Skip to main content
Back to News
news/AI Models

Nvidia Reaffirms Open Models and Keeps Hugging Face Deal

Nvidia's interim CEO says the open model strategy and the $12.9B Hugging Face deal are unchanged, with Nemotron topping US open models on OpenRouter.

Stefan Trbojevic

Stefan Trbojevic

18 November 20263 min read
LinkedIn
Two dark glass cubes with warm amber light between them, representing two companies merging

The takeaway

Nvidia confirmed its open model strategy and the Hugging Face acquisition survive the leadership change, so open weights remain a safe planning assumption for builders, with concentration risk as the tradeoff.

Why it matters for builders

Open weight model access and the Hugging Face roadmap stay intact, so agent stacks built on those assumptions do not need migration, but teams should keep a second model provider because one vendor now controls the registry, the compute and a growing share of the models.

Nvidia Reaffirms Open Models and Keeps Hugging Face Deal

Nvidia's interim chief executive used her first earnings call since Jensen Huang's resignation to answer the question the open source community had been asking for a week: the company's open model strategy is not changing.

"At Nvidia, we are committed to fostering a vibrant open model ecosystem, so every country, company, and industry can build its own intelligence," Colette Kress told analysts on Tuesday. She added that the $12.9 billion Hugging Face acquisition, agreed in September, remains on track, and that together the two companies will "scale its platform and make intelligence more open, more capable, and more accessible."

What Kress Committed To

Two things were on the line. Huang had been the loudest internal advocate for open weight models, personally pushing the Nemotron family built for long-running agentic workloads. His exit over undisclosed relationships, including one with an executive suspected of working for Chinese intelligence, raised the obvious question of whether Nvidia would quietly retreat to closed, margin-friendly systems.

Kress answered that directly, framing open models as an adoption engine rather than a charity project. The developer base Nvidia is buying makes the point: Hugging Face serves more than 18 million developers and 200,000 companies.

The Hugging Face deal was not raised during the Q&A portion of the call, and no updated closing date was given. That silence is the honest signal to watch. A reaffirmed strategy in prepared remarks is a commitment; a completed acquisition is a fact.

What the Numbers Say

Abstract glass layers rising like a staircase, illustrating stacked open model layers

Kress also put figures behind the open model pitch. Weekly token volume on OpenRouter is up 35-fold year over year, including a 3.5-fold increase in the past three months alone. Nemotron, she said, is now the most widely used US open model on that router, and it is "where developers start to build their own intelligence."

That claim matters beyond bragging rights. OpenRouter is the closest thing the industry has to a public scoreboard for model demand, and it measures what builders actually run, not what vendors report.

Why Builders Should Care

For teams shipping agents, the practical read is that current stack assumptions hold. Open weight Nemotron models stay available for fine-tuning and self-hosting, the largest model hub on the internet stays inside a hardware vendor that benefits from distribution, and pricing pressure from open models keeps disciplining closed API costs.

The risk is concentration. One company would own the dominant model registry, a large share of the compute it runs on, and a growing share of the open models themselves. That is a single point of failure for anyone whose architecture depends on Hugging Face as neutral infrastructure.

The counterweight is legal, not technical. The SEC is reportedly investigating Nvidia's disclosures around Huang's departure, and Congress still has chip export legislation in motion, both covered in our roundup of November seventeen and in our analysis of the export enforcement era.

Share𝕏

The Automation Brief

Read 5 AI stories instead of 50.

The essential moves in AI agents, models, automation and infrastructure — filtered for builders and operators, with the part that actually matters.

No noise. Unsubscribe anytime.

Editorial notes

Reported by

Stefan Trbojevic

Edited by

n8n Lab Editorial

Published

18 November 2026

Updated

18 November 2026

AI disclosure: AI assisted with research and drafting. Factual claims are reviewed by an editor.

n8n Lab is an independent service provider. We are not affiliated with, endorsed by, or sponsored by n8n GmbH. “n8n” is a trademark of n8n GmbH and is used here only to describe the platform-specific implementation and automation services we provide.