Insights      Artificial Intelligence      Why Georgian invested in Tenstorrent

Why Georgian invested in Tenstorrent

Over the last few years, we have observed how AI is significantly reshaping the tech stack, particularly by strengthening the relationship between the hardware and software layers. The scaling laws of AI – that hardware appears to govern the size and capabilities of AI models – help demonstrate this interconnection. At Georgian, we are watching how AI is impacting the whole tech stack, looking for companies with the right combination of technical vision and operational maturity.

Enter Tenstorrent, a semiconductor company that designs and sells chips specialized for AI workloads (called AI accelerators). The company’s first generation of hardware comes in a variety of form factors, ranging from standalone chips to developer-oriented desktops to servers. These direct hardware sales are complemented by the company’s commercial IP portfolio, which sells Tenstorrent technology into more customizable form factors. The company also provides two software libraries that are intended to make its hardware accessible to developers.

These are some of the reasons why Georgian has participated in Tenstorrent’s $693M Series D, alongside Samsung, AFW Partners, Fidelity, XTX Markets, LG Electronics, HOOPP and others.

Led by industry veterans Jim Keller and Keith Witek, the Tenstorrent management team brings together highly-experienced professionals with significant hardware expertise. Keller, Tenstorrent’s CEO, was previously Chief Architect at AMD where he worked on the Zen line of chips, a product credited for the company’s successful turnaround.[1] He also oversaw the creation of Tesla’s hardware division and was a VP of Engineering at Apple during the launch of the company’s first in-house designed iPhone chip.[2] The company’s COO, Keith Witek, complements Keller’s technical vision, with experience from Google, Tesla and AMD. Chief Architect Wei-han Lien led microarchitecture design for Apple’s A6 and A7 CPUs.[3]

In our view, under Keller’s leadership, Tenstorrent has been remarkably forward thinking. With hardware, product design typically starts years before a shipped product. Design choices anticipate the future. Through the course of our diligence, we saw that Tenstorrent has positioned itself for a number of evolving trends:

  • Built for AI: There is growing demand in the market for specialized AI processors, particularly as AI workloads shift from creating models to serving them. In designing chips, Tenstorrent took a first principles approach, focusing on data flow and compatibility with the models of tomorrow in the fast moving world of AI.
  • Scalability: Tenstorrent’s technology is designed to scale the same building block, the company’s proprietary AI Tensix cores, across use cases from small form factors (the ‘edge’) to data centers. The company’s upcoming hardware furthers this scalability with a chiplet design – smaller, modular chips that can be mixed-and-matched together into custom sizes and capabilities. 
  • Open source: Tenstorrent’s chips use an instruction set architecture called RISC-V.  Unlike its competitors, Arm or x86, RISC-V is open source. There are indications that the broader market is seeking an open source alternative with reduced licensing requirements.[4] Moreover, we see RISC-V’s growing adoption in universities as a potential opportunity, seeding relevant skills and experience in the next generation of engineers.[5]
  • Markets of the future: Tenstorrent’s IP business has established partnerships across sectors, including automotive and consumer electronics.  For instance, Tenstorrent’s partnership with LG enables Tenstorrent technology to power smart TVs.[6] While these use cases are extensions of today’s technology, we think that, over time, these use cases may evolve into areas like IoT and autonomous vehicles. 

In assessing AI hardware, our perspective is that technical wherewithal is not enough; competency across operations, capital, supply chain, IP and other areas are also important elements. We think that Tenstorrent embodies this approach, which is reflected in the collective industry experience of its management team.

AI, chiplets, open source – these were decisions Tenstorrent made years ago. We are delighted to partner with Tenstorrent and look forward to watching these anticipations unfold.

  1. Katie Tarasov, CNBC: How AMD became a chip giant and leapfrogged Intel after years of playing catch-up, Nov 29, 2022.  
    Majeed Ahmad, EDN: The story of Jim Keller and his pioneering work on chip design and architecture, Jan 9, 2021.
  2. Dr Ian Cuttress, AnAndTech: Interview with Jim Keller: “The Laziest Person at Tesla”, Jun 17, 2021.
    Aaron Pressman, Why Intel is betting its chips on microprocessor mastermind Jim Keller, May 18, 2020.
  3. The A7 was the world’s first 64-bit mobile chip. Daniel Dilger, Apple Insider: A7: How Apple’s custom 64-bit silicon embarrassed the industry, Dec 2, 2019.
  4.  Ian King, Bloomberg: Arm to Scrap Qualcomm Chip Design License in Feud Escalation, Oct 22, 2024.
  5. For a sample of universities incorporating RISC-V in their curricula, see: DFRobot, RISC-V: Revolutionizing Higher Education and Computer Science Curricula, Feb 14, 2024.
  6.  Tenstorrent, Tenstorrent Partners with LG to Build AI and RISC-V Chiplets for Smart TVs of the Future, May 30, 2023.

Read more like this

Why Georgian Invested in You.com

We are pleased to announce that Georgian has led You.com’s $50M Series…

Why Georgian Invested in Coder

We are excited to announce that Georgian has led Coder’s $35M fundraise…