Stay up thus far with free updates
Simply register for Artificial intelligence myFT Digest – delivered straight to your inbox.
Nvidia's competitors are mobilizing their forces to interrupt the corporate's dominance within the AI chip market, raising a whole bunch of hundreds of thousands of dollars and launching recent products to capitalize on the boom in artificial intelligence technology.
Cerebras, d-Matrix and Groq are amongst a gaggle of smaller corporations that need to take a chunk of the multi-billion dollar marketplace for AI chips from Nvidia. The company has to this point dominated the primary wave of investment with its graphics processing units (GPUs).
They are riding the wave of expectations that demand for artificial intelligence “inference” – the computing power required for models like OpenAI’s ChatGPT and Google’s Gemini to generate responses to queries – will grow exponentially as chatbots and other applications of generative AI develop into more popular.
Nvidia's Hopper GPUs are well-suited to the extremely resource-intensive task of coaching world-class AI models and have develop into one of the sought-after commodities on the planet.
Cerebras, d-Matrix and Groq as a substitute give attention to cheaper, more specialized chips designed to run AI models.
On Tuesday, Cerebras announced its recent Cerebras Inference platform, based on its dinner-plate-sized CS-3 chip. Cerebras claims its solution is 20 times faster at AI inference than Nvidia's current generation of Hopper chips – and at a fraction of the value, citing tests conducted by benchmarking analytics provider Artificial Analysis.
“You can beat the 800-pound gorilla by bringing a a lot better product to market,” Cerebras CEO Andrew Feldman told the Financial Times. “In my experience, the higher products normally win, and we've taken necessary customers away from (Nvidia).”
The CS-3 chip forgoes using a separate high-bandwidth memory chip like Nvidia uses, as a substitute offering another architecture with memory integrated directly into the chip wafer.
Limitations on memory bandwidth are a fundamental limitation on an AI chip's inference speed, Feldman said. Combining logic and memory in a single large chip produces results which might be “orders of magnitude faster,” he said.
d-Matrix, founded in 2019 by Sid Sheth, can be launching a brand new funding round, lower than a yr after it raised $110 million in a Series B round led by Singapore's state fund Temasek. Sheth said the corporate is aiming to boost $200 million or more later this yr or early next yr. d-Matrix is still within the early stages of the funding process and said the ultimate amount is subject to vary.
d-Matrix plans to launch its own chip platform, Corsair, later this yr. Sheth said the corporate is combining its products with open software reminiscent of Triton, which competes with Nvidia's Cuda, a widely used software platform that provides developers the tools to construct AI applications and optimizes the performance of the chips.
Nvidia's biggest customers depend on using open source software like Triton. “App developers don't like being locked into a selected tool,” says Sheth, and “individuals are starting to understand that Nvidia has a stranglehold on the training side with Cuda.”
Groq, one other contender in the factitious inference space led by a former founding member of Google's Tensor Processing Unit team, raised $640 million this month from investors led by BlackRock Private Equity Partners and is valued at $2.8 billion.
A enterprise capitalist warned that despite the hype surrounding the sector, it has been a significant challenge for semiconductor start-ups to penetrate the market.
Chipmaker Graphcore was bought by SoftBank last month for just over $600 million, lower than the roughly $700 million the corporate had raised in enterprise capital since its founding in 2016, people acquainted with the deal said.
Groq and Cerebras were also founded in 2016. “There has been an almost insatiable desire amongst public investors to search out and back the following Nvidia,” said Peter Hébert, co-founder and managing partner of enterprise capital firm Lux Capital. “It's not nearly chasing the newest trend. The momentum can be benefiting several VC-backed chip startups which were hard at work for nearly a decade.”
