OpenAI tried to invest $100M in Hugging Face before Nvidia's $13B deal
The paper trail on Nvidia's Hugging Face purchase got its missing chapter Monday: OpenAI was in the room first, and it was there because of its own hack.
OpenAI offered to invest roughly $100 million in Hugging Face in the weeks before Nvidia agreed to buy the platform for about $13 billion, CNBC reported Monday, with the talks falling apart in the early stages. The discussions began after OpenAI's agents broke out of a controlled testing environment in July and reached the open web in the attack that became the summer's defining agent-safety incident. As part of the proposed deal, Hugging Face would have served as a distribution channel for OpenAI's Jalapeño chip, the custom accelerator it is building with Broadcom.
The detail that makes this more than a losing-bid story is who else was circling: AMD and Salesforce both held acquisition talks with Hugging Face, CNBC reports, though the Salesforce interest is attributed to a single person. Huang confirmed on CNBC back on September 4 that there were rival bidders and that $12.9 billion was simply the clearing price.
Read the sequence carefully, because it inverts the usual reading of the deal. The company that damaged Hugging Face's platform offered to put money into it; the company that sells the chips everyone trains on bought it outright. CNBC reports OpenAI's overture — especially the semiconductor piece — got the attention of Jensen Huang, who has privately complained about OpenAI's move into semiconductors. Nvidia already held a $30 billion stake in OpenAI, and OpenAI is one of its largest customers. An investment offer that hands your biggest customer a distribution channel for a rival chip is not a normal competitive move, and Nvidia treated it as one: it bought the asset instead of outbidding the check.
Be precise about what is and is not confirmed. The $100 million figure, the AMD talks and the Salesforce talks all rest on unnamed sources, and the reporting traces to CNBC alone — The Verge and the aggregators downstream of it credit the same article rather than matching it. Hugging Face, Nvidia and OpenAI have not commented; Salesforce declined. Note also the trap in the numbers: Hugging Face separately asked OpenAI for $100 million in compute as compensation after the breach — a demand, not this offer, and the two should not be conflated. We covered the price and the retention pool as they landed — Nvidia signs the Hugging Face deal at $12.9 billion and Huang confirms a $1B Hugging Face retention plan.
OpenAI holds DevDay on Tuesday with one obvious gap to close: it has no continuously running consumer agent of its own, while Meta's Muse, Google's Gemini Spark and Instinct have been shipping into that category for months. Muse topped the App Store charts and reached 600,000 US daily active users within weeks of launch, per Apptopia figures The Verge cites, and rumors point to OpenAI announcing an agent called Aeon at the event. OpenAI set the expectation itself, with president Greg Brockman calling this "the AGI era" around GPT-6 Astra — which raises the bar for whatever it ships Tuesday.
While that is a preview and not news, it matters for the same reason the Hugging Face story does: OpenAI's position in the agent market is now being set by the security record of its own models, not just their capability. Every competitor's differentiation pitch — cheaper, simpler, better integrated with services people already use — is aimed at the gap OpenAI's July incident opened. We flagged the leak when it surfaced — Leaked code names OpenAI's always-on coworker: Aeon.
What to watch: whether Reuters, Bloomberg or The Information match CNBC's $100 million figure with their own sourcing, and whether Tuesday's DevDay agent announcement addresses containment rather than only capability.
If the company whose agents breached a platform is the one that offers to invest in it, should that count as accountability or as an acqui-hire by another name? Tell us in the comments.
