Cerebras shares plunge nearly 20% after missing earnings expectations — hardware sales drop but AI cloud revenue climbs 281%

Cerebras shares plummeted about 18% after the company missed earnings expectations, according to Reuters. The second quarter financial results, released on Wednesday, showed a 23% year-over-year decrease in hardware sales, while its cloud services revenue surged by 281%. Despite a nearly doubled year-over-year earnings figure, Cerebras' shares fell in after-hours trading due to the earnings miss and the drop in hardware sales, according to Reuters.

The company's cloud services revenue reached $125.99 million, significantly higher than the $33.03 million in the same quarter last year. However, hardware sales dropped to $54.12 million from $70.3 million in Q2 2025. Wall Street analysts had expected Cerebras to earn $194.23 million during the quarter.

Cerebras' operating expenses rose to $502.79 million, a substantial increase from $89.28 million a year ago, and its gross margin fell to 14%. The company incurred a loss of approximately $450.53 million. The primary cause of the increase in operating expenses and losses is stock-based compensation, which rose from $13.3 million in the previous quarter to $377 million in the second quarter of 2025. This compensation, triggered by Cerebras' May IPO, is expected to lead to further losses.

Despite the negative earnings and hardware sales figures, analysts were more disappointed by the earnings miss, the declining hardware sales, and the uncertain returns generated by Cerebras' new business model. The company's transition from a hardware sales model to a cloud services model has led to increased capital expenditure on infrastructure before it can earn cloud revenue.

This shift requires Cerebras to invest heavily in infrastructure, such as Wafer Scale Engines manufactured by TSMC, and deploy hardware in customers' data centers. The company's new model allows customers to pay for AI inference/training compute access, rather than owning the machines themselves.

Cerebras' new model has proven successful, with demand for inference AI compute increasing, as demonstrated by the $20 billion OpenAI agreement. Cerebras has committed to providing 750 MW of inference capacity over several years, with an option for an additional 1.25 GW. However, the company must invest in infrastructure before earning revenue from these agreements, raising concerns about its ability to generate attractive and sustainable profits.