Cerebras IPO Prices at $150–$160, Targeting $50B Valuation

Cerebras, a company that makes AI chips with wafer-scale processors made just for inference tasks, is set to start trading on Thursday. The company priced its initial public offering (IPO) at $150 to $160 per share, with the goal of raising $4.8 billion and reaching a market value of nearly $50 billion.

The company raised its target range from $115 to $125 to meet higher investor demand, which Morningstar senior equity analyst Brian Colello called "remarkable" considering that Cerebras was only worth $8 billion in October, before its big deals with OpenAI and AWS changed its business story.

With $50 billion, Cerebras is the biggest U.S. IPO in the last 12 months, beating CoreWeave's $23 billion start in March 2025. However, it's still less than the $135 billion that SpaceX and OpenAI are expected to raise when they go public later this year.

What Cerebras Actually Does — and Why the Timing Is Right

Cerebras is in the inference segment of the AI chip market. This is the phase where trained models run in real life and people interact with them. This is different from the training phase, which is when models are made. As Dimitri Zabelin, senior investment research analyst at PitchBook, put it, "The AI hardware market rotated from training-cycle dominance towards inference-cycle scaling, where token generation speed and cost per query determine competitive positioning." This difference is very important for 2026.

The structure of cerebras' chips is different. There are 58 times more wafer-scale processors made by this company than Nvidia's B200. Its fault-tolerant architecture gets around manufacturing flaws that would make standard large-die chips unusable in the business world.

The chips don't use traditional HBM memory, but instead use SRAM memory, which gives up some capacity for speed. This lets the chips respond to inference requests up to 15 times faster than top GPUs for some tasks. For uses where latency is the main problem, like real-time AI bots, conversational AI, and live inference at scale, that speed advantage is useful for business.

The proof from business is real. An agreement with Amazon Web Services to offer Cerebras compute to cloud customers and a $20 billion compute deal with OpenAI (structured as a chip procurement promise in exchange for an OpenAI equity stake and a $1 billion loan) give the company a way to make money that most AI companies in the IPO stage don't have. These deals raised the value from $33.6 billion when the IPO was filed in March to the $50 billion goal set now.

The Sovereign AI Wildcard That Most Coverage Is Missing

It's interesting that 86% of Cerebras's revenue came from two companies in the UAE in 2025, mostly through a deal made in 2024 with Abu Dhabi's G42 technology company. This concentration caused U.S. national security concerns, which slowed down the company's IPO last year and could happen again if geopolitical conditions change. That concentration led to U.S. national security concerns, which slowed down the IPO last year and could happen again if global conditions change.

Zabelin says the story of independent AI is now a tailwind instead of a headwind. "Now that countries have finished training their national models for the first time, they are turning their attention to putting inference-focused hardware in place." That's starting a huge buying cycle for infrastructure that is designed for inference and is paid for by state capital. "Cerebras is in a good position to gain."

It's important to note the framing: It used to be a hardware company that was affected by regulations. Now, Zabelin describes Cerebras as "a diversified infrastructure player positioned at the centre of two converging tailwinds—the sovereign AI buildout and the incoming inference tsunami." It's still unclear whether that description will hold up under the close scrutiny that public markets apply after an IPO.

The Risk Factors Investors Need to Weigh

Any honest evaluation will show that the competition is very strong. Colello named Nvidia's Groq business unit as Cerebras' main rival; both are focusing on AI inference but use fast SRAM memory instead of HBM. Cerebras rivals with AMD, Google, AWS, Microsoft, and Meta on custom inference chips, as well as Broadcom, Marvell, and MediaTek, which are specialised chip designers that work with hyperscalers. "We anticipate plenty of competition in AI inference over time," Colello added.

The biggest financial risk right now is having too many customers. Colello himself says that the OpenAI deal is very important to Cerebras' value. It's set up in a way that depends on OpenAI's ability to make more money, because that will affect whether it keeps buying Cerebras chips at the agreed-upon amounts. "

A potential risk is that OpenAI fails to meet its own revenue targets and scales back," Colello pointed out. That concentration isn't a tail risk for a company with less than 800 workers that makes most of its money from two customers and two countries; it's the base scenario risk.

Should You Buy on IPO Day? What History Actually Says

When AI companies go public, the stock usually goes through a time of price discovery that offers better entry points than the IPO price. This price discovery period usually lasts for a few days to a few months after the IPO. SANDISK—up almost four thousand percent in a year. CoreWeave: it's 172% higher since its initial public offering. These two cases show that people who put money into AI infrastructure early on have been rewarded. Historically, though, dips after an IPO have been better starting points than the IPO price itself.

Cerebras' value went from $8 billion in October to $50 billion at IPO. This is because of the speeding up of business due to the deals with OpenAI and AWS, which was already reflected in the price before Thursday's first trade. There is a real long-term case for investors who believe in the inference cycle thesis and Cerebras' speed edge. It's possible for those buyers to find better times to buy in in the months after the IPO than when the price was $150 to $160.

The change from training to inference makes real demand for Cerebras's tailored chips. The company was founded at the perfect time in AI's history. OpenAI and AWS's deals give the companies business validation that most companies in the IPO stage don't have.

The autonomous AI boost is real and getting stronger. All of the risks—concentration of customers, circular deal structures, and competition from Nvidia's Groq—are real, and the present valuation doesn't fully address them. The market values at $50 billion the ability of a company with less than 800 workers to carry out a long-term plan with high goals. From what we know about the past, long-term buyers may do better by waiting for prices to settle after an IPO rather than buying on Thursday's open.