Cerebras shares slide 16% after quarterly revenue falls short of estimates

Cerebras Systems saw its shares tumble 16% in extended trading on Wednesday after the AI chip designer reported quarterly revenue that fell short of Wall Street expectations. The decline underscores the pressure on richly valued AI companies to deliver on every key metric.

The stock had closed up 11.6% in regular trading and remains more than 41% above its IPO price, buoyed by optimism about demand for its specialized processors that compete with Nvidia's offerings. This was Cerebras' second earnings report as a public company, and the market's reaction suggests investors are still calibrating how to value a firm that must prove it can scale profitably.

Second-quarter sales rose 74.3% year-over-year to $180.11 million, but that missed the $194.23 million analysts had expected, according to LSEG data. Adjusted loss narrowed to $6.91 million from $40.5 million a year earlier.

The company's adjusted gross margin fell to 40.6% from 46.5% in the prior quarter. CFO Bob Komin attributed the roughly 5-percentage-point decline to higher costs for renting computing capacity that Cerebras had previously deployed at other customer sites.

Despite the miss, Cerebras raised its 2026 adjusted revenue forecast to $880 million to $890 million, up from a prior range of $855 million to $865 million. The company also lifted its annual adjusted gross margin outlook to 41%-43%, from 38%-41% earlier, and said it expects to more than triple revenue in 2027.

Cerebras' cloud business, which provides access to its chips remotely, roughly quadrupled to $126 million in the quarter. Hardware sales, however, declined to $54.1 million from $70.3 million a year ago.

CEO Andrew Feldman said the company's design, which places memory directly on the chip, has cushioned it from surging memory prices that have hit competitors. "Nvidia's prices have gone through the roof because of HBM prices," he said, referring to high-bandwidth memory. "This is a battleground, and if they can't deliver or they're having significant component price increases, of course that helps."

Cerebras uses TSMC's 5-nanometer process for its chips, which Feldman said gives it better access to supply than companies relying on more advanced nodes. The company is also working to expand chip volumes to support a $20 billion multiyear agreement to provide AI compute to OpenAI.