Losing Jensen Was Bad Enough. Losing NVIDIA’s Bench Would Be Worse

Nov 16, 2026
losing-jensen-was-bad-enough.-losing-nvidia’s-bench-would-be-worse

Losing Jensen Was Bad Enough. Losing NVIDIA's Bench Would Be Worse ©@crstrbrt from Getty Images via Canva, @HYWARDS from Getty Images via Canva

Losing Jensen Was Bad Enough. Losing NVIDIA’s Bench Would Be Worse ©@crstrbrt from Getty Images via Canva, @HYWARDS from Getty Images via Canva

Key Takeaways

  • With Jensen Huang gone, the bigger risk to NVIDIA is losing the leaders who run the business, not the CEO search.

  • NVIDIA’s products for the next two to three years are largely designed, so a brain drain would hurt most later, when analysts expect revenue to pass $1 trillion.

  • SpaceX, a big NVIDIA customer planning a chip fab with Tesla, is the rival to watch if engineers leave.

  • The first test of whether the bench holds is interim CEO Colette Kress’s Nov. 17 report.

NVIDIA (NVDA:NASDAQ) lost Jensen Huang as CEO on Nov. 11. The question now is whether it keeps the people who actually run the company.

Altimeter Capital CEO Brad Gerstner made the case for that bench on X the day NVIDIA announced the change: “Jensen is a friend & legend. But he was not a one man show.” His list ran from Jonah and Ian to Colette Kress, now interim CEO.

Elon Musk’s SpaceX (SPCX:NASDAQ) has already shown interest. Musk wrote on X on Nov. 11 that hiring Huang “would be an honor.”

A Bench This Big Takes Years to Build

NVIDIA had about 42,000 employees at the end of fiscal 2026, 31,000 of them in research and development, its annual report shows. R&D spending more than tripled, from $5.27 billion in fiscal 2022 to $18.5 billion in fiscal 2026:

NVDA R&D Expenses: Fiscal 2022-2026 (TIKR)

NVDA R&D Expenses: Fiscal 2022-2026 (TIKR)

The Next Two Years Are Already Designed

A brain drain would take time to show up in the products. Tae Kim, author of The Nvidia Way, said on TBPN that “the next two, three years are actually baked in the cake” for NVIDIA’s products. Its roadmap already runs through Rubin Ultra in 2027 and Feynman in 2028. Kim’s bigger worry is strategy three to five years out, NVIDIA’s biggest years by consensus:

NVDA Revenue, Actual and Consensus Estimates: 2024-2029 (TIKR)

NVDA Revenue, Actual and Consensus Estimates: 2024-2029 (TIKR)

Consensus has revenue climbing from about $412 billion in the year ending January 2027 to $896 billion two years later, with eight analysts seeing $1.08 trillion by January 2030. The chips that ship after Feynman are being designed now, by the bench Gerstner described.

Why SpaceX Is the Rival to Watch

SpaceX runs on NVIDIA’s chips. On SpaceX’s first post-IPO earnings call in August, Musk said, “We’ve decided to build exclusively on Nvidia because we think the Vera Rubin architecture is the best architecture.” That AI unit brought in $7.6 billion of SpaceX’s $14.6 billion of third-quarter revenue, and NVIDIA’s SpaceX stake was worth about $21 billion on June 30.

SpaceX is also planning Terafab, a chip fab with Tesla, and its IPO filing warns it may not succeed. NVIDIA engineers could help close that gap, so losing them to Terafab could cost NVIDIA twice: the people, and a big customer that could, in time, need it less.

Leave a comment