Cerebras Systems has officially filed for an initial public offering (IPO), positioning itself as a significant challenger to Nvidia in the high-performance AI hardware market. The announcement comes after a previous IPO attempt in 2024 was delayed due to a federal review of an investment from G42, an Abu Dhabi-based technology company, which was ultimately withdrawn. Following this setback, Cerebras has focused on enhancing its financial stability and expanding its customer base.
CEO Andrew Feldman has consistently touted the company’s technology as “the fastest AI hardware for training and inference,” and this IPO filing represents a crucial step in proving that claim in the public arena. In the past year, Cerebras raised over $2 billion, closing a $1.1 billion Series G funding round last year and following it with a $1 billion Series H round in February, which valued the company at approximately $23 billion.
The momentum behind the IPO is bolstered by two significant partnerships. Cerebras has secured a deal with Amazon Web Services (AWS) to integrate its chips into Amazon’s data centers, establishing a foothold with one of the largest cloud service providers globally. More notably, the company has reportedly struck a contract with OpenAI valued at over $10 billion, further solidifying its competitive position. In a recent interview, Feldman remarked, “Obviously, [Nvidia] didn’t want to lose the fast inference business at OpenAI, and we took that from them.”
The financials presented in the IPO filing indicate solid growth. Cerebras generated $510 million in revenue for 2025, registering a net income of $237.8 million on a GAAP basis. However, it reported a non-GAAP net loss of $75.7 million after excluding certain one-time items, reflecting the typical financial trajectory of a high-growth hardware enterprise that is investing heavily in research and scaling up production.
Central to Cerebras’ competitive edge is its Wafer-Scale Engine, a silicon wafer the size of a dinner plate that incorporates hundreds of thousands of AI cores. This design contrasts with traditional systems that connect numerous smaller GPUs, which often introduce latency, power inefficiencies, and software complexities. By consolidating workloads onto a single massive chip, Cerebras aims to deliver superior speed and memory bandwidth essential for the most demanding AI models, a space where even Nvidia’s formidable clusters sometimes falter.
The timing of the IPO is notable given the surging demand for AI computing capabilities. Major industry players are actively seeking alternatives that offer better performance per dollar or watt. OpenAI’s decision to award a substantial contract to Cerebras instead of relying solely on Nvidia illustrates a market readiness to explore new architectural solutions.
The partnership with AWS further validates Cerebras’ transition from experimental technology to real-world applications. However, the company faces significant challenges ahead. Nvidia’s well-established ecosystem—including its widely-used CUDA software platform, extensive developer community, and years of optimization—presents a formidable barrier. For Cerebras to succeed, it must continue to demonstrate that its wafer-scale chips are not only faster but also easier to program and reliable at scale. Moreover, manufacturing such large chips in volume poses inherent technical and supply-chain risks, even with robust foundry partnerships.
While the company has yet to disclose its fundraising target or the timeline for the IPO—currently aimed for mid-May—the filing itself is a pivotal moment. After navigating various regulatory hurdles, securing significant funding, and landing high-profile clients, Cerebras is poised to enter the public market at a time when investors are eager for dedicated AI infrastructure opportunities.
A successful IPO could enable Cerebras to accelerate its manufacturing processes, broaden its software offerings, and deepen its involvement in both training and inference workloads. This listing would signify more than just another hardware entry into the market; it could illustrate that meaningful competition to Nvidia is not only achievable but is already making significant inroads with top-tier clients in the tech industry.
See also
MediaTek Expands AI Chip Portfolio to Drive Growth in U.S. and Global Markets
China Doubles AI Accelerator Chips to 60,000, Boosting Scientific Computing Power
BMW Group and University of Zagreb Optimize Battery Cell Production with AI, Reducing Costs by 50%
AI Will Redefine Legal Education, Says University of Sydney’s Law School Leadership
NPL Integrates NVIDIA Ising AI to Automate Quantum Calibration and Enhance Stability
