US AI chipmaker Cerebras targets $4b IPO

Sunnyvale, California-based AI chipmaker and data center operator Cerebras Systems is seeking up to US$4 billion in an IPO that could value it at about US$40 billion.

Formal marketing could begin as soon as May 4 and banks have received more than US$10 billion in indications of interest though the size, valuation, and timing may still change.

That target is up from about US$2 billion in March and follows a February funding round that raised about US$1 billion at a US$23 billion valuation, including new capital.

A recent SEC filing showed Abu Dhabi-based AI firm G42 provided 24% of 2025 revenue versus 85% a year earlier and Cerebras said in March 2026 that issues tied to a US national security review had been resolved.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Reported profit came from an accounting gain while customer risk stayed concentrated

Cerebras posted net income of US$237.8 million in 2025, though it still lost money from operations 1.
The net income came from a US$363 million non-cash accounting adjustment. Without it, operating loss was US$145.9 million 1.
Revenue also depended on two customers in the United Arab Emirates 2.
Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), an Abu Dhabi AI research university, made up 62% of 2025 revenue. G42, an Abu Dhabi AI company, made up 24% 2.

OpenAI tie-up includes a loan and share purchase rights

Remaining performance obligations reached US$24.6 billion as of Dec. 31, 2025. More than US$20 billion came from an OpenAI agreement 1.
OpenAI gave Cerebras a US$1 billion loan to build data center infrastructure and provide services under a broader agreement 2.
OpenAI holds warrants, which are securities that give the holder the right to buy shares later, for as many as 33.4 million shares. They vest in full only if OpenAI buys 2 gigawatts of computing power from Cerebras 2.
The structure gives an AI developer financing and equity-linked incentives from its computing supplier, a setup other AI labs could copy as they look for capacity beyond Nvidia 2.