Cerebras files for IPO as OpenAI and AWS fuel chip demand

The AI chip market is about to get a public-market test for one of Nvidia's most visible challengers. TechCrunch reports that Cerebras Systems has filed to go public, reviving an IPO path it abandoned after a 2024 filing was delayed by federal review of an investment from Abu Dhabi-based G42.

Cerebras is known for wafer-scale AI hardware designed for training and inference. CEO Andrew Feldman has described the company's systems as the fastest AI hardware for those workloads, and the pitch has become more relevant as frontier labs and cloud providers run into the economics of large-scale inference. The question for investors is no longer whether AI compute demand exists. It is whether Cerebras can turn that demand into durable, profitable revenue.

The company has raised aggressively while waiting for the public window to reopen. TechCrunch cites a $1.1B Series G last year, followed by a $1B Series H in February at a $23B valuation, based on Wall Street Journal reporting. That places Cerebras among the highest-profile infrastructure bets in the current AI cycle.

Recent customer signals explain why the IPO filing matters now. Cerebras has announced an agreement with Amazon Web Services to use Cerebras chips in Amazon data centers. It has also been linked to an OpenAI deal reportedly worth more than $10B. Feldman told the Wall Street Journal that Nvidia did not want to lose OpenAI's fast inference business and that Cerebras took it from them, according to TechCrunch.

The filing also gives investors real numbers to test against that story. TechCrunch says Cerebras reported $510M in revenue for 2025. Net income was $237.8M, though excluding certain one-time items, the company posted a non-GAAP net loss of $75.7M. That mix is exactly why a public listing is important: AI infrastructure demand can be enormous, but hardware margins, customer concentration, and supply-chain execution still decide whether the business compounds.

Cerebras has not disclosed how much it wants to raise in the IPO. A spokesperson told TechCrunch the offering is planned for mid-May. If it proceeds, the listing will become a benchmark for how public markets value non-Nvidia AI compute suppliers after two years of private-market enthusiasm. The company does not need to replace GPUs everywhere to matter. It needs to prove that enough high-volume inference workloads need a different architecture and will pay for it repeatedly.