RISC-V AI Startup Tenstorrent Gets Another $100M Infusion From Samsung And Hyundai
Tenstorrent is a relatively recent startup that is focused on creating new processors built around the RISC-V ISA and specifically targeted at providing extremely high integer math performance for AI training and inference. We wrote back in April about how optimistic predictions for the performance of Tenstorrent's upcoming Ascalon processor put it in tight competition with AMD's EPYC, NVIDIA's Grace, and Intel's Xeons.
It's stories of that kind that have likely convinced many funds, including two of the largest chaebols in Korea (Hyundai Motor Group and Samsung), to invest another $100 million in the startup. Tenstorrent has already raised over $230 million to this point, so the extra cash infusion is a big one relative to the company's size.
While it may seem surprising at first, both Hyundai's and Samsung's involvement in Tenstorrent make a lot of sense. Tenstorrent makes (or at least, plans to make) advanced AI processors, and it's exactly that sort of hardware that will enable automotive AI applications such as obstacle avoidance and, eventually, fully-automated driving.
Meanwhile, Samsung has its fingers in all kinds of semiconductor pies, and this investment comes from Samsung's "Catalyst Fund," a venture capital group specifically created to "invest in disruptive ideas." While Samsung has at least two different processor design divisions within the company, it never hurts to make friends with more processor manufacturers, particularly given Samsung's capable chip foundries.
Jim Keller describes the investment as "humbling" and remarks on Hyundai's "aggressive adoption of technology" such as the group's purchase of Boston Dynamics (remember those cute robot dogs?) and its joint venture with Aptiv, an Irish autonomous vehicle company with an emphasis on green technology. Similarly, he says he is "excited" by the chance to work with Samsung given its position in the electronics industry.
Tenstorrent says that this round of funding will be used to "accelerate the company's product development, the design and development of AI chiplets, and its ML software roadmap."
