Cerebras files for IPO with 20x revenue growth but heavy customer concentration

AI hardware company Cerebras has filed for an initial public offering amid rapid revenue growth but persistent profitability challenges and heavy reliance on a small number of customers. The company's revenue surged from $24.6 million in 2022 to $510 million in 2025, representing 20-fold growth. However, 86% of that revenue comes from just two customers: Abu Dhabi-based G42 and Mohamed bin Zayed University of Artificial Intelligence. Cerebras maintains a $24.6 billion backlog of orders, including a $20 billion deal with OpenAI. The company expects to recognize approximately 15% of this backlog revenue within the first 24 months, 43% during months 25 to 48, and the remainder thereafter. Despite the revenue growth, Cerebras reported operating losses of $145.9 million in 2025. The company recorded GAAP net income of $237.8 million for the year, but this resulted from a $363 million accounting adjustment related to a forward contract liability reduction rather than operational profitability. Gross margins have improved from 12% in 2022 to 39% in 2025 as the company scaled its wafer-scale engine business. Cerebras has never been profitable since its founding. The company postponed IPO plans in 2024 after a national security review examined its ties with G42, which is both a customer and investor holding a 1% stake acquired for $40 million in 2021. Cerebras designs complete wafer-scale AI systems, packing approximately 900,000 compute cores, 44 GB of on-chip SRAM, and 21 PB/s of on-chip bandwidth into single silicon wafers. The architecture avoids inter-chip communication bottlenecks but faces manufacturing yield challenges. Recent agreements to supply AI hardware to Amazon Web Services and OpenAI are expected to diversify the company's revenue streams beyond its current concentrated customer base. Market expectations suggest Cerebras could raise approximately $3 billion in its IPO, significantly higher than earlier $1 billion plans, reflecting both rapid growth and substantial AI infrastructure ambitions.