Samsung and Hyundai support AI startup Tenstorrent, according to CEO Keller, as everyone seeks to rival Nvidia.

Samsung and Hyundai support AI startup Tenstorrent, according to CEO Keller, as everyone seeks to rival Nvidia.

The Battle for AI Supremacy: Is Nvidia’s Dominance Sustainable?

Image

Nvidia’s Monopoly: The Quest for Alternatives

In the realm of artificial intelligence (AI), chip giant Nvidia has emerged as the dominant force, overshadowing tech giants like Microsoft, Google, and OpenAI. With its GPU chips serving as the industry’s go-to for AI computing, Nvidia’s control has shown resilience against startup competitors trying to break its monopoly. However, despite its powerful position, industry players are hungry for competition and searching for alternatives to Nvidia.

Speaking exclusively to ENBLE, Jim Keller, CEO of AI chip startup Tenstorrent, shares his concerns about Nvidia’s dominant position: “Nvidia has a monopoly [profit margins]. If you want to build a high-performance solution with AI inside of it, Nvidia will command most of the margin in the product. The problem with the winner-take-all strategy is it generates an economic environment where people really want an alternative.”

While Nvidia’s grip on the AI market seems unshakeable, Keller believes that emerging technology, such as the open-source chip instruction set RISC-V, presents an opportunity for change. Keller’s credibility in the computer-chip world, from his successes at Advanced Micro Devices (AMD) and Apple to his work on Tesla’s Autopilot chip platform, lends weight to his opinions.

Tenstorrent’s Funding and Hyundai’s Interest in Alternatives

To challenge Nvidia’s dominance, Tenstorrent recently secured $100 million in funding from Hyundai Motor Group and Samsung Catalyst Fund, among other investors. The substantial investment, along with previous funding of almost $250 million, puts Tenstorrent in a strong position to bring multiple AI chips to market.

Hyundai’s interest in alternatives to Nvidia stems from its goal of incorporating AI into car technology and its recent acquisition of MIT robotics spinoff Boston Dynamics. Hyundai aims to build next-generation products with AI, realizing that relying on Nvidia for standard products is economically impractical. Heung-soo Kim, Hyundai’s executive vice president, stated, “Tenstorrent’s high growth potential and high-performance AI semiconductors will help the Group secure competitive technologies for future mobilities.”

Samsung’s involvement is equally strategic, as one of the world’s leading semiconductor contract manufacturers. Having produced chips for Tesla’s Autopilot, Samsung recognizes the potential for startups like Tenstorrent to become major customers in chip-making.

RISC-V: The Open-Source Path forward

Keller’s vision aligns with the economic pressures and technological advancements in the AI industry. He believes that AI and RISC-V will merge, with computation predominantly driven by AI. While Nvidia’s GPUs based on CUDA and PyTorch have been successful, Keller sees them as a temporary solution, not the end game for AI. He points to the rise of open-source alternatives, such as TensorFlow and PyTorch, as promising developments in software collaboration.

To match the open-source software effort, Keller embraces RISC-V, an open-source chip instruction set developed at the University of California, Berkeley. RISC-V’s scalability and open nature offer control to Tenstorrent and its customers, unlike dealing with Nvidia’s monopoly. “Slowly RISC-V is gonna replace everything,” claims Keller, alluding to the potential displacement of ARM, Nvidia’s instruction set, and the legacy x86 code that underpins Intel’s empire.

Tenstorrent’s Approach: AI and General-Purpose Computing

Tenstorrent’s strategy involves the development of a dedicated AI chip alongside a RISC-V-based general-purpose CPU. This combination addresses the need for AI and general-purpose computing to coexist seamlessly. Keller emphasizes the importance of a tightly embedded relationship between the two, driving Tenstorrent to build a RISC-V processor.

With a team consisting of talented designers from AMD, Apple, and Nvidia, Keller is confident in Tenstorrent’s potential. The company aims to provide high-end processors that have licensing value, alongside AI accelerator chips that can be sold to other players in need of an alternative to Nvidia.

Seeking Alternatives to Nvidia

While Nvidia’s dominance seems unassailable, industry players are exploring alternative solutions due to the economic burden imposed by high-priced Nvidia products. Startups, power-supply companies, microcontroller companies, autonomous driving startups, and data center edge server manufacturers all seek AI engines for their products without incurring exorbitant costs.

Tenstorrent’s innovative approach, coupled with the open-source advantages of RISC-V, addresses these concerns. By leveraging RISC-V and developing collaborative technologies, Tenstorrent offers a potential escape from Nvidia’s control.

The Adventure Continues

For Keller, computers are an adventure. Throughout his career, he has consistently challenged the status quo, delivering breakthroughs with the world’s fastest chips, contributing to Apple’s transition away from Intel, and revitalizing AMD’s chip development. In Tenstorrent, Keller saw the opportunity to push the boundaries of chip design in response to the increasing demand for performance in deep learning AI models.

Reflecting on the dynamic nature of the chip industry, Keller notes that the war for computation has undergone numerous transformations, with different players emerging victorious at different times. The convergence of AI, RISC-V, and economic pressures creates the perfect storm for change.

As Tenstorrent pioneers the fusion of AI and RISC-V, Keller holds the conviction that this new battle will shape the future of computing. With his passion for innovation and the support of Hyundai and Samsung, Keller and Tenstorrent are determined to provide a compelling alternative to Nvidia, fueling healthy competition in the AI landscape.

Image