A $6 Billion Signal: Open AI Chips Are Challenging Nvidia

Nearly every AI tool you use today runs on hardware made by one company: Nvidia. That near-monopoly is now being seriously challenged. Yesterday, chip startup Tenstorrent closed a $1.5 billion funding round, valuing the company at $6.2 billion — more than double what it was worth just two years ago. Tenstorrent is ten years old and spent most of that time as an obscure underdog. That is rapidly changing.

Tenstorrent, led by celebrated chip designer Jim Keller, is building something specific: chips made for open-weight AI models. Open-weight models — like Meta's Llama or DeepSeek — are AI systems that any company can download and run on their own hardware. Instead of paying a fee every time you ask an AI a question, you run the model yourself. Tenstorrent demonstrated its Galaxy servers running DeepSeek at speed, and that demo helped close the round.

What makes this especially significant is what Tenstorrent turned down. Earlier this year, Intel and Qualcomm reportedly offered to acquire the company for up to $5 billion. Tenstorrent said no and raised independently instead. That decision tells you something: the people building this technology believe the open AI infrastructure market is worth far more than any buyout. Staying independent means building for the open-weight world on their own terms.

For anyone building software or running a business that uses AI, the practical meaning is clear. Running your own AI models is about to get easier and cheaper. Until recently, self-hosting a capable model required deep technical expertise and expensive Nvidia hardware. As companies like Tenstorrent scale up alternatives, that barrier drops. More competition in chips means lower costs and more control for the teams that use them — and that is good news for every builder working with AI today.