Cerebras seeks higher IPO range on strong demand

Cerebras Systems, may raise its IPO range to US$150-US$160 a share from US$115-US$125 and increase the offer to 30 million shares from 28 million as investor demand surges, according to sources.

At the top of the revised range, the Sunnyvale, California-based AI chipmaker could raise about US$4.8 billion instead of the previously expected US$3.5 billion, though the terms could still change.

Orders reportedly exceed 20 times the available shares, potentially making Cerebras the world’s largest IPO so far this year, according to Dealogic.

Cerebras is making its second attempt to go public after withdrawing a 2024 filing tied to a US security review involving G42. The company now counts Amazon and OpenAI among its customers.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

Cerebras reshaped its IPO path over the past year

Cerebras pulled its 2024 IPO after delays from a US national security review tied to G42, an Abu Dhabi AI company and minority investor, which led it to cancel its roadshow 1.
In March 2025, Cerebras got clearance from the Committee on Foreign Investment in the United States (CFIUS), a US panel that reviews deals for national security risks, for G42’s planned share purchase and could press ahead with listing plans 2.
A deal with OpenAI helped reduce reliance on G42, which made up 87% of revenue in the first half of 2024 3.
The company also moved from mainly selling chips to running them in its own data centers as a cloud service. It reported US$24.6 billion in remaining performance obligations, or contracted future revenue not yet recognized, on Dec. 31, 2025 3.

Cerebras’s IPO gives specialized AI inference chips a market test

Strong demand for the offering suggests the AI chip market could spread beyond Nvidia, which holds 94.4% of the graphics processing unit (GPU) market according to Futurum Research, a technology advisory and research firm 4.
That momentum supports the view that AI inference, or replying to user queries, can reward chips built for that task over general-purpose GPUs 4.
It may also speed up spending on other chip designs as cloud providers such as Amazon weave specialized hardware into their AI platforms 4.
Still, Cerebras faces execution risk as it shifts from shipping hardware to running data-center infrastructure at far larger scale and delivering cloud services needed to meet its contracts 4.