Cerebras, an AI chip startup, reported 88% revenue surge ahead of earnings, though the company remains unprofitable; OpenAI orders drive growth.
As AI chip upstart Cerebras Systems prepares to release its second-quarter earnings, market attention is sharply focused on the tension between its surging revenue and suddenly pressured profit margins. Whether the company can demonstrate a credible path to scalable profitability—following landmark client wins such as OpenAI—will be pivotal for its share price.
After U.S. market hours on Wednesday, August 12 (Eastern Time), Cerebras Systems—which went public in May—will deliver its second earnings report as a publicly listed company. According to prior guidance, core revenue for the second fiscal quarter is expected to reach approximately $194 million, representing an 88% year-over-year surge.
However, this robust year-over-year growth translates to only marginal sequential growth compared to the $191.3 million in core revenue reported in the first fiscal quarter, indicating a near-term deceleration in momentum. Core revenue is a non-GAAP metric that excludes items such as amortization of customer warrants and pass-through data center revenue. Wall Street's current consensus estimate aligns closely with company guidance, forecasting revenue of roughly $191 million.
In the first quarter, Cerebras reported total revenue of $193.4 million, up 94% year-over-year. Cloud and other services revenue soared by 178% to $82.8 million, while hardware revenue rose 59% to $110.6 million.
Investors will scrutinize Cerebras' earnings quality more critically this time around. Wall Street currently expects the company to post a net loss of $0.17 per share, a stark reversal from the $1.44 earnings per share reported a year earlier. Furthermore, in its previous earnings release, the company warned that core gross margin for the second fiscal quarter would plummet to 36%–38%, significantly below the 46.5% recorded in the first fiscal quarter. Core operating margin is also projected to fall into negative territory, ranging from –30% to –32%.
The sharp decline in gross margin stems primarily from the company's need to temporarily lease and rely on customers' existing systems—such as expensive compute capacity from G42—to fulfill surging demand for AI inference ahead of full-scale deployment of its own data centers. This reliance has severely eroded profitability.
Wedbush analyst Matt Bryson explicitly noted that data center capacity—not wafer supply—is currently Cerebras' most significant operational constraint. To date, Cerebras has secured approximately 215 megawatts of compute capacity for 2026. Analysts believe the faster the company brings its proprietary data centers online, the sooner it can shed its dependence on high-cost external compute resources. Any sign of margin improvement or an inflection point could serve as a major catalyst for the stock price.
Despite near-term profit pressure, Cerebras continues to rapidly expand its business footprint. The company previously announced a multi-year compute contract with OpenAI valued at over $20 billion and spanning 750 megawatts of capacity. Additionally, Cerebras has established a multi-year collaboration with Amazon AWS to bring its high-speed AI inference technology to the cloud. Furthermore, the company recently entered into partnerships with CrowdStrike and AMD, under which it will integrate its Helios architecture with AMD's offerings and plans to deliver the solution via Cerebras Cloud later this year.
Wedbush believes that the partnerships with CrowdStrike and AMD were likely not incorporated into management's initial earnings guidance, implying that these projects could provide additional upside to results in the coming quarters.
On the hardware iteration front, Cerebras will host its annual Supernova conference on August 18. Wedbush expects this event to be a likely venue for the launch of the company's next-generation WSE-4 processor, though management could also preview related product updates during the upcoming earnings call. The WSE-4 is expected to deliver a significant performance leap over the WSE-3, offering Cerebras an opportunity to raise prices and improve margins—a positive catalyst that is not yet fully reflected in current expectations.
Amid fierce competition from giants such as NVIDIA, Advanced Micro Devices, and Broadcom, Cerebras' stock price has declined approximately 30% since its IPO. However, ahead of its earnings release, Wedbush maintained its "outperform" rating and $280 price target.
Analyst Bryson emphasized that Cerebras' long-term investment thesis remains intact: its Wafer Scale Engine (WSE) leverages SRAM instead of the industry-wide constrained and deteriorating HBM memory, bypasses CoWoS advanced packaging limitations, and continues to utilize the 5nm node rather than the even tighter 3nm process. As the industry moves into the second half of 2026 and approaches 2027, worsening memory supply constraints and pricing conditions across the sector will increasingly highlight the value of Cerebras' distinctly different architectural approach.
The market is now awaiting clearer signals from management in this earnings report: when new capacity will come online at scale, when the OpenAI and AWS contracts will begin contributing meaningful revenue, and whether full-year guidance can be maintained or raised. These answers will determine whether this AI chip challenger can regain market confidence.