The End of an Era! Semianalysis: Without Jensen Huang, Nvidia Will Become the Next Apple

Jensen Huang has left Nvidia, the company he helmed for 33 years.

“Nvidia was never built to operate without him.” This is what renowned semiconductor research firm Semianalysis wrote in its latest report.

Semianalysis likens Nvidia to the next Apple.

At such scale and size, the only other founder-led visionary company that has thrived after its founder’s departure is Apple under Steve Jobs.

SpaceX and Tesla cannot run without Elon; Nvidia cannot run without Jensen.

Semianalysis believes that with professional managers taking over, the most likely trajectory for Nvidia is: massive share buybacks, shrinking the balance sheet, and abandoning aggressive bets on the ecosystem. “The company’s pace of technological advancement will be slower than before”—acceptable to shareholders in the short term, but it is a negative signal for the entire AI industry. “Without him, the technology world will be vastly different.”

The bigger concern is: the supply chain advantages, customer trust system, and open source ecosystem moat personally built by Jensen are difficult for his successors to replicate, and by 2028, over 70% of total capacity may be shifted away from Nvidia.

Currently, the most promising internal successors at Nvidia are Ian Buck (current Vice President of Data Center, creator of CUDA) and Jonah Alben (Senior Vice President of GPU Engineering), but Semianalysis has reservations about both: “Both are excellent, but may not grasp the essence of supply chain management like Jensen.”

A Kingdom, Only One King

Jensen Huang has approximately 60 direct reports, covering nearly all key functions. He has publicly stated: “The CEO should have the most direct reports, because these people require the least management.” He does not hold one-on-one meetings, nor does he care about hierarchy or titles.

This extremely flat structure allows Nvidia, with about 42,000 employees, to support one of the top market caps in the world—Jensen himself refers to it as “the smallest big company in the world.”

But the price is: almost no one is groomed as a true successor.

Semianalysis points out that only five executives are given a prominent introduction under Jensen’s name on Nvidia’s website, and two of them have legal and finance backgrounds. Earlier this year, an executive told Semianalysis: “Ian (Buck) would rather die than take Jensen’s place, and doesn’t want that job. Jonah (Alben) is more respected, people say he could do it, but again, no one really wants that job.”

Currently, Colette Kress has been appointed interim CEO, but Semianalysis maintains, “To unify 60-plus vice presidents around a single company vision requires a certain type of individual; otherwise, the company's direction could simply vanish.”

The Founder’s Bets, Managers Can’t Place

Nvidia’s history of success is essentially a history of making bold bets.

CUDA lost money for years, and was continually questioned by Wall Street analysts. Jensen persisted. A former employee told Semianalysis:

Despite strong opposition from many within the company against using limited transistors for programmable shaders, CUDA, and tensor cores, Jensen got the entire company behind these efforts.

Now, Nvidia internally has two more projects Jensen labels as “zero to $10 billion bets”—Omniverse and autonomous driving. They don’t make money today, but he believes they are as important as CUDA.

Semianalysis asserts: successor professional managers will not make such bets. Their accountability to the board and shareholders is fundamentally different. “Long-term bets and conviction will become much less likely.”

The comparison is right there—Apple after the Steve Jobs era.

With the founder gone and professional managers in charge, the company becomes more conservative and returns more cash to shareholders. Investors become happy, but the company's technological development slows compared to the past.

Customers Buy Jensen, Not Just GPUs

Jensen’s departure will also have a structural impact on Nvidia’s customer base.

Jensen is the chief decision-maker for chip allocation and the central figure in all major customer negotiations. As Semianalysis wrote:

New cloud service providers around the world secure funding because Jensen supports them—not because they compared chips and chose Nvidia.

Specifically, Nvidia’s “endorsement program” enables operators without investment-grade ratings to borrow at investment-grade pricing—Nvidia sells GPUs at full profit on day one, then sets revenue floors for buyers and guarantees their revenue for landlords or personally signs lease agreements.

As of the latest fiscal quarter, Nvidia disclosed off-balance-sheet guarantees totaling $530 billion, a sharp increase from $184 billion in the previous quarter, including $36 billion in “take-or-pay” minimum guarantees and $99 billion in equity stakes.

Without Jensen’s personal guarantees, this (AI financing) ecosystem could shrink dramatically.

The Supply Chain Moat is “Personal”

Nvidia’s control over suppliers is also highly dependent on Jensen’s personal reputation.

Semianalysis revealed a crucial detail in its report: there is an implicit preferential pricing agreement between Nvidia and SK Hynix. They estimate that in 2027 for HBM4, Nvidia’s pricing from SK Hynix is approximately $3.0 to $3.3 per Gb, while other top-tier customers such as AMD, Broadcom, Amazon, and Google pay $3.7 to $4.1 per Gb. In return, Nvidia gives SK Group preferential GPU allocation rights.

Nvidia’s supply and capacity commitments jumped within a single quarter from $119 billion to $279 billion, primarily to lock in memory supply.

All of these deals were personally negotiated by Jensen, CEO to CEO.

Without him, these companies will no longer prioritize Nvidia to the same extent, and others will find it easier to secure capacity.

Semianalysis predicts that by 2028, over 70% of total capacity may no longer flow to Nvidia.

The Fracture in the Open Source Ecosystem

Jensen also plays another, less-discussed role: the bridge between Nvidia and the global AI open source community.

Semianalysis points out that this is an “open secret”—Nvidia’s Developer Technology team (DevTech) closely collaborates with multiple cutting-edge AI labs; their engineering exchanges have produced a series of kernel library innovations optimized for Blackwell GPUs, such as DeepGEMM, DeepEP, and MiniMax MSA. These results are then disseminated to the global open source community and adopted by widely-used inference engines like SGLang and vLLM.

Without Jensen as a spokesperson offering financial and engineering support for open source models, these labs might have to again deny any Nvidia GPU holdings and restrict Nvidia software to internal use only.

Chinese innovation would no longer be able to circulate between the Nvidia ecosystem and the broader machine learning ecosystem, which is fundamentally negative. It is positive for closed AI labs, but a setback for startups.