This year’s IPO market has been a stalled car.
You know the feeling. You turn the key, the engine hums, you’re about to hit the road…and then the car stops. That’s sort of what’s happened in 2024, as the scattered IPOs of companies like Reddit and Astera Labs inspired hope that IPOs were back to the races. But in the end, while the IPO engine has hummed, the car hasn’t started.
At least, not yet. This week, AI chipmaker Cerebras filed its S-1, marking the beginning of its journey to the public markets. There’s hope that a runaway-success debut for Cerebras could finally get things motoring again. On the surface, that makes sense—entrenched in the AI boom, Cerebras is backed by big-name VCs like Benchmark and Altimeter, billing itself as a prospective challenger to Nvidia.
The company’s top line provides some strong support for that story. Cerebras’s 2023 revenue came in at $78.7 million, more than triple the prior year’s $24.6 million, with momentum on the rise. Revenue in the first half of this year surged to $136.4 million, up more than 1,400% from the first half of 2023.
What’s more, as deep tech investor Patrick Beatty pointed out to me, the company’s got a “proven team,” led by Andrew Feldman, who cofounded and sold SeaMicro, a server startup, to AMD for $357 million in 2012.
Look a little closer at Cerebras’s S-1, however, and it’s difficult to ignore some big questions. Cerebras has narrowed its losses, but they’re still substantial, totaling $127.2 million in 2023 versus $177.7 million in 2022. But there’s something else that every person I’ve spoken to brought up almost immediately when discussing the filing—Cerebras is heavily reliant on one customer.
In the first half of this year, a staggering 87% of Cerebras’s revenue was accounted for by G42, an Abu Dhabi-based company backed in part by Mubadala, Silver Lake, Microsoft, and Ray Dalio’s family office. (Microsoft reportedly has sought to pull back on its partnership with G42.) G42, an AI-focused data-center and cloud-services business, also led Cerebras’s $250 million Series F in 2021. (That round valued the company at $4.25 billion.)
I’ve seen some commentary out there that says something to the effect of “well, that’s a problem Nvidia has too.” And, again on the face of it, you’d be right. About 46% of Nvidia’s revenue links back to four unnamed customers, a group that reportedly includes Amazon, Meta, Microsoft, and Alphabet. So, yes, customer concentration is perhaps a feature that Cerebras shares passingly with the world’s most famous chipmaker—but there’s a difference. The stability of your customers matters deeply when so few mean so much. G42 is private, backed by the United Arab Emirates’ sovereign wealth fund among others, and was only founded in 2018. (Mired in geopolitical risk, G42 divested from China earlier this year amid U.S. pressure.) Nvidia’s mega-customers, by contrast, are established Big Tech businesses that generate hundreds of billions of dollars in annual revenue in aggregate.
Additionally, Cerebras’s ‘Nvidia Challenger’ moniker is going to be hard to live up to—the company would have to substantially take on Nvidia in AI training, while Cerebras’s inference business (an area less well-trod by Nvidia) is still nascent.
It’s still possible that public markets investors ultimately bite, but it would likely be more about AI generally than about Cerebras specifically.
"If you’re talking about chip investors who’ve done due diligence, I’d be surprised,” said Thomvest Ventures managing director Umesh Padval. “It may end up being about the FOMO of missing out on chip companies, in the market of AMD, Broadcom, and Nvidia. So, if they feel ‘I’m getting the next Nvidia,’ that could be the primary reason why there would be so much demand.”
In the end though, the name of the game is growth—and if Cerebras can convince investors that its own business and the broader market fueling it are real and expanding, the IPO engine may very well start.
“You have to continue to provide growth to keep investors excited,” said Simon Yu, Listed Ventures managing partner. “So, you don't go public to just be stagnant, because then you'll just die on the vine.”
See you tomorrow,
Allie Garfinkle
Twitter: @agarfinks
Email: alexandra.garfinkle@fortune.com
Submit a deal for the Term Sheet newsletter here.
Nina Ajemian curated the deals section of today’s newsletter.
