Cerebras files for the first generative artificial intelligence IPO

If Cerebras can overtake its original benefactor, G42, this IPO will be a rocket. If not,…

It may be hard to believe, but there has not been a single public offering of an AI hardware company. Period. Not one. Why? Because neither startup has yet achieved the escape velocity needed to become standalone again as Nvidia. Perhaps this will change soon. Cerebras filed with the SEC for an IPO; their IPO could take place this year. Let’s dive in.

Okay, who is Cerebras, again?

If you follow me, you know that I have been a long-time fan of Cerebras Systems, their CEO Andrew Feldman, whom I have known for 12 years, and the Wafer Scale Engine (WSE) approach to AI computing. The idea is so simple and so compelling. Instead of cutting a board into chips, packaging them in high-bandwidth memory, installing them on a computer motherboard, and then reconnecting hundreds of chips to communicate with each other over an expensive network, why not just connect the resident chips together? on a waffle? This is wafer-scale calculations that I think will become a huge trend in the next 5 years. Oh, and they don’t need an HBM because Cerebras uses on-chip SRAM supplemented with a streaming server. Yes, powering and cooling an entire wafer is difficult and you have to deal with inevitable defects on the wafer, but the resulting math is too compelling not to succeed. That is, if you have the right software.

Why an IPO and why now?

The moment may be good. Cerebras’ revenue increased more than 14-fold in the first half of 2024 to $136.4 million, and the company’s net loss decreased from $77.8 million to $66.6 million. This provides the perfect backdrop to try to raise more funds to accelerate growth. But rather than return to (expensive) venture capital, co-founder and CEO Andrew Feldman believes public markets are ready for the offering of generative AI, and he’s doing it.

The problem is customer concentration

While Nvidia has a similar problem, with four customers accounting for more than 40% of revenue last quarter, Cerebras’ customer concentration may depend on where they are in their lifecycle. Getting a prospect to the table takes a lot of selling and convincing, especially if you’re just another startup with a deck of slides. This IPO could allow Cerebras to detect more G42 aircraft flying under the hyperscaler’s radars.

Cerebras’ prospectus shows that G42 accounted for 87 percent of its revenue in the first half of the year. G42 has committed to purchasing $335 million worth of Cerebras shares by April next year, giving it a stake of more than 5 percent. G42 and its subsidiaries are already purchasing and building a 36 exaflop distributed supercomputer powered by the Cerebras processor.

As an analyst, investors often ask me how Cerebras can grow, given that their largest potential customer is already building its own AI chips and, of course, buying hundreds of thousands of Nvidia GPUs. I ask them if they had even heard of G42 before they started buying Cerebras WSE exchanges, and I assure them that there are a dozen more G42s being funded to build local AI factories as sovereign AI farms and second-tier public clouds. Perhaps they just need permission to purchase.

Conclusions

Although we will have to wait for the market to accept Cerebras’ planned public offering, we can bet that it will be strong. Stock investors are eager to invest in an alternative to Nvidia, and many believe AMD is fully valued. The AI ​​market is also eager to have an alternative it can implement to implement AI, and Cerebras’ value proposition is extremely compelling in terms of both training and inference.

If they can pull off a successful IPO, it could help them grow from a small base; many enterprises will not buy from startups. And nothing promotes success like success.

Disclosures: This article expresses the opinions of the author and

should not be relied upon as advice to purchase or invest in any companies mentioned. Cambrian-AI Research is fortunate to have many, if not most, semiconductor companies as our clients, including Blaize, BrainChip, Cadence Design, Cerebras, D-Matrix, Eliyan, Esperanto, GML, Groq, IBM, Intel, Micron , NVIDIA, Qualcomm Technologies, Si-Five, SiMa.ai, Synopsys, Ventana Microsystems, Tenstorrent and dozens of investment clients. We do not have investment positions in any of the companies mentioned in the article and do not plan to initiate any in the near future. More information can be found on our website https://cambrian-AI.com.