Enki

Nvidia's open source push faces test after Huang exit

By

Published

Reporting from TechCrunch

More onNvidia

Jensen Huang's exit puts Nvidia's $12.93 billion Hugging Face deal and open model work in doubt. Hugging Face's CEO says it will keep fighting.

What it means for founders

If you serve or fine-tune open weights, three things to track.

  • Hosting terms. The pledge that Hugging Face stays chip neutral is Nvidia's, made in a post under Huang's name. Watch whether Kress repeats it, and whether any DOJ conditions write it into the deal. Keep your own copies of the weights you ship, so a policy change on the hub does not stall your product.
  • Nemotron cadence. An Nvidia insider told TechCrunch they do not expect the leadership change to affect Nemotron development. Treat the next release date as the test before you plan a roadmap around it.
  • Policy exposure. TechCrunch argues the scandal could strengthen the case in Washington for tougher limits on Chinese AI companies, and Huang was among the most prominent opponents of restricting the open AI ecosystem. Teams building on Chinese open models like Kimi or DeepSeek should plan for the chance of new rules and test a US alternative now.

What we don't know yet

TechCrunch says it is unclear how Kress will handle the Hugging Face deal or Nvidia's wider support for open models. Nor do we know what, if anything, the Justice Department will demand before the purchase closes.

The story

Nvidia's biggest open source bets now sit with a leader who did not make them. Jensen Huang resigned as CEO on Thursday, days after his relationship with Alice Huang, who ran Singapore data center firm Megaspeed, became public. CFO Colette Kress is interim CEO. CBS News cites a New York Times report that U.S. surveillance suggested she was passing information to Beijing. Megaspeed says it followed U.S. law, and President Trump said Huang told him he knew of no intelligence links. TechCrunch reports that the open model community now asks whether Nvidia keeps backing open weights without him.

What Huang built

The largest piece is Hugging Face. In a September 3 post on Nvidia's blog, Huang announced a $12.93 billion agreement to buy the model hub. He promised it would stay an open platform where developers pick their own models, clouds and chips, and wrote that Nvidia hardware would not be required to build or deploy there. Nvidia says it has published more than 500 models and 250 open datasets on the site, and that the hub serves over 18 million users and 200,000 companies.

The deal has not closed. TechCrunch says it is still set to finish in 2027's first half, and it cites an October Reuters report that the Justice Department began asking how the purchase would affect competition.

Huang's open push reached further. Nvidia agreed to a $6 billion deal for a non-exclusive license to Poolside's coding technology, it invested in Reflection AI, maker of the Beam open-weight model, and in Thinking Machines Lab, and in October it added a 120 billion parameter Nemotron model for enterprise agents plus two small Bolt models for PCs. In July he published a letter, with Microsoft, Meta, IBM and Hugging Face itself among its 25 corporate signers, asking Washington to avoid over-restricting open-weight models.

Hugging Face's response

Clem Delangue, Hugging Face's CEO, posted on X that the news was sad and that "we'll keep fighting for open-source AI, harder than ever," as TechCrunch reported. TechCrunch says its questions on the deal status, and on whether Hugging Face has talked to Kress, went unanswered.

TechCrunch frames open source as business strategy for Nvidia. Big customers such as Google and OpenAI design their own chips, so a wider field of companies training and running models means more GPU buyers.

Sources

More in Models & Labs

How Enki covers newsCorrectionsReport an error