Cerebras, the IPO offering price set at $185... Expectations for an AI inference sector "Nvidia replacement" are heating up

Cerebras Systems sets its initial public offering (IPO) price at $185 per share.
This price far exceeds the previously announced range of $150 to $160, once again confirming market expectations for one of this year’s most closely watched tech IPOs.

Through this offering, Cerebras Systems will raise at least $5.5 billion, approximately 8.2104 trillion Korean won.
Recently, the semiconductor industry has seen strong overall buying interest, and analysts believe that the investor psychology seeking alternatives to Nvidia has driven the success of this issuance.

The largest tech IPO of the year… company valuation reaches $56.4 billion

Cerebras Systems will trade under the ticker symbol “CBRS” after listing.
Based on the offering price, its fully diluted company valuation is approximately $56.4 billion, about 84.1939 trillion Korean won.
CEO Andrew Feldman’s holdings are valued at about $1.9 billion, roughly 2.8363 trillion Korean won.

This IPO is seen as a rare large-scale tech IPO in recent years.
Since Uber raised about $8 billion in 2019, cases of larger scale are few.
Subsequent notable large IPOs include Snowflake raising about $3.8 billion in 2020, and Rivian raising $12 billion in 2021.

Especially, the process of raising the offering price also reflects market expectations.
In documents filed on May 4, Cerebras Systems stated it would sell 28 million shares at $115 to $125 per share, but a week later increased the offering to 30 million shares and raised the expected price to $150 to $160.
The final offering price was set at $185.

Expected to be an Nvidia alternative… AI inference market is key

Cerebras Systems has attracted attention because it is seen as an alternative in the “inference” domain of the AI semiconductor market.
Inference is the process of applying trained AI models to real-world services to produce answers or results.
So far, the market has heavily relied on Nvidia’s graphics processing units (GPUs), but companies are seeking new options in terms of cost and performance.

The company’s flagship product is a large chip called “WSE-3.”
This chip, nearly the size of a dinner plate, is much larger in structure than Nvidia’s Blackwell B200 GPU.
Its core advantage is its 44GB SRAM memory.
SRAM is much faster than the DRAM used in typical servers, but also significantly more expensive.

Cerebras Systems states that the 900k cores on the WSE-3 can access on-chip SRAM with a one-clock-cycle latency.
Thanks to this architecture, it can deliver faster processing performance than existing GPUs in large-scale AI inference tasks.
In other words, its competitiveness in real-world service operation is more prominent than during AI training phases.

Overcoming dependence on G42 and diversifying clients

However, the listing process was not smooth sailing.
Founded in Silicon Valley in 2016, Cerebras Systems announced an IPO plan in September 2024 but withdrew the application about a year later.
The prospectus at that time showed that UAE AI company G42 accounted for 80% of its annual chip sales, raising concerns about excessive reliance on a specific customer.

Additionally, the U.S. Committee on Foreign Investment in the United States (CFIUS) reviewed Cerebras Systems’ partnership with G42 from a national security perspective, adding uncertainty.
Although both companies were later found not to have violated any laws, market attention to the stability of its revenue structure persisted.

The company subsequently adjusted its business model.
It no longer solely focuses on hardware sales but shifted its emphasis to cloud services based on its chips.
This strategic shift means Cerebras Systems now competes not only with Nvidia but also with Google Cloud, Amazon Web Services, Microsoft ($MSFT), Oracle ($ORCL), CoreWeave, and others.

According to the latest prospectus, G42’s revenue share has fallen to 24% in 2025.
However, most new revenue comes from another UAE entity—Mohammed bin Zayed Artificial Intelligence University, which accounts for 62% of 2025 revenue.
While client diversification is ongoing, the high proportion of Middle Eastern clients remains a challenge to be addressed.

Contracts with OpenAI and AWS boost investor sentiment

Nevertheless, investors believe that revenue composition could change rapidly in the future.
In January this year, Cerebras Systems announced a three-year contract with OpenAI, providing 750 MW of computing capacity valued at $20 billion.
Converted to Korean won, this is approximately 29.856 trillion KRW.

Then, in March, Amazon Web Services announced it would provide access to Cerebras Systems’ chips through its cloud AI service “AWS Bedrock.”
This is seen as a signal that Cerebras Systems is no longer just an experimental alternative but is entering the core AI infrastructure supply chain.

Market views suggest such contracts could be opportunities to expand the customer base and reduce dependence on specific regions and clients.

Major investment firms and Silicon Valley figures are also involved

The composition of major shareholders is also notable.
Fidelity Investments is reportedly the largest investor, holding shares worth about $3.8 billion, approximately 5.6726 trillion KRW.
Benchmark Capital is also said to hold shares worth about $3.3 billion, roughly 4.9262 trillion KRW.

OpenAI President Greg Brockman estimates his holdings are worth about $14.4 million, approximately 82.1k KRW.
OpenAI CEO Sam Altman’s holdings are valued at about $16.5 million, roughly 841.94k KRW.

Holger Mueller of Constellation Research commented that Cerebras Systems offers a “substantial alternative” to Nvidia in AI inference, and the market shows a strong willingness to fund it.
Analysts believe that increased competition will lower the costs for companies adopting AI, potentially accelerating industry innovation.

The success of Cerebras Systems’ IPO is seen not only as a company going public but also as a sign that the AI semiconductor market is gradually shifting from a GPU-centric focus toward diversification.
However, the stock price after listing will likely depend on whether the company can truly diversify its revenue and demonstrate profitability from large contracts.

TP AI Notice: This article is summarized based on the TokenPost.ai language model.
Main content may be omitted or differ from actual facts.