Funding a New Rival in AI Hardware
The shift is happening right on the balance sheets of Silicon Valley as artificial intelligence infrastructure expands. AI chip startup Tenstorrent has closed a $1.5 billion Series E funding round, according to Bloomberg and BetaKit. This new capital pushes the company valuation to $6.2 billion, more than double the $2.6 billion it held during its previous funding round in 2024. The round was headed by Trace Capital and Innovation X, with participation from Agostinelli Group, Sandisk, Tether Investments, T. Rowe Price Group, Cadence Ventures, Ark Invest, and Fidelity, according to Bloomberg.
Before closing this massive equity round, Tenstorrent drew takeover interest and held talks with Intel and Qualcomm about a possible sale, according to Bloomberg and Startup Fortune. Chief Executive Officer Jim Keller stated that the company opted to bring in new funding rather than sell itself. Keller pointed to a May demonstration proving that company Galaxy servers could handle DeepSeek AI models as a turning point, according to Bloomberg. “Once we had the ability to fundraise and demonstrate the performance and started to ship, the investor interest came in,” Keller said, according to Bloomberg. “Many of our investors and I personally think there’s an opportunity to grow quite a bit. I think that’s the best return for the investor and the employees.”
Building Open Architecture Against Nvidia
Tenstorrent builds processors, computers, and software to compete with larger rivals like Nvidia, Advanced Micro Devices, and Broadcom, according to Bloomberg. Founded in 2016 in Toronto by Ljubisa Bajic, Milos Trajkovic, and Ivan Hamer, the firm employs 1,150 people and offers customizable hardware and software solutions, according to BetaKit and Bloomberg. The company relies on the open-source RISC-V architecture, avoiding licensing fees tied to other instruction sets while providing flexibility for clients who want to customize silicon, according to Startup Fortune and BetaKit.
The company plans to use the new capital to fund volume manufacturing of its Galaxy servers and support global expansion, according to BetaKit. Keller has contributed to chip programs at firms including AMD, Apple, Intel, and Tesla before joining Tenstorrent, bringing industry experience to the startup’s hardware efforts, according to Bloomberg and Startup Fortune. The firm sells AI compute at multiple levels, ranging from single cards and workstations to racks and superclusters, serving customers that include chip designers, sovereign infrastructure providers, and inference platforms, according to BetaKit.
Scaling Up for General-Purpose AI
Watch this space as Tenstorrent works to meet growing demand for hardware that can adapt to changing models. “People are looking for alternatives,” Keller said, according to Bloomberg, adding that customers want general-purpose AI computers able to run any type of model. The company also licenses its AI and RISC-V intellectual property to clients who want to own and customize their silicon, such as LG and Samsung, according to BetaKit.
Serving a wide range of customers, Tenstorrent takes on orders that may be too small for larger competitors, tailoring either software or hardware to what a given client requires, according to Bloomberg. “People say, ‘How do you grow to be a $10 billion business?’ Well, you start with $10 million and $100 million, then $500 million,” Keller said, according to Bloomberg. “Our mission is to grow organically and meet the customers where they are and figure out how to service them.” With fresh capital secured and volume manufacturing underway, what’s changing is the viable path for startups attempting to challenge traditional graphics processing unit dominance in AI data centers.
Sources
- AI chip startup Tenstorrent raised $1.5 billion in Series E funding
- Intel and Qualcomm Both Court Jim Keller's AI Chip Startup Tenstorrent – Startup Fortune
- Tenstorrent secures $1.5 billion USD at $6.2-billion valuation to challenge GPU dominance
Photo: Jim Keller (CC BY-SA 4.0)
