US AI chip startup Tenstorrent raises $1.5b to scale hardware

Tenstorrent raised US$1.5 billion in a new funding round. That round values the US chip startup at US$6.2 billion. It will use the money to expand hardware for open-weight AI models and video generation. The company focuses on inference, meaning running trained models.

A concrete example is Tenstorrent’s Galaxy servers, which ran Pramana models. That result helped the company choose independence over reported acquisition interest from Intel and Qualcomm. The funding can support more hardware capacity and development while keeping control in Tenstorrent’s hands.

The round more than doubled its US$2.6 billion valuation from late 2024. It also supports talks about advanced manufacturing and 2nm technology. Keller said Tenstorrent is discussing production with Intel, TSMC, and Samsung. It is also talking about 2nm technology with TSMC, Rapidus, and Samsung. Practically, the money should help turn demonstrated performance into a larger hardware business.

An AI chip startup is a young company that designs specialized processors for artificial intelligence. These chips aim to make AI faster, cheaper, or more energy-efficient than general-purpose processors. AI inference means using a trained model to produce an answer, prediction, image, or video. Training teaches the model; inference uses what it learned.

Tenstorrent says its Galaxy servers can run Pramana models, demonstrating its hardware’s inference role. A server receives a request, runs the trained model, and returns a result. Tenstorrent is targeting open-weight AI models and video generation, where handling many requests efficiently can matter.

Tenstorrent raised US$1.5 billion at a US$6.2 billion valuation, more than doubling its US$2.6 billion valuation from late 2024. The company plans to expand hardware and is discussing 2nm technology with TSMC, Rapidus, and Samsung. Its focus positions it to compete on practical AI use, not only model creation.

The company raised a huge $1.5 billion, lifting its valuation to $6.2 billion. Its valuation increased by $3.6 billion from late 2024. That is more than double its earlier $2.6 billion value.

The funding will help Tenstorrent expand hardware for open-weight AI models and video generation. Its Galaxy servers reportedly ran Pramana models, helping convince the company to remain independent. Trace Capital and Innovation X led the round, rather than an acquisition by interested technology companies.

Intel and Qualcomm reportedly considered buying Tenstorrent at valuations reaching $5 billion. The new financing therefore places Tenstorrent above those reported acquisition levels. Keller also says Tenstorrent is discussing chip production with Intel, TSMC, and Samsung. It is separately discussing 2nm technology with TSMC, Rapidus, and Samsung.

Tenstorrent stayed independent because it had demonstrated that its own Galaxy servers could run Pramana models. The article connects this decision with that demonstration, but does not provide a more detailed reason.

Galaxy servers are Tenstorrent’s hardware platform for running AI models, especially during inference. After showing Galaxy could run Pramana models, the company chose independence despite reported acquisition interest from Intel and Qualcomm.

The decision brought a $1.5 billion funding round at a $6.2 billion valuation. That valuation more than doubled from $2.6 billion in late 2024, giving Tenstorrent capital to expand hardware for open-weight models and video generation. Keller has also said Tenstorrent is open to chip production with Intel, TSMC, and Samsung. So independence means pursuing partnerships without becoming part of an acquiring company.

Tenstorrent does not have to manufacture every chip itself. It can work with outside manufacturers such as Intel, TSMC, Rapidus, or Samsung to turn its designs into physical chips. This gives the startup more production options as it scales hardware for open-weight AI models and video generation.

The company is already discussing 2nm technology with TSMC, Rapidus, and Samsung. It has also said it is open to producing chips with Intel. These partnerships could provide advanced manufacturing, factories, and production expertise, while Tenstorrent focuses on chip design and AI performance.

Tenstorrent’s choices remain under discussion, rather than finalized in the article. The company raised $1.5 billion at a $6.2 billion valuation, giving it resources to expand. Working with several manufacturers could reduce dependence on one supplier, but no partner or manufacturing agreement has been confirmed.

“2nm technology” refers to an advanced manufacturing generation, not necessarily transistors exactly two nanometers wide. Smaller designs can place more transistors in the same area, potentially delivering more computing power while using less energy.

Tenstorrent showed that its Galaxy servers could run Pramana models, highlighting its focus on AI inference. If future Tenstorrent chips use 2nm, denser circuitry could perform more calculations per second. It could also reduce energy used for each result. That matters for large AI systems, where electricity and cooling are major costs.

Tenstorrent is discussing 2nm technology with TSMC, Rapidus, and Samsung, but the article does not say its chips already use it. Manufacturing at this scale remains difficult, so benefits also depend on chip design, production yields, memory, and software. Its $1.5 billion funding could help expand hardware for open-weight AI models and video generation.