes the price of survival. Some will raise on their own merits. Others will find their cost of capital has moved against them.
Counterpoint expects a pause on founder-driven experiments, which gives suppliers and engineers breathing room and eases pressure on R&D and capital spending. The cost is slower quarter-on-quarter growth across the supply chain, the same coin viewed from the other side: less forced acceleration and less guaranteed demand.
The podium at the Nasdaq is empty
In the months before the November announcement, Huang stood at the Nasdaq with six of the biggest names in technology and unveiled a $500 billion buildout plan. The valuation that made Nvidia the first $5 trillion company rested on more than chips. It rested on a cadence that lined up TSMC's nodes, SK Hynix's memory, Foxconn's racks and the hyperscalers' orders, and on a checkbook that made smaller players believe they had a guarantee.
The cadence left with the man. Kress has the balance sheet discipline to run the company. Whether she or a permanent successor can run the calendar is the open question, and the board will grade the next CEO on capital allocation. The first public test will be a memory or packaging partner announcing a slip. The first real casualty will be a neocloud that learns Nvidia's name was never a guarantee.
