Samsung Foundry to Build Tenstorrent’s Next-Generation AI Chips
The South Korean tech company said on Monday that Tenstorrent, a Canadian startup, has become a new client for Samsung’s contract chip manufacturing division for artificial intelligence chips. Tenstorrent is one of several firms attempting to overtake Nvidia, the industry leader in AI processors. The business supplies semiconductors and intellectual property to data centers, but it also works to supply businesses like the automobile industry. As a part of the agreement, Tenstorrent intends to manufacture the chips using one of Samsung’s cutting-edge manufacturing techniques, known as 4nm.
A chiplet, Tenstorrent, is a Samsung product that is made to fit within a single package with other chiplets. RISC-V, an open-source semiconductor architecture used in some of Tenstorrent’s chips, is a rival to the Arm and x86 architectures used by Intel (INTC.0) and Advanced Micron Devices (AMD.0). RISC-V is identified as an open-standard architecture, and the idea is that by working together, the member businesses may advance new horizons in processor innovation while encouraging new levels of design flexibility. However, the chip that Samsung will produce is not based on RISC-V technology, and it is named the Quasar chip.
In August, Samsung invested in Tenstorrent as part of a $100 million capital raising that also included Hyundai Motor Co. (005380.KS) and other companies. This chip transaction comes after Samsung’s investment. Tenstorrent had a $1 billion value and had received $234.5 million prior to the fundraising round in August. Both the amount of the deal and the anticipated output of chips were kept a secret by the two companies.
CEO of Tenstorrent, Jim Keller, opened up, Tenstorrent’s focus is on developing high-performance compute and delivering these solutions to customers around the world.”
Tenstorrent sells AI chips and licenses AI software solutions and intellectual property (IP) based on RISC-V, an open source instruction set architecture used to create bespoke processors for a variety of applications.