Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseInsight · On RadarLatent Space (swyx)→ original

Jensen Huang Steps Down as NVIDIA CEO After 33 Years, CFO Colette Kress Becomes Interim CEO

Jensen Huang, who led NVIDIA for 33 years, has stepped down from his roles as CEO, president, and board member. Colette Kress, the company's CFO since 2013, is now interim CEO and keeps her CFO title as well. Before the resignation, The New York Times reported on Huang's ties to a Chinese woman charged with espionage, and a second such story followed. For developers building on CUDA, the direction of the new leadership will become clearer after next week's quarterly report.

AI-processed from Latent Space (swyx); edited by Hamidun News
Jensen Huang Steps Down as NVIDIA CEO After 33 Years, CFO Colette Kress Becomes Interim CEO
Source: Latent Space (swyx). Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceLatent Space (swyx) ↗
Editor-in-Chief
Key metric #1Kress remains CFO and serves as interim CEO at the same time
Key metric #2NVIDIA's report next week could become the most-viewed in its history
Critics2–3 thousand Blackwell chips at Chinese labs, roughly 6 MW out of 24 GW

Jensen Huang, who ran NVIDIA for 33 years and turned a maker of gaming graphics cards into the main supplier of AI hardware, has stepped down. He gave up three positions at once: CEO, president, and board member. Colette Kress, NVIDIA's CFO since 2013, has become interim CEO. She keeps her previous post, so she now holds two jobs. For anyone training models, renting GPUs, or building products on CUDA, this is a reason to look at dependence on a single supplier.

What is known about the departure

NVIDIA announced the resignation on November 12, 2026. Swyx, the author of the Latent Space newsletter, calls it the most significant leadership change in the AI industry since the reshuffles at Google DeepMind. The day before, The New York Times published a piece reporting on Huang's ties to a woman accused of spying for China, and on the board preparing to replace him. Then reports of a second such story appeared, and the company announced the departure the following day.

The official reason has not been disclosed in detail. Swyx writes directly that the connection between the publications and the resignation remains a matter of dispute, but allows that the board may have made the decision. The accusations remain press reports, not established facts.

The timing is awkward. Next week NVIDIA presents its quarterly report to investors, and according to Swyx's estimate it will become the most-viewed report in the history of the NVDA ticker. The main question for analysts will most likely be not revenue, but who is running the company and in which direction.

Why this matters to developers

The community's main question is what will happen to open software and American leadership in AI. NVIDIA releases its own open models and tools, and CUDA remains the foundation of almost all training and inference. So far there is no announced change to products or supply, and a new strategy will only appear together with a permanent CEO.

In parallel, a study by Will Depue on GPU "fingerprints" in model weights has circulated. He claims that initialization patterns in published weights reveal what hardware a model was created on. He writes that six Chinese models, including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3, were initialized on Blackwell systems, four of them on Grace Blackwell. As evidence for the method he cites a match of 24 out of 24 on models with known hardware. For the largest runs he estimates the scale at 2–3 thousand Blackwell chips.

Critics push back. That is about 6 MW against roughly 24 GW of Chinese computing capacity, so the result can also be read as a sign that export controls are working. Others note that Chinese labs' use of Blackwell was already known. The data has not been publicly confirmed: Depue says senior researchers at OpenAI and Anthropic checked the result, but does not name them.

What to do in practice

There are few details on policy consequences, but one scenario is already being discussed. According to reports, the US Department of Commerce is preparing a rule that would restrict Chinese companies' access to foreign AI data centers. It would apply only in some countries, only to future shipments, and only to Blackwell-class chips and above. By one estimate, it would affect no more than 14% of Chinese computing in 2027. Some commentators expect fast bipartisan bills after Huang's departure, but that is still a forecast.

Practical takeaways, without panic:

  • If you depend on CUDA and NVIDIA infrastructure, you do not need to change anything urgently: no one has reported a halt to products or shipments.
  • Build the possibility of new export rules into your procurement and contract planning, especially if you work with international data centers.
  • It is sensible to keep critical weights and datasets in your own reserve. This practice is useful regardless of the news.
  • The quarterly report and Colette Kress's first statements will show whether the previous course holds.

The main questions remain open: who will become permanent CEO, and whether the new team will keep its commitment to open models. Until then, it is wiser to rely on NVIDIA's official documents rather than rumors.

Story

Key Questions Answered

What is known about the departure?

NVIDIA announced the resignation on November 12, 2026. Swyx, the author of the Latent Space newsletter, calls it the most significant leadership change in the AI industry since the reshuffles at Google DeepMind. The day before, The New York Times published a piece reporting on Huang's ties to a woman accused of spying for China, and on the board preparing to replace him. Then reports of a second such story appeared, and the company announced the departure the following day.

Why this matters to developers?

The community's main question is what will happen to open software and American leadership in AI. NVIDIA releases its own open models and tools, and CUDA remains the foundation of almost all training and inference. So far there is no announced change to products or supply, and a new strategy will only appear together with a permanent CEO. In parallel, a study by Will Depue on GPU "fingerprints" in model weights has circulated.

What to do in practice?

There are few details on policy consequences, but one scenario is already being discussed. According to reports, the US Department of Commerce is preparing a rule that would restrict Chinese companies' access to foreign AI data centers. It would apply only in some countries, only to future shipments, and only to Blackwell-class chips and above. By one estimate, it would affect no more than 14% of Chinese computing in 2027.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…