Chip Disruption: 3 Canadian Challenger Threatens Nvidia’s AI Dominance with Affordable Tech

• Trump says Taiwan owes defense
• Trump’s comments drop Nvidia, TSMC
• Tenstorrent offers affordable, efficient AI
• Tenstorrent aims to cut AI costs
• Tenstorrent rivals Nvidia with efficiency
• Tenstorrent partners with Hyundai, LG, Rapidus
• Tenstorrent boosts ties with Japan

Trump says Taiwan owes defense

Chip Disruption: 3 Canadian Challenger Threatens Nvidia’s AI Dominance

Amid investor reevaluations of Nvidia and TSMC following Donald Trump’s comments on Taiwan, Tenstorrent, a Canadian company specializing in AI chip, software, and hardware design, emerges as a notable competitor to TSMC. Unlike TSMC, Tenstorrent collaborates with integrated circuit foundries in South Korea and Japan.

In a Bloomberg Businessweek interview from July 16, Trump remarked, “I have great respect for the people of Taiwan. They have taken nearly all of our chip business, and I believe Taiwan should contribute to our defense costs, as they currently don’t provide us with anything in return.”

 

Trump’s comments drop Nvidia, TSMC

Chip Disruption: 3 Canadian Challenger Threatens Nvidia’s AI Dominance: Trump criticized Taiwan, saying, “They took our chip business from us. How could we be so foolish? They’ve become incredibly wealthy, and it feels like we’re just an insurance policy for them. Why are we doing this?”

Following these remarks, Nvidia and TSMC saw their share prices fall by over 8% within the week.

Tenstorrent, which is privately held and not reliant on TSMC, was founded in 2016. The company, headquartered in Toronto, has additional offices in Santa Clara and three other U.S. locations, as well as Bengaluru, Beijing, Gyeonggi-do, Tokyo, and Belgrade.

 

Tenstorrent offers affordable, efficient AI

The company, now with around 460 employees, is expanding quickly and currently has 45 job openings, including roles like CPU architect, CAD engineer, systems software engineer, and account executive listed on its website.

CEO Jim Keller, who joined Tenstorrent in 2020, is renowned for his work in integrated circuit design at AMD, Apple, Tesla, and other major tech firms. His approach focuses on creating AI solutions that are both more efficient and less expensive compared to Nvidia’s offerings.

Chief CPU Architect Wei-han Lien, also a former AMD and Apple veteran, recently told DigiTimes, “Customers don’t need to pay premium prices like those of a Porsche or Ferrari to run their Generative AI models; they just need affordable solutions that perform efficiently.”

 

Tenstorrent aims to cut AI costs

Chip Disruption: 3 Canadian Challenger Threatens Nvidia’s AI Dominance: Tenstorrent specializes in designing AI Graph Processors, RISC-V CPUs, and configurable chiplets tailored for AI applications. Their products are often more affordable than Nvidia’s because they are designed for cost-effective manufacturing, leverage open-source RISC-V technology, and do not rely on expensive high-bandwidth memory.

Wei-han Lien mentioned, “We’ve been approached by several multinational companies seeking cost-effective solutions, as they find mainstream products too pricey for running inferences.”

In February, when OpenAI CEO Sam Altman estimated that $5 trillion to $8 trillion would be needed to produce enough AI chips to meet future demand, Jim Keller responded on X, claiming he could achieve the same goal for $1 trillion.

 

Tenstorrent rivals Nvidia with efficiency

Additionally, Tenstorrent develops AI servers, PCIe boards, and workstations tailored for AI model developers, complemented by cloud computing services. The company asserts that its Galaxy AI compute platform is three times more efficient than Nvidia’s DGX platform while costing a third less.

Their TT-Buda and TT-Metalium software development kits, which support AI model execution on Tenstorrent hardware and allow for custom development, are positioned as competitors to Nvidia’s CUDA software.

Tenstorrent has secured funding from South Korea’s Samsung Catalyst Fund and Hyundai Motor Group, along with several U.S. institutional investors including Fidelity, Eclipse Ventures, Real Ventures, Maverick, and Moore Capital.

In October 2023, Tenstorrent announced that Samsung Foundry will manufacture its RISC-V CPU and AI acceleration chiplets using 4nm process technology, with shipments expected to commence this year from Samsung’s Texas facility. Meanwhile, TSMC is preparing to produce Nvidia’s new Blackwell GPU in Taiwan using its own 4nm process.

 

Tenstorrent partners with Hyundai, LG, Rapidus

Chip Disruption: 3 Canadian Challenger Threatens Nvidia’s AI Dominance: In February of this year, CEO Jim Keller visited Hyundai Motor’s Asan factory in South Korea to meet with executives and engineers. According to the Korea JoongAng Daily, they discussed developing AI technologies for Hyundai’s autonomous vehicles, robots, and air mobility products, such as flying cars.

In March, Tenstorrent COO Keith Witek was appointed to the board of directors at Hyundai Mobis, a leading auto parts maker within the Hyundai Motor Group. Hyundai Mobis is driving advancements in connected and autonomous driving with both hardware and software solutions. Witek brings experience from senior roles at Google, Tesla, and AMD.

Tenstorrent is also collaborating with LG Electronics in South Korea to apply RISC-V AI technology to TVs, consumer appliances, and automotive electronics. In November 2023, Tenstorrent and Japan’s Rapidus, an advanced logic IC foundry venture, announced a partnership to develop AI semiconductor design capabilities. Rapidus, established in 2022, plans to begin pilot production at the 2nm node in 2025 and mass production in 2027 at its new facility in Hokkaido.

Rapidus is supported by major entities such as Sony, Toyota, Toyota’s auto parts affiliate Denso, national telecom carrier NTT, telecom equipment maker NEC, investment firm Softbank, and Mitsubishi UFJ, Japan’s largest bank. Its technology, starting at 2nm and advancing further, is developed in partnership with IBM, aiming to revitalize Japan’s semiconductor industry and reduce its reliance on Taiwan.

 

Tenstorrent boosts ties with Japan

Chip Disruption: 3 Canadian Challenger Threatens Nvidia’s AI Dominance: David Bennett, Tenstorrent’s Chief Customer Officer, emphasizes the significance of Japan to both Tenstorrent and himself. Prior to joining Tenstorrent in 2022, Bennett was CEO of NEC Personal Computers, president of Lenovo Japan, a visiting lecturer at Yamagata University, and a board member at Sanrio, the company behind Hello Kitty.

In February 2024, Japan’s Leading-edge Semiconductor Technology Center (LSTC) announced a project commissioned by NEDO (Japan’s New Energy and Industrial Technology Development Organization) focused on advancing semiconductor manufacturing technology for post-5G telecom infrastructure. LSTC is associated with Rapidus.

The project, titled “Development of edge-AI accelerator implemented in 2-nm logic technology,” aims to integrate Tenstorrent’s RISC-V CPU and chiplet design with an accelerator chip from Japan’s AI Chip Design Center, a research collaboration between the National Institute of Advanced Industrial Science and Technology and the University of Tokyo. Rapidus is working on a Rapid and Unified Manufacturing Service to produce these AI devices quickly.

Tenstorrent plans to establish a high-performance computing design center in Japan to support this initiative and future projects. Additionally, LSTC intends to send around 200 engineers to Tenstorrent in the US for training over the next five years. With funding from Japan’s Ministry of Economy, Trade and Industry, Tenstorrent’s expansion costs will be reduced, and its relationship with Japan will be strengthened. Many of these engineers are expected to return to Japan, helping to address the country’s shortage of IC design talent.