Cerebras Files for IPO Despite Heavy Reliance on Two Customers and Persistent Losses

Cerebras, the pioneering supplier of wafer-scale AI processors, has filed for an initial public offering (IPO) after postponing such plans last year due to national security concerns over its ties with G42, an Abu Dhabi-based AI company backed by sovereign wealth fund Mubadala. The filing reveals the company has achieved remarkable revenue growth but faces significant challenges around customer concentration and profitability.

Rapid Revenue Growth Masks Underlying Fragility

The company's financial results paint a picture of explosive growth tempered by concerning dependencies. Cerebras' revenue has grown from $24.6 million in 2022 to $510 million in 2025—a 20x increase that positions it among the fastest-growing AI hardware companies. However, this impressive growth figure masks a critical vulnerability: approximately 86% of revenue comes from just two customers—G42 and Mohamed bin Zayed University of Artificial Intelligence (MBZUAI).

This heavy concentration means Cerebras depends on a handful of large, project-based deployments rather than the diversified, repeatable revenue base that companies like AMD or Nvidia typically maintain. The remaining 14% of revenue comes from a fragmented base of smaller enterprise, government, and cloud customers, but none contribute enough individually to reduce the company's reliance on its top two clients.

Recent Wins May Diversify Revenue Streams

Cerebras has recently inked agreements to supply its AI hardware to Amazon Web Services and OpenAI, which could help diversify revenue streams and reduce dependence on G42 and MBZUAI. These partnerships represent significant validation of Cerebras' wafer-scale architecture approach.

The company currently has a massive $24.6 billion backlog, including the $20 billion OpenAI deal, providing strong demand visibility. Cerebras expects to recognize approximately 15% of this backlog within the first 24 months through December 31, 2027, 43% during months 25 to 48, and the remainder thereafter.

However, the company warns that converting this backlog into revenue depends on manufacturing capacity of its partners, infrastructure deployment, and power availability. Cerebras further notes that all manufacturing services and components are purchased on a purchase order basis without capacity or volume commitments, posing significant risks given the company's long supply chain.

Profitability Remains Elusive

Despite the revenue growth, Cerebras has never been profitable. The company reported GAAP net income of $237.8 million in 2025, compared to a net loss of $481.6 million in 2024. However, this income statement is misleading as the profit did not come from core business operations but from an accounting adjustment.

Cerebras recorded a $363 million gain from a change in the fair value (and extinguishment) of a forward contract liability. The company had a financial obligation whose value was reduced, allowing it to book that reduction as income. Without this accounting benefit, Cerebras would have remained unprofitable, with operating losses totaling $145.9 million in 2025.

Gross margins have improved significantly from 12% in 2022 to 39% in 2025 as the scale of its business and product mix improved, but the company continues to operate at a loss.

Wafer-Scale Architecture: Innovation with Challenges

Cerebras positions itself as an AI infrastructure company, not just a maker of AI accelerators. The company designs a full stack: wafer-scale engine (WSE, literally a full silicon wafer turned into one processor), systems, and software, delivered as rack-scale solutions.

While Nvidia sells everything from AI GPUs to fully built rack-scale solutions, Cerebras only sells systems. The WSE packs around 900,000 compute cores, 44 GB of on-chip SRAM, and 21 PB/s of on-chip bandwidth, avoiding inter-chip communication bottlenecks by design. Essentially, Cerebras trades system complexity for silicon complexity.

However, wafer-scale chips are notoriously hard to yield. The WSE features plenty of redundant cores and memory cells to maximize yields, but this remains a significant technical challenge.

National Security Concerns and IPO Timing

Cerebras postponed its IPO plans in 2024 after a national security review examined its ties with Abu Dhabi-based G42 amid concerns about potential foreign access to advanced AI processors. G42 is both a customer and investor, controlling a 1% stake in Cerebras that it acquired for $40 million in 2021.

The company has not specified an official fundraising target in its IPO filing, but current market expectations point to a roughly $3 billion raise. This is significantly higher than earlier $1 billion plans, reflecting the company's rapid revenue growth and the scale of its AI infrastructure ambitions.

The IPO filing comes at a time when investors are showing renewed interest in AI infrastructure companies, despite concerns about profitability and customer concentration. Cerebras' wafer-scale approach represents a fundamentally different architecture from traditional GPU-based solutions, potentially offering advantages for certain AI workloads.

As the company moves forward with its public offering, it will need to demonstrate that its innovative technology can translate into sustainable, diversified revenue and eventually profitability—challenges that have proven difficult for many high-growth AI hardware companies.