Nvidia, Stripe Spend $26B on Open-Weight AI Acquisitions

Tech giants are making massive bets on open-weight AI models, with more than $26 billion in acquisitions announced or reported in August alone.

Nvidia is reportedly closing a $13 billion deal to acquire Hugging Face, the platform that has become the de facto hub for sharing and deploying open-weight AI models. The acquisition follows Nvidia's $6 billion agreement with Poolside, an open-weight model builder, and Stripe's purchase of OpenRouter for more than $7 billion two weeks earlier, according to TechCrunch.

The spending spree reflects a strategic shift as major tech companies seek alternatives to dependence on frontier labs like OpenAI and Anthropic.

Why it matters

These acquisitions signal that the AI market is diversifying beyond the dominant proprietary model providers. As companies face rising inference costs and seek more control over their AI infrastructure, open-weight models offer a hedge against vendor lock-in. The deals also reveal competitive pressure on Nvidia from model builders developing their own chips, including OpenAI's newly announced Jalapeño inference chip.

Strategic motivations behind the deals

For Nvidia, the Hugging Face acquisition addresses a critical vulnerability. Major AI labs including OpenAI and Google are building their own inference chips, threatening Nvidia's dominance in AI hardware. By controlling the largest U.S. developer platform for open models, Nvidia gains direct access to a user base it can steer toward its chips and technical standards.

Nvidia already produces its own Nemotron family of open-weight models, but adoption has been limited. Hugging Face brings an established ecosystem of developers building and deploying models outside the frontier labs.

Stripe's OpenRouter acquisition targets a different opportunity. CEO Patrick Collison framed the deal around efficiency: "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."

Current adoption remains limited

Despite the acquisition activity, open-weight model adoption remains modest. Just 6% of companies use open-weight models according to spending data analyzed by Ramp, while only 2% of software engineers surveyed by Jellyfish use them.

Nik Albarran, AI product lead at Jellyfish, told TechCrunch that open-weight models currently serve specific use cases—primarily high-volume, repetitive inference tasks like customer service chatbots where models can be tuned for efficiency. For coding and agentic workflows requiring varied reasoning, proprietary frontier models still dominate.

"There are not many companies where that is the case yet…[but] if the prices continue to go up from the frontier labs, more and more companies will be forced to at least consider it," Albarran said.

The long-term bet on model diversity

Lin Qiao, CEO of Fireworks—another leading open-weight model platform often mentioned as an acquisition target—argues the market is moving toward specialized models. Her company processes 40 trillion tokens daily, exceeding both Gemini and OpenAI's API volumes.

"Every single app company should consider hiring an in-house researcher," Qiao told TechCrunch. "They can use their product and product data to build their own model. The future is actually specialized intelligence."

The acquisitions suggest tech giants are preparing for a future where AI capabilities are more distributed across multiple model types rather than concentrated in a few frontier labs.

The details were first reported by TechCrunch.