AI chip startup Tenstorrent has raised $693 million in Series D funding, positioning itself as a potential challenger to NVIDIA in the competitive AI hardware market. The round, which values the company at $2.6 billion, drew investment from notable backers including Samsung Securities, Amazon founder Jeff Bezos, and Hyundai Motor Group.
Led by celebrated chip architect Jim Keller, Tenstorrent is taking an unconventional approach to AI processing. Unlike NVIDIA's proprietary technology stack, the company is betting on open-source architecture and standard components to deliver more cost-effective AI solutions.
The Context: NVIDIA is currently offering developers an integrated, proprietary technology stack—chips, data center layouts, and everything in between—to ensure optimal performance. By contrast, Tenstorrent is betting on a more modular, interoperable ecosystem that could allow companies greater freedom to integrate AI solutions in the way that works best for them.
Keller says the company's approach focuses on affordability and open-source technology, steering clear of NVIDIA's reliance on high-bandwidth memory (HBM). "You can't beat Nvidia if you use HBM," Keller notes, pointing out that NVIDIA's buying power makes it impossible to compete on cost if you use the same components. Instead, Tenstorrent aims to cut through the complexity and offer something cost-effective that can scale without a massive price tag.
What Sets Tenstorrent Apart: Tenstorrent's chips use RISC-V, an open standard for processors that is beginning to gain traction as an alternative to the dominant ARM designs. This focus on open standards isn't just about technology; it's also a strategic play to attract engineers and developers who prefer the flexibility that comes with open-source platforms.
Yes, but: While Tenstorrent has made promising strides—securing nearly $150 million in customer contracts so far—they are still a fledgling contender compared to NVIDIA's staggering dominance in AI datacenter revenue (which is now in the tens of billions each quarter). But Keller remains undeterred, planning to release new AI processors every two years, compared to NVIDIA's annual release cycle.
The fresh capital will fund engineering team expansion and the development of AI training servers to showcase the company's technology. Tenstorrent is already working with major semiconductor manufacturers, including TSMC and Samsung, on chips using 2-nanometer fabrication technology.
The bottom line: While the funding represents a significant milestone for Tenstorrent, the company's success will ultimately depend on its ability to deliver on its promises of more affordable and efficient AI processing solutions.