Open-weight AI firms have emerged as some of the most sought-after acquisition targets in the tech industry, as major players race to secure their place in the growing market for AI models that aren’t controlled by frontier labs. The most closely watched deal centers on Nvidia, which is reportedly preparing a EUR 11 billion acquisition of Hugging Face, the platform widely used for sharing open-weight AI models and benchmarks.
Often described as a kind of GitHub for the AI era, Hugging Face sits at the center of a broad ecosystem of developers who build and deploy large language models independent of the biggest research labs. If confirmed, the deal would give Nvidia direct access to the largest U.S. developer community for open models.
The reported Hugging Face talks follow a string of similar moves. Nvidia recently reached a EUR 5 billion agreement with Poolside, an open-weight model builder, in a deal that would bring most of its employees into the chipmaker. Two weeks earlier, Stripe acquired OpenRouter, a leading supplier of open-weight models to businesses, for more than EUR 6 billion.
Why Chipmakers and Payment Giants Are Buying In
The wave of spending reflects shifting dynamics across the AI landscape. For Nvidia, the acquisitions offer a way to reduce dependence on deals with major hyperscalers and frontier labs. That concern is heightened as companies like OpenAI and Google develop their own inference chips, including OpenAI’s newly detailed Jalapeño processor. As model builders move into hardware, Nvidia is pushing further into the model-making business.
Nvidia already produces its own Nemotron family of open-weight models, though adoption has been limited. Controlling a major open-model developer hub would give the company a large user base it can steer toward its chips and standards.
Cost is another driving factor. Rising questions about the price of AI inference have pushed some companies to explore cheaper models from Chinese firms such as Moonshot, DeepSeek, and Alibaba. Even so, adoption remains modest. Just 6% of companies use open-weight models, according to spending data compiled by Ramp, while only 2% of software engineers use them, per Jellyfish, a developer tools maker.
Where Open Models Fit Best
According to Nik Albarran, AI product lead at Jellyfish, open-weight models are mostly used by companies running repeated inference workloads, such as customer service chat systems. Because those tasks are high-volume and repetitive, an open model can be tuned to answer questions at a lower cost.
Stripe framed its OpenRouter deal in similar terms. “Tokens are the central currency for companies building with AI, and it’s clear that the real-world economic potential will depend on making good use of scarce compute resources,” said Patrick Collison, Stripe’s co-founder and CEO.
For coding and agentic tasks, however, frontier models often prevail, thanks to easier access and, in some cases, token subsidies from proprietary labs. Albarran said companies currently favor open models mainly for control and configurability rather than cost, but that could change. “If the prices continue to go up from the frontier labs, more and more companies will be forced to at least consider it,” he said.
Source
Image: techcrunch.com
