CBRS · Cerebras Systems Inc.

Cerebras Systems Inc. is pioneering a fundamentally differentiated approach to AI infrastructure through wafer-scale integration—the WSE-3 processor is 58 times larger than NVIDIA’s B200 chip and delivers 2,625 times more memory bandwidth—enabling up to 15× faster inference on leading open-source models. This architectural advantage underpins its business model, which spans on-premises AI supercomputers, Cerebras Cloud, and partner cloud integrations (AWS Marketplace, Microsoft Marketplace, Hugging Face, etc.), allowing customers to consume AI compute via token, dedicated capacity, or hardware purchase. The company has rapidly scaled revenue from $24.6 million in 2022 to $510.0 million in 2025, fueled by strategic partnerships with OpenAI (a $20+ billion multi-year deal), AWS (binding term sheet for hyperscaler deployment), and UAE-based sovereign AI leaders G42 and MBZUAI. Its financial profile shows a sharp inflection: net income of $237.8 million in 2025 versus a $481.6 million net loss in 2024—driven largely by a $363.3 million gain from extinguishing a forward contract liability—notably masking underlying non-GAAP net losses of $75.7 million in 2025 and $21.8 million in 2024. Gross margin declined to 39.0% in 2025 from 42.3% in 2024, reflecting rising data center costs tied to cloud expansion. The $3.24 billion IPO—priced at $120.00 per share—is among the largest in AI infrastructure history and will fund aggressive scaling of inference capacity, including long-term leases across North America and Europe, while also addressing $230.0 million in tax obligations arising from RSU net settlement. Key risks are highly specific: extreme customer concentration (MBZUAI alone contributed 62% of 2025 revenue and 77.9% of year-end accounts receivable), material weaknesses in internal controls over financial reporting, and total reliance on TSMC for wafer fabrication amid geopolitical volatility in Taiwan. Governance is dominated by a dual-class structure granting Class B shareholders 20 votes per share—resulting in 99.2% of voting power held by pre-IPO investors post-offering—while the executive team includes seasoned semiconductor architects and operators with deep experience at AMD, SeaMicro, and Sunrun. The offering’s significance lies not only in scale but in validation: it positions Cerebras as a foundational AI infrastructure provider at the intersection of sovereign AI investment, hyperscaler co-design, and real-time reasoning workloads—where speed directly translates into user engagement, developer productivity, and new application categories like coding agents and digital twins. With over 100 exaflops of deployed compute and contracts requiring delivery of 750MW to OpenAI and additional capacity to AWS, Cerebras is transitioning from a hardware vendor to a mission-critical, vertically integrated AI compute platform—making this IPO a pivotal moment for both the company and the broader AI infrastructure ecosystem.