Cerebras has stepped out in front in the coming era of AI IPOs.
The California-based company, which designs and manufactures supersized AI chips that compete with Nvidia’s GPUs, filed for a Nasdaq IPO on Sept. 30. The listing is a bellwether of the public markets’ appetite for loss-generating AI companies.
Cerebras, which also designs AI software, is not yet profitable: Its net loss was $66.6 million for the first six months of 2024, a 14% improvement from the previous year, according to the filing. But it is betting that demand for chips, accelerated by the AI boom, will close that gap.
Cerebras’ growth has been extremely rapid. Its revenue soared from $8.7 million in the first half of 2023 to $136.4 million in H1 2024, a year-over-year increase of 1,467.8%. But there’s a big asterisk: 87% of its revenue in the first half of 2024 came from a single customer, Abu Dhabi-based AI company G42.
G42 led Cerebras’ $250 million Series F at a $4.25 billion valuation in late 2021. Foundation Capital, Benchmark and Eclipse Ventures are the company’s largest outside shareholders.
Cerebras thinks demand will support its growth, projecting in the filing that the market for hardware, software and services to train and run AI systems will balloon from $131 billion today to $453 billion by 2027.
Nevertheless, it may not be smooth sailing on the stock market. The stock price of Astera Labs, the only other pure-play AI company to go public this year, remains about where it opened on its first day of trading.
Cerebras has developed a solution for AI training and inference—the process of running live data through trained models— powered by a processor that has more computational power than the leading commercially available cluster of GPUs. Its “AI Supercomputer” can train a large language model equivalent in size to Chat GPT-3 with 97% fewer lines of code compared to GPU clusters like those produced by Nvidia, according to its IPO filing.
Like other AI infrastructure companies, Cerebras has spent heavily on R&D: over $77 million in the first six months of 2024. Its cumulative losses hit $728.2 million in June.
As AI chipmakers in the private markets dig deeper into VCs’ pockets, Cerebras can serve as a test case of if that capital expenditure can pay off.
Featured image courtesy of Cerebras
