Cerebras reported its first earnings as a public company on Tuesday, and the numbers tell two different stories.
Revenue hit $193.4 million in the first quarter - nearly double what the AI chipmaker brought in a year ago. The company cut its quarterly loss to $14 million, or 22 cents a share, compared with a $23.9 million loss in the same period last year.
That is the good news.
Earnings reports hide more than they reveal, and Market Briefs pulls out what actually moves a stock every weekday morning - you also get a 45-minute investing masterclass as a bonus.
The bad news came in the outlook. Management warned that core gross margin would land between 36% and 38% in the April-through-June period, a step down from the 46.5% it recorded in the first three months of the year. Gross margin is the profit a company keeps after covering production costs. A shrinking margin means Cerebras is spending more to make each chip relative to what it charges.
Investors did not love the forecast. Shares fell about 10% in extended trading after the report.
The stock has had a volatile run since Cerebras went public in May. The company priced its IPO at $185 per share. On its first day of trading, the stock opened at $350 and settled at $311.07. Since that debut, the share price has fallen 28%. It sat at $226.72 when markets closed Tuesday, before the after-hours drop.
The offering raised more than $6 billion, making it the largest IPO for an American tech firm since Uber listed in 2019. That kind of demand showed how hungry investors were for a pure-play AI chip company that could challenge Nvidia.
Management guided for full-year core revenue of $855.5 million to $865 million. At the midpoint, that would be about 69% growth from last year. The company also said it expects core revenue of $914 million in the current fiscal year, which would represent 88% growth from the prior year.
Cerebras is trying to carve out a piece of the AI chip market dominated by Nvidia. The company does not compete across the full range of Nvidia's product line. Instead, it focuses on a specific slice of the market where its chip architecture gives it an edge.
The company also operates a cloud service where customers can run AI models on its processors. That business model - selling both chips and compute time - mirrors what Nvidia does with its DGX Cloud offering.
In the first quarter, Cerebras revealed two big partnerships. Its processors will be deployed in Amazon Web Services data centers, giving it a distribution channel that reaches thousands of enterprise customers. And it inked a $20 billion-plus agreement to provide computing capacity to OpenAI, the company behind ChatGPT. That deal alone could reshape Cerebras's revenue trajectory if it scales as planned.
A June 8 note from Mizuho pointed to a technical advantage. Cerebras fits significantly more SRAM memory onto each chip than what Google's latest TPU or Nvidia's Groq 3 LPU can hold. SRAM is a fast, expensive type of memory that sits close to the processor. Having more of it means the chip can handle certain AI workloads without constantly fetching data from slower memory. That translates to a real performance edge in tasks like inference, where a model is running a prediction rather than training.
The question now is whether Cerebras can grow fast enough to offset the margin pressure. Revenue is growing at nearly 70% annually. But if margins keep shrinking as the company scales production, the market may keep marking the stock down. The next few quarters will show whether the margin compression is a temporary cost of ramping up or a structural feature of the business.
Want the story behind the numbers on names like this? Join the 350,000+ investors reading Market Briefs in five minutes a day, plus a free 45-minute investing course.