Jensen Huang Resigns: NVIDIA's 33 Years and the "Post-Jensen Era"

Jensen Huang has stepped down.

A one-sentence announcement from NVIDIA's board of directors has drawn an abrupt end to the era of Silicon Valley's most legendary founder.

On November 12, NVIDIA announced that co-founder Jensen Huang has resigned from his positions as President, Chief Executive Officer and member of the board of directors, effective immediately. The current Executive Vice President and Chief Financial Officer Colette Kress will serve as interim CEO, while continuing to hold the CFO post until a formal successor is confirmed. Harvey Jones, who previously led Synopsys, has been appointed as the first board chairman in NVIDIA's history.

The capital market reacted almost instinctively. After the news was released, NVIDIA's pre-market decline once widened to 10%; and since Jensen Huang issued an apology letter acknowledging extramarital affairs on November 8, the company's market value has evaporated by about 450 billion US dollars, and the market's predicted probability of his leaving before the end of the year once soared to 88%.

This company, regarded as the global AI computing power backbone, completed the whole process from independent investigation, public apology to power handover within four days.

This is not a "glorious retirement" after achieving great success. Jensen Huang is 63 years old this year. A few months ago, he publicly stated that he hoped to work until the end of his life, and claimed that he "does not believe in succession plans". His departure is a governance crisis caused by the superposition of three layers of pressure: personal scandal, national security concerns and geopolitics.

In short, Jensen Huang's departure seems rather hasty.

01. The Founder Who Built an AI Empire From a Graphics Card Manufacturer

Jensen Huang's contributions cannot be summed up by any single hit chip. What he really did was bet the fate of the company three times in 33 years on directions that had not yet been proven by the market.

The first bet was a gamble for survival. In 1993, Jensen Huang and two engineer partners founded NVIDIA on the dining table of a chain restaurant.

The first product NV1 stubbornly adopted the quadrilateral surface rendering route, which was completely incompatible with Microsoft-led DirectX standard that only supported triangle modeling. It became unsalable as soon as it was launched, forcing the company to lay off about 70% of its staff. The cash on the account was only enough to support the company for another 30 days, and the phrase "we are only 30 days away from bankruptcy" became the sense of crisis that he repeatedly emphasized.

In desperate situation, he quickly admitted his mistakes and restarted from scratch. In 1997, he launched the RIVA 128 compatible with industry standards to turn the tide; in 1999, the company went public, and released GeForce 256 in the same year, putting forward the concept of "Graphics Processing Unit (GPU)" for the first time. Later, it acquired the assets of 3dfx and recruited core engineers, laying its dominant position in the graphics card market. Even during the dark period of GeForce FX that was jokingly called "Dust Lord" in 2002, he never wavered in his adherence to the independent graphics card route.

The second bet is the bet on ecosystem, which best reflects his long-termism. In 2006, NVIDIA launched the CUDA parallel computing architecture, enabling GPUs that were originally only responsible for rendering images to undertake general-purpose scientific computing. At that time, this was a thankless task: the general-purpose computing market did not yet exist, Wall Street could not understand the significance of burning money, and the stock price remained low for a long time. But Jensen Huang withstood the pressure and continued to invest for nearly a decade.

In 2012, AlexNet trained with NVIDIA GPUs made a breakthrough in the image recognition competition, kicking off the wave of deep learning. In 2016, NVIDIA delivered the first DGX-1 supercomputer to OpenAI, which was still little-known at that time. When ChatGPT took the world by storm in 2022 and everyone realized that AI requires massive computing power, the development tools, software libraries and usage habits accumulated by NVIDIA around CUDA had formed an unshakable ecological barrier. Competitors can produce chips with similar parameters, but it is difficult to replicate the software ecosystem with millions of developers and more than ten years of code precipitation.

The third bet is the bet on system. In 2020, Jensen Huang advocated the acquisition of network manufacturer Mellanox for 6.9 billion US dollars, reshaping NVIDIA from a chip-selling company to a system-level company that sells computing, network and software as a bundled package. Subsequent full-rack solutions including DGX, HGX and GB200 NVL72 transformed NVIDIA from a component supplier to the chief designer of AI data centers, and NVLink and high-speed networks have become new moats to lock in customers.

After 2023, years of accumulation were fully realized. Data center revenue quickly exceeded the gaming business, and its market value exceeded 5 trillion US dollars in 2025, firmly ranking in the global first tier. In April 2026, Jensen Huang was awarded the IEEE Medal of Honor, which is the highest-level recognition of his contributions in the engineering and technology circle.

In the author's opinion, what is most irreplicable about Jensen Huang has never been a single chip, but his judgment to keep betting before the market is proven, and the willpower to bear the lack of understanding for a long time. CUDA sat on the bench for nearly a decade, and such a decision is exactly the most difficult to make for the professional manager system that pursues short-term performance. The real legacy he left behind is a system that interlocks hardware, network, software and developer ecology.

02. Cracks on the Trillion-Dollar Throne

Judging only from financial reports, NVIDIA is still a high-speed running money printer.

In the second quarter of fiscal year 2027, NVIDIA's revenue reached 96.2 billion US dollars, a year-on-year increase of 106%. Among them, data center business revenue was 89 billion US dollars, up 117% year on year, with a gross margin of 75% and a net profit of nearly 59.7 billion US dollars.

NVIDIA's revenue guidance for the next quarter is about 108 billion US dollars, and it is estimated in advance that the revenue of the next fiscal year will still grow by about 70%. The new generation platform Vera Rubin has been shipped since August, and Blackwell Ultra is still in large-scale deployment. However, behind the dazzling report card, five cracks are clearly visible.

The first crack is that the gross margin has peaked and fallen back. The management has lowered the gross margin expectation for the third quarter of fiscal year 2027 to 74%, and further bottomed out to 71% to 72% in the fourth quarter, due to the price increase of memory chips. This means that the high-margin "golden age" is being pulled back to reality by the cost law of the hardware business.

The second crack is the high concentration of customers. Ultra-large-scale cloud vendors account for more than half of the data center revenue, and the capital expenditure rhythm of a few large customers is almost the barometer of NVIDIA.

Goldman Sachs estimates that global AI capital expenditure will reach about 800 billion US dollars in 2026 and rise to about 1.2 trillion US dollars in 2027. Its essence is that the free cash flow of cloud vendors is transferred to chip vendors in one direction in the form of procurement. This mechanism created a profit myth during the expansion period, but it can also run in reverse. Once the capital expenditure of cloud vendors slows down, equipment suppliers will immediately feel the chill, and customers also have to bear years of depreciation.

The third crack is the controversy over profit quality and "circular financing". In the second quarter of fiscal year 2027, the company's "other income" reached 7.77 billion US dollars, accounting for about 11% of pre-tax profit, which mainly came from the valuation rise of equity held, rather than the chip sales business itself.

NVIDIA has invested nearly 50 billion US dollars in cutting-edge AI labs, and promoted a 500 billion US dollar financing plan to help enterprises purchase hardware. The cycle of "I invest in customers, and customers use the money to buy my chips" has long been discussed by the market for its vulnerability in an environment with high interest rates and unproven AI commercialization returns.

The fourth crack is geopolitics and compliance. Since 2022, the United States has continuously controlled the export of high-end AI chips, and the policy of the special chip H20 for the Chinese market has fluctuated repeatedly. Singapore once contributed about 22% of the company's revenue, an unusually high proportion for a city-state, but the company later stopped disclosing country rankings.

In this incident, Megaspeed, an enterprise associated with Jensen Huang, was accused of transshipping 7.5 billion US dollars worth of NVIDIA chips to Chinese enterprises through networks in Singapore and Malaysia, and the relevant parties were regarded as suspected foreign intelligence personnel by the security authorities. NVIDIA has thus been pushed to the center of the game between US national security and the two parties in the US, and its congressional investigation is almost inevitable.

The fifth and most core crack is key person risk. Jensen Huang has more than 60 direct reports, most of whom are from the engineering field, covering GPU, CPU, storage, network and other businesses. A large amount of technical information is directly collected to him alone, and he judges the priority; externally, he is the company's most important business representative and personalized brand carrier. His iconic leather jacket, classic quotes and "Jensen Huang craze" have long served as symbols of customer trust and capital expectations.

Since the CUDA ecosystem, system capabilities and supply chain have been highly systematized, the revenue in the next 12 to 18 months will not disappear suddenly due to the change of CEO. But the real loss of "key person risk" is not in the current revenue, but in the narrative ability, ecological centripetal force and expectation management — Jensen Huang is essentially the "chief explanation officer" of this round of AI capital expenditure cycle, responsible for making customers, investors and the industrial chain believe that the story can continue.

03. Interim CEO Takes Over and the Reform in the Post-founder Era

To understand the successor's reform, we must first understand the nature of this power arrangement. Kress is not an outsider parachuted in. She has served as NVIDIA's Chief Financial Officer since September 2013, and has nearly 25 years of financial experience in companies such as Cisco and Microsoft, having personally experienced every cycle of NVIDIA's rise. But her title is "interim", and she will continue to serve as CFO during the transition period, while the board of directors has officially launched a formal search that takes into account both internal and external candidates.

What is more thought-provoking is the position of chairman. NVIDIA has never set up a board chairman in its history, and now the appointment of senior Harvey Jones is itself a strong signal of governance checks and balances. What the board wants at this moment is stability, compliance and accountability, rather than recreating a strongman who "sets the tone alone".

In the short term, the main tone is "maintaining the established success" rather than "turning direction". The product roadmap has been scheduled to Vera Rubin, and the third quarter financial report on November 17 will be released as usual, and customer contracts will not become invalid due to the change of CEO. Kress's top priority is to stabilize the core technical team represented by Ian Buck, maintain the relationship with several major cloud vendors and the supply chain rhythm, and ensure that products are delivered as planned.

Looking at a longer timeline, a leader with a financial background plus a strong chairman will most likely bring four foreseeable reform directions.

The first is institutionalized governance: saying goodbye to the flat structure where 60 subordinates report directly to one person, reducing dependence on individuals by adding levels, redistributing decision-making power and introducing process management, and the organization evolves from "agile" to "standardized".

The second is financial discipline: capital allocation will become more prudent, the thresholds for repurchase, dividends and mergers and acquisitions will be raised, and the "circular financing" will be actively cooled down, so that profits will return to the chip sales business itself.

The third is compliance first: under the pressure of espionage, chip smuggling investigations and congressional pressure, export control and business in China will be more conservative, and the "pragmatic flexibility" that wandered on the edge of rules in the Jensen Huang era will shrink significantly.

The fourth is lowering the narrative tone: external communication will shift from relying on the founder's personal IP to systematic investor relations.

The historical reference that the market most often cites is Cook succeeding Jobs. After taking over, Cook grew Apple's annual revenue from about 108 billion US dollars to more than 400 billion US dollars, proving that professional managers can fully amplify a mature system. But there are essential differences between the two: Cook was the successor personally selected by Jobs and went through a long handover period.

Jensen Huang left forcibly with "immediate effect" and almost no handover period. Moreover, he publicly claimed before that he "does not believe in succession plans", and the company has never cultivated a clear second-in-command. This zero-buffer handover is the biggest risk.

Different routes can also be read from the candidate list. Internal candidate Ian Buck represents the continuation of technology and ecology, while external candidates including Arm CEO Haas, AMD's Lisa Su and even the just-retired Cook correspond to different imaginations of industry integration, competition reconciliation and global operation respectively.

Therefore, NVIDIA is likely to temporarily degenerate from a "cycle-creating company" to a "cycle-following company". When AI capital expenditure moves towards 1.2 trillion US dollars in 2027 and the cycle reverses, what the company needs most is exactly the kind of person like Jensen Huang who can redefine demand and find the next CUDA, which is exactly the capability that interim arrangement and financial-oriented governance structure are most difficult to provide.

In addition, for the Chinese market, Jensen Huang has always been a pragmatic faction who tried every means to do business. Successors will most likely be more conservative under political pressure, and NVIDIA's flexibility in China will shrink, which means computing power pressure in the short term, but opens a more certain window for domestic substitution in the medium term.

However, there is no need to be overly pessimistic. A company that can go through multiple cycles of gaming, cryptocurrency and AI never relies on personality worship, but on organization and ecology. Jensen Huang's greatest legacy is exactly CUDA, system capabilities and a capable team, which can exist independently of himself.

The real test for a legendary founder is to make the company still great after he leaves.

Jensen Huang did not leave himself a chance to verify the answer — he was not defeated by the technology cycle, but was pushed off the table by his own behavior and this highly politicized era. This is his personal tragedy, and also the inevitable price to be paid for the "strongly founder-bound" model.

But for NVIDIA, the "post-Jensen Huang era" is not necessarily the end of the legend. It is more like a coming-of-age ceremony that was long overdue, evolving from one person's vision to the governance of a complete system.

When the tide of the AI cycle turns for the first time, whether this system can see the next direction on its own.

This article is from WeChat official account "DoNews" (ID: lovedonews), author: Yang Bochen, published with authorization from 36Kr.