Cerebras Stock Sinks 31% Below IPO Price Amid Lockup Expirations

From a blockbuster debut to below the offer price

Cerebras Systems was the hottest listing of 2026 five months ago. Today its shares trade well below the price institutions paid for them. The AI chipmaker priced its initial public offering at $185 per share on May 13, selling 30 million Class A shares22. The deal raised $5.55 billion, the largest IPO by a U.S. tech company since Uber in 201921. Orders came in at more than 20 times the shares available, so the price landed above a marketed range of $150 to $160. That range had already been raised from $115 to $12523.

The first trading session was dramatic. Shares opened at $350, hit $386 during the day, and closed at $311.07, a 68% gain on the offer price21. The latest trading tells a very different story. On October 16, CBRS closed at $127.21 after a 10.52% one-day drop, which left the company's market value near $30.22 billion16. That price is about 31% below the IPO price and roughly two-thirds below the first-day peak.

The slide happened in stages. Several overlapping pressures hit at once: scheduled share releases, insider sale filings, doubts about Cerebras's most important customer, and a wider selloff in AI infrastructure stocks. Coverage gives different weight to each factor. Taken together, the reporting points to one conclusion: Cerebras priced for perfection and then met conditions that were far from perfect.

What valuation, exactly?

The size of the drop depends partly on which valuation you start from, and published figures vary widely. Reuters put the fully diluted valuation at $56.43 billion at the IPO price23. IPOScoop gave a market cap of $39.82 billion at pricing, apparently counting only outstanding shares. CNBC said the first-day close valued Cerebras at about $95 billion21. Yahoo Finance said the debut pushed market capitalization to almost $70 billion on outstanding shares, or about $86 billion fully diluted, citing Bloomberg data25.

The differences come from whether restricted stock, options, and warrants are counted. That accounting question is closely tied to the lockup story. Any measure that leaves out restricted shares makes the tradable float look scarcer than it will be once those shares are freed. On the current ~$30 billion figure, the stock now trades below every valuation calculated at the IPO price, not just below the first-day spike16.

An unusual lockup that front-loaded supply

Cerebras did not use the standard design where every insider becomes free to sell on day 180. Its lockup has staggered early releases tied to earnings dates. The main restriction ends at the earlier of two dates: the second trading day after third-quarter results, or 180 days after the May 14 prospectus4. The IPO filing estimated that up to about 171.1 million shares could be released during the lockup period4. Including the underwriters' full option, the offering itself was only 34.5 million shares, so the potential release pool is close to five times the original float4.

Some analysts flagged this risk right after the debut. A Seeking Alpha piece published days after listing noted that more than 60 million shares would come free by the second-quarter earnings release9. An August analysis put the first earnings-linked release at about 27.7 million shares after first-quarter results. It said the largest single wave, roughly 36.4 million shares, followed the August 12 second-quarter report1.

On the October release dates, the reporting does not agree. 24/7 Wall St. said a lockup ended on September 30, making about 19.4 million shares eligible for sale11. Other outlets put the release of an estimated 19.4 million shares on October 14815. A German market report described the October 14 tranche as about 19.4 million shares held by non-executive employees10. The most likely explanation is that the staggered schedule produced more than one release this autumn. Readers should be careful with any single date. Startup Fortune's statement that the October 1 release covered "roughly 13% of IPO shares" does not easily match the other figures13.

The more careful coverage repeats one point: being allowed to sell is not the same as selling10. That is true. Still, the stock fell on heavy volume around both windows, which suggests supply is weighing on the price whether or not holders actually sold15.

Insiders file to sell

Insider filings added to the pressure. On the September 30 window, chief operating officer Dhiraj Mallick filed a Form 144 to sell 396,000 shares worth about $77.9 million. Chief financial officer Robert Patrick Komin Jr. filed for 32,500 shares worth about $6.39 million11. A Form 144 records an intention to sell, and only a later Form 4 confirms a completed sale11.

Startup Fortune reported earlier insider activity. It said CTO Sean Lie filed a Form 144 for more than $153 million and CEO Andrew Feldman sold $5.92 million. It described these sales as linked to taxes on vested equity13. That outlet argued the timing looks bad even if the motive was routine, and that seems right2. When officers' sale filings arrive alongside a lockup release and a falling stock, they undercut the case that management sees the decline as a buying opportunity.

The OpenAI question

The customer story probably matters more than share supply. On September 30, SemiAnalysis reported that the Ultrafast tier of OpenAI's GPT-6.1 Sol runs on Nvidia GPUs instead of Cerebras hardware. By the same account, the earlier GPT-5.6 Sol had run on Cerebras systems since August8. Neither company confirmed the claim15.

The report hits hard because of how concentrated Cerebras's business is. The company disclosed $25.4 billion in remaining performance obligations at the end of June, and a large share of that is tied to OpenAI15. One early analysis estimated the backlog was nearly 80% dependent on OpenAI9. TipRanks added a second worry: reports that Microsoft wants to move to its own Maia chips by 20276.

Cerebras is working to diversify. It recently agreed to supply AI systems to Gimlet Labs for about 100 megawatts of planned capacity, though financial terms were not disclosed8. One deal of undisclosed size does not offset that level of concentration.

Fundamentals versus the sector tide

The hardest question is whether Cerebras is being punished for its own results or for its sector. Coverage leans different ways. Startup Fortune emphasizes improving fundamentals. It reports second-quarter revenue of $209.9 million, up 103% from a year earlier, with cloud revenue up 287%, and notes that UBS raised its price target to $33013. The Motley Fool says the second-quarter report disappointed investors and sent the stock down 20%16. Both can be true. Strong growth can still fall short of expectations built into a stock trading at roughly 217 times forward earnings11.

The pattern of concern started early. The stock first dropped below its IPO price on June 24, after the company guided second-quarter core gross margins to 36%–38%, down from 46.5% in the first quarter19.

The broader market has clearly made things worse. CoreWeave dropped about 16% on October 13, and its bonds yielded close to 15%, a sign of doubt about AI infrastructure financing13. A Financial Times report that OpenAI's annualized revenue was below some earlier estimates added to concerns about demand5. Cerebras's 11% fall on Friday, October 16, came during a broad chip retreat linked to inflation and worries about Federal Reserve rate hikes, not any company news10.

Analysts have not cut their targets to match the stock. MarketBeat showed an average 12-month target near $299.9015, and 24/7 Wall St. counted three strong-buy and seven buy ratings11. Targets that sit more than double the share price usually mean analysts are lagging the market, not that the stock is cheap.

What comes next

My reading is that the lockup releases set the timing of the decline, but the OpenAI uncertainty and the sector selloff set its size. The next catalyst combines both. Financial calendars disagree on the third-quarter report date, with estimates ranging from November 10 to November 19, and the company has not confirmed one15. That report will also trigger the final lockup expiration4. Investors will get the biggest test of operating results and the biggest wave of new supply within days of each other. Cerebras needs revenue in line with its guidance and a clear answer on its OpenAI workloads. Without both, the gap between its fundamentals and its share price is likely to widen.