Nvidia's Biggest Rival: The Surprising Answer

In the high-stakes world of AI hardware, Nvidia has long been the undisputed king. Its GPUs power some of the most prominent AI products, including OpenAI’s ChatGPT, and the company controls more than four-fifths of the market for add-on GPUs. But who is Nvidia's biggest rival? The answer might surprise you.

Nvidia's dominance in the AI hardware segment is hard to overstate. The company's GPUs are in high demand for AI applications, and its market share has grown significantly in recent years. According to Jon Peddie Research, the market for add-on GPUs grew 47.9% between the second quarter of 2023 and the second quarter of 2024, highlighting the increasing demand for AI applications that Nvidia's GPUs are well-positioned to meet.

But Nvidia isn't the only player in the game. Advanced Micro Devices (AMD) is one of Nvidia's main rivals, with a budding GPU and data center business. AMD has been a direct beneficiary of AI tailwinds over the last couple of years. For instance, AMD's third-quarter earnings report released on Oct. 29 showed an 18% year-over-year increase in total revenue to $6.8 billion, with adjusted gross profit and operating profit increasing by 23% and 34% year over year, respectively. AMD has also been an aggressive acquirer of smaller AI platforms, which has helped it accelerate sales and expand profit margins while integrating new products and services to bolster its AI portfolio.

But AMD isn't the only company challenging Nvidia's dominance. Intel, which is newer to the market for GPUs and other AI chips compared to Nvidia and AMD, launched both its first GPU and its first specialized AI chip in 2022. Although Intel hasn't made much of a dent in the market shares of Nvidia and AMD yet, it is a household name, with its processors being widely used in Windows computers. Intel is still catching up in the AI chip market, but its entry into the space is a sign that the competition is heating up.

Another company that is challenging Nvidia's position is Cerebras Systems. Cerebras specializes in a completely different type of chip architecture called a wafer-scale engine (WSE). According to Cerebras' filings, a larger wafer-style chip has the potential for considerably more compute power, memory, and bandwidth compared to incumbent GPU infrastructure. Cerebras' WSE-3 processor carries 44GB of on-chip memory, which is 880 times that of an Nvidia H100. The memory bandwidth of 21 Petabytes a second beats an H100 seven thousand times over, making it a potential game-changer in AI training and inference.

Cerebras has a list of prominent clients, including Meta, Docker, Aleph Alpha, and Nasdaq, which utilize its technology for AI Model Services and other offerings. This client base provides concrete belief in the company’s future and its potential to challenge Nvidia's dominance. Cerebras' technology is designed to be a far closer match to what AI models require for nourishment than a cluster of GPUs tethered together through Ethernet or Nvidia’s own InfiniBand. This design can significantly reduce the time required for AI training and inference, making it a more efficient solution for AI workloads.



In summary, while Nvidia maintains a dominant position in the AI hardware segment, competitors like AMD, Intel, and Cerebras Systems are employing strategies such as aggressive acquisitions, specialized chip architectures, and in-house development to challenge Nvidia's position. The battle for AI dominance is far from over, and the next few years will be crucial in determining who emerges as the ultimate winner.