Cerebras Systems filed to go public. What does that mean for AMD, Intel and Nvidia?

The artificial intelligence boom has been the biggest topic in financial markets for nearly two years, following the November 2022 launch of ChatGPT. And while Nvidia (NASDAQ:NVDA) has been the biggest beneficiary of the trend — as have its investors — the AI semiconductor space is about to get a bit more crowded.

Artificial intelligence startup Cerebras Systems filed for an initial public offering earlier this week. Terms or size of the offering has not yet been disclosed, but investor interest has been noticeable, given the returns that Nvidia investors have seen over the past two years.

But what does it mean for the AI GPU market?

The AI accelerator market has so far been dominated by Nvidia, which may have a market share approaching 95%, according to some estimates.

Right behind is AMD (NASDAQ:AMD), with its MI300x offerings. In its most recent quarter, AMD generated more than $1B in AI revenue and the company is believed to be approaching a 10% market share, just three quarters after its products launched.

Intel (NASDAQ:INTC), which has long dominated the CPU market for PCs and traditional data centers, is in third place, with an estimated market share of less than 1%.

The entry of Cerebras into the space, which claims to have solved a decades-long problem in the semiconductor industry’s history, i.e. developing chips as big as full silicon wafers, is something of an unknown for now.

“Whether Cerebras can compete with Nvidia and AMD for AI chip spending ultimately comes down to economics, yields, and the technological differences between traditional GPUs and Cerebras’ wafer-scale engine,” Seeking Alpha analyst Jeremy of Kumquat Research said via email.

Through some guesswork, Cerebras’ wafer-scale engine likely offers faster interconnects between cores and lower average power consumption, but has significantly less memory than a traditional GPU, Jeremy said.

And while that may help with a company’s power bill, it could result in it being “significantly slower [than the competition] for larger inference models because it would need to communicate with off-chip memory,” he added.

Nonetheless, its presence and products — including the August introduction of an AI inference solution which Cerebras said is 20 times than Nvidia GPU-based hyperscale clouds — has paid off for Cerebras’ top line.

In its IPO filing, Cerberas said it saw a 220% year-over-year rise in revenue in 2023, as sales totaled $78.7M, up from $24.6M in 2022.

That rapid growth has also come with heavy losses, however. Cerebras lost $66.6M and $77.8M in the six months ended June 30, 2024, and 2023, respectively, according to the filing.

The company raised $250M in a series F financing round in 2021, valuing it at more than $4B.