AI chip newcomer Cerebras (CBRS.US) falls 10% after first earnings report: Annual guidance only up 69%, though exceeding expectations, still difficult to satisfy market appetite.

AI chip manufacturer Cerebras Systems (CBRS.US) saw its stock price drop by about 10% in after-hours trading, as the annual sales forecast given by this recently listed chip manufacturer disappointed investors the market had expected the company to capture a larger share of the artificial intelligence data center market. In a statement on Tuesday, Cerebras said that its revenue in 2026 will reach $8.55 billion to $8.65 billion, a year-over-year increase of 69% when calculated at the midpoint. According to compiled data, analysts' average forecast was $8.248 billion. Its first post-IPO financial report showed that first-quarter sales soared 94% to $1.934 billion; during this period ending March 31, the net loss was $14 million, or $0.22 per share, exceeding expectations of $0.07. Analysts had estimated revenue of $1.814 billion and expected a loss of $58.6 million. Cerebras' hardware business brought in sales of $1.106 billion. Sales from cloud and other services totaled $82.8 million. Like other competitors, Cerebras must meet investors' high expectations. Due to the boom in data center construction driven by artificial intelligence workloads, investors have become accustomed to strong profitability and rapid revenue growth driven by surging demand. NVIDIA Corporation and a handful of chip manufacturers often exceed Wall Street's expectations by a large margin, making it no longer a guarantee for stock price increases even with solid quarterly earnings reports. This was the first financial report released by the company after raising $5.5 billion in the largest IPO in the history of the chip industry. Using innovative technology built around a giant chip, Cerebras has carved out a unique space for itself in the field of artificial intelligence infrastructure. The company claims that this technology can better run large AI models and generate quick responses for users. The fluctuation in the company's stock price reflects the high hopes that the market has for chip manufacturers in the AI era. After debuting in May at an opening price of $185, Cerebras' stock price had previously reached a high of $311.07. Following the earnings report on Tuesday, the stock briefly dropped to a low of $202.25 in after-hours trading. Since going public, the stock price has risen by 23%, closing at $226.72 in regular trading on the New York Stock Exchange. As Cerebras released its financial report, chip manufacturer stock prices experienced a sharp decline on a regular trading day, with one Wall Street analyst calling it a "chip disaster." The Philadelphia Semiconductor Index fell by 7.9%, and all 30 constituent stock companies saw their prices drop. CEO Andrew Feldman stated that the biggest challenge currently is obtaining enough data center space. "It is incredibly ironic that after we and NVIDIA Corporation invented all these technologies, that buildings have become the limiting factor," he said in an interview before the earnings report was released. CFO Bob Komminen said on a conference call after the release of the financial report that the shortage of data center space is causing Cerebras to rent back some of its systems from customers and "actively" expand its own capacity. He stated that these costs will lead to a decrease in profit margins of approximately 10 to 15 percentage points this year. The underwhelming performance overshadowed the previous agreement with OpenAI. Under this multi-year agreement, OpenAI will deploy Cerebras' 750-megawatt high-speed inference systems in the coming years. Cerebras stated that this agreement is worth over $20 billion. Additionally, the two companies have jointly launched Codex-Spark, a model designed for near-real-time coding and optimized for interactive work. Cerebras also highlighted its longstanding partnership with Amazon.com, Inc. AWS. The company stated that it will "introduce a decoupled inference strategy, where AWS's Trainium 3 chip handles pre-filling, while Cerebras CS-3 is responsible for running high-speed inference for decoding."