Tenstorrent is a Toronto-based AI hardware startup that develops AI processors and training servers. The company plans to use its new funding to build out its engineering team and to build artificial-intelligence training servers to demonstrate its technology. CEO Jim Keller, a famed microprocessor engineer, said Tenstorrent plans to release a new AI processor every two years. The company has signed customer contracts totaling nearly $150 million. Tenstorrent was founded in 2016 and is one of several startups aiming to chip away at Nvidia’s dominant AI chip market share. Other startups mentioned alongside Tenstorrent include Axelera, Etched and Groq. The company recently closed a large financing that values it at more than $2.6 billion. Tenstorrent is a Toronto, Canada–based computing company that designs and sells AI processors and licenses AI and RISC‑V IP to customers who want to own and customize their silicon. Led by CEO Jim Keller, the company brings together experts in computer architecture, ASIC design, advanced systems, and neural network compilers. It operates U.S. offices in Austin and Silicon Valley and has global offices in Belgrade, Tokyo, and Bangalore. Tenstorrent closed a $100M strategic financing up‑round. The company intends to use the funds to accelerate product development, design and development of AI chiplets, and advance its ML software roadmap. Tenstorrent is a hardware startup developing next-generation computers and AI-focused processors. Its flagship Grayskull processor was slated to go to market in the second half of 2021. The company is also launching DevCloud to let developers run models without purchasing hardware. Larger systems using Tenstorrent’s technology will be available through partners. The company said it will use the new funding to continue creating AI-focused hardware for developers. Tenstorrent is headquartered in Toronto and maintains U.S. offices in Austin and Silicon Valley. Leadership includes CEO Ljubisa Bajic and CTO Jim Keller. Tenstorrent develops Grayskull, an all-in-one AI compute system designed to accelerate model training and inference using a novel architecture that eliminates unnecessary computation. Grayskull implements conditional execution and packs 120 proprietary Tensix cores stitched with a custom torus interconnect, 120MB of local SRAM and up to 16GB external LPDDR4 across PCIe Gen4. The Tensix core combines a packet processor, a programmable SIMD processor, a dense math block, and five RISC cores to drive high utilization. In preliminary tests the system reportedly hit 368 TOPs at a 75W PCIe-card TDP and processed 23,345 sentences/sec on BERT-Base (SQuAD 1.1), a claimed 26x advantage versus competing solutions. Tenstorrent is sampling Grayskull with partners and expected the system to be production-ready in fall 2020, and plans to target datacenters, public and private cloud servers, on-premises and edge servers, and automotive markets. The company is headquartered in Toronto with offices in Austin and Silicon Valley and has raised over $34 million to date.

Total raised

$1.0B

Funding rounds

4

Latest round

Series D

Latest activity

Dec 2024