CBRS · Cerebras Systems Inc.

Cerebras Systems Inc. is pioneering wafer-scale AI infrastructure, positioning itself at the forefront of the most demanding segment of the AI compute market—low-latency, high-bandwidth inference. Its core innovation, the Wafer-Scale Engine (WSE-3), integrates 900,000 cores and 44GB of on-chip SRAM on a single 46,225 mm² silicon die—58x larger than NVIDIA’s B200 chip—enabling 2,625x greater memory bandwidth. This architectural advantage underpins its claim of up to 15x faster inference speeds, validated by third-party benchmarks and deployed with marquee clients like OpenAI and AWS. The company’s business model spans on-premises hardware sales, dedicated cloud capacity, and consumption-based inference services, supported by co-designed software (CSoft compiler, Inference Serving Stack) and AI model services. Revenue grew dramatically—from $24.6M in 2022 to $510.0M in 2025—driven heavily by sovereign AI initiatives in the UAE (MBZUAI and G42), which together accounted for 86% of 2025 revenue. While this customer concentration fuels near-term growth, it introduces significant execution and geopolitical risk, especially given the Middle East’s exposure to regional conflict and evolving U.S. export controls. Financially, Cerebras achieved net income of $237.8M in 2025 after a $481.6M loss in 2024, though non-GAAP net loss remained $75.7M due to $150.5M in stock-based compensation tied to the IPO liquidity event. Its cash position stood at $701.7M as of December 31, 2025, bolstered by a $1.0B Working Capital Loan from OpenAI and $1.0B in Series H preferred financing in January 2026. The IPO offers 30M Class A shares at a $155.00 midpoint price, targeting $4.4875B in net proceeds—$329.6M of which is earmarked for RSU-related tax withholding. Proceeds will fund data center expansion, R&D, and general corporate needs, while the company simultaneously ramps a $250M revolving credit facility. Governance is highly centralized: post-IPO, Class B holders (including founders and major investors) will own ~85.3% of shares but control ~99.2% of voting power, with a classified board and multiple anti-takeover provisions. Key risks include reliance on TSMC without volume guarantees, unremediated material weaknesses in financial controls, and regulatory uncertainty around AI, export controls, and data privacy—particularly given its global footprint and deployment in sensitive jurisdictions. The company’s success hinges on scaling its differentiated architecture amid intensifying competition from GPU incumbents and neo-clouds, while navigating complex supply chains, sovereign demand volatility, and the operational challenges of transitioning from hardware-first to hybrid cloud/hardware monetization.