Several prominent AI names are being snapped up by bigger companies in the space.
NVIDIA has agreed to buy Hugging Face for $12.9 billion, according to a report from The Information that cites a person familiar with the deal. Reuters has since corroborated the story, though neither NVIDIA nor Hugging Face has issued an official comment outside business hours. If it closes, this would be one of the largest acquisitions in NVIDIA’s history, and it hands the company control of the platform that has functioned as the default home for open-source AI models, datasets, and tools for the better part of a decade.
The timing is interesting. Just two days before this report surfaced, Business Insider had written that Hugging Face was quietly working with a bank to gauge buyer interest at a valuation north of $13 billion. That process apparently moved fast. It’s also a reversal of sorts for NVIDIA, which had reportedly offered Hugging Face $500 million for a stake last year at a $7 billion valuation, an offer the company turned down. Eighteen months later, NVIDIA is paying nearly double that valuation to own the whole thing outright.
A Pattern Of Consolidation
Hugging Face isn’t the only AI infrastructure name changing hands this year. Stripe recently finalized its own acquisition of OpenRouter, the AI model routing platform, in a deal reportedly worth more than $7 billion. OpenRouter had raised money at a $1.3 billion valuation just months before the deal was struck, so the markup mirrors what’s happening with Hugging Face — companies sitting in the middle of the AI stack, without necessarily building frontier models themselves, are suddenly worth an enormous amount to whoever wants to own the plumbing. OpenRouter’s own CEO, Alex Atallah, had long called his company “the Stripe of AI,” which makes Stripe’s decision to actually buy it feel less like a coincidence and more like Stripe reading its own marketing copy literally.
The common thread across both deals is distribution. Neither OpenRouter nor Hugging Face makes the underlying models that developers care about. What they own is the layer where those models get discovered, compared, routed, fine-tuned, and shipped into production. That layer has quietly become one of the more valuable pieces of real estate in AI, and the companies sitting on top of the actual compute — NVIDIA for chips, Stripe for payments — are the ones moving to buy it before someone else does.
Why Hugging Face, Why Now
NVIDIA’s interest in Hugging Face isn’t new. It was one of the investors, alongside Salesforce and Google, in Hugging Face’s $235 million round back in 2023 that valued the company at $4.5 billion. NVIDIA has also built increasingly deep product ties with the platform since then, including running NVIDIA-accelerated inference for models hosted on the Hub. What’s changed is the scale of NVIDIA’s ambition and the competitive picture around it.
Anthropic and OpenAI, the two labs at the frontier of closed-source models, have both been reported to be developing their own chips, a move aimed squarely at reducing how much they depend on NVIDIA hardware. That’s a real threat to NVIDIA’s business model, because a handful of frontier labs buying enormous quantities of GPUs has been the engine behind the company’s growth. Owning Hugging Face gives NVIDIA a hedge against that concentration risk. Instead of its future being tied to whether three or four frontier labs keep buying chips at the same pace, NVIDIA gets a stake in an entire ecosystem of open-source developers, startups, and enterprises who build on models that are typically trained and served on NVIDIA GPUs anyway.
This lines up with a broader shift NVIDIA has been signaling for a while. In July, Jensen Huang used his very first post on X to share a letter titled “Open Weights and American AI Leadership,” co-signed by more than 20 companies including Hugging Face, Meta, Microsoft, and IBM, arguing that open models are essential to US competitiveness. What stood out at the time was who didn’t sign it: Anthropic and OpenAI were both absent from the initial list. NVIDIA has also been releasing its own open-weight models under the Nemotron line, with Nemotron 3 positioned as one of the most capable open models to come out of the US, shipped with full weights and training recipes on Hugging Face rather than kept behind an API.
Every frontier lab that ships an open model, from Google’s Gemma line to Moonshot’s Kimi releases, tends to land on Hugging Face first. Google’s Gemma 4 went out under an Apache 2.0 license with day-one support across the usual open-source serving stack, and when Moonshot released Kimi K3’s weights, it saw what was reportedly the fastest release growth the platform had recorded. Owning that distribution hub means NVIDIA isn’t just supplying the hardware underneath these releases, it’s sitting at the gate through which nearly all of them pass.
The Open-Source Bet
There’s a strategic logic here that goes beyond simple defensiveness. The more the AI industry standardizes around open weights, the more that spreads NVIDIA’s customer base across thousands of companies rather than concentrating it inside a handful of labs that could, in theory, walk away to custom silicon. A frontier lab building its own chip is a genuine risk to NVIDIA’s revenue. A thousand startups fine-tuning open models on Hugging Face and deploying them on NVIDIA GPUs is not. Every dollar NVIDIA puts into growing the open-source ecosystem is, in a roundabout way, a dollar spent insulating itself from the handful of customers big enough to eventually stop needing it.
That’s also probably why reaction to the deal, at least in its early hours, has skewed positive within the open-source community itself. Hugging Face’s neutrality has been part of its appeal for years, and there’s an obvious tension in a hardware giant now owning the platform that’s supposed to serve every model, including ones built for AMD chips or Google’s TPUs. But NVIDIA has more to gain from keeping that platform open and thriving than from narrowing it, at least for as long as its own dominance in AI hardware holds. Whether that neutrality survives the deal in practice, rather than just in NVIDIA’s messaging around it, is the part worth watching once the ink actually dries.
No terms have been disclosed on deal structure or expected closing timeline, and both companies have yet to confirm the report on record.