- Cerebras priced its IPO at $185 per share and raised $5.5 billion, well above its initial $115 to $125 range.
- Shares opened at $385, a 108% pop, putting the fully-diluted valuation at $56.4 billion on day one.
- 2025 revenue climbed 76% to $510 million and the company swung from a near half-billion-dollar loss to a $237.8 million profit.
- Customers include OpenAI, AWS, G42 and Saudi Arabia’s MBZUAI, positioning Cerebras as a direct inference rival to Nvidia.
A year ago, this day looked like it would never come for Cerebras. The wafer-scale chip designer’s first IPO attempt in 2024 was strangled by a CFIUS review tied to its Abu Dhabi backer, G42. On Thursday, May 14, 2026, the same company rang the Nasdaq opening bell, raised $5.5 billion, and watched retail investors push its stock up 108% in a single morning. The question is no longer whether Cerebras has a story. It is whether anyone still wants to bet against an AI chip company that is not Nvidia.
The Day Cerebras Doubled
From $185 to $385 in a Single Morning
The pricing path tells the story on its own. Cerebras initially guided to an IPO range of $115 to $125 per share, then revised it upward to $150 to $160 as demand built. On Wednesday evening it priced at $185, a level already 16% above the high end of the updated range, while expanding the offering to 30 million shares. By the time public trading opened Thursday morning, the stock printed at $385, more than double the IPO price, before cooling to above $330 mid-day.
At the $185 IPO price, the company carried a fully-diluted valuation of $56.4 billion. At the $330 to $385 trading range, that figure reaches roughly $100 billion on a fully-diluted basis. Co-founder and CEO Andrew Feldman’s stake is worth nearly $1.9 billion at the IPO price alone. Co-founder and CTO Sean Lie’s stake clears $1 billion. For founders who watched their original 2024 listing collapse under regulatory review, the reversal is total.
Trend Insight — A 108% first-day pop is not a normal IPO outcome. It signals that institutional bookrunners under-priced demand and that retail and hedge-fund desks are willing to pay any multiple for a credible non-Nvidia AI chip story. The same pattern showed up in late-cycle 1999 networking IPOs. The signal is real demand; the question is sustainability.
Why Investors Are Salivating Now
Revenue Doubled, Losses Erased
The financial profile that finally unlocked the IPO was filed in April. Cerebras reported 2025 revenue of $510 million, up 76% year-over-year. More striking, the company swung from losing nearly half a billion dollars in 2024 to posting $237.8 million in net income in 2025. For a fabless semiconductor company that had been burning cash to develop and ship the largest single chip ever built, that swing is the difference between a science project and a profitable platform.
The 2024 IPO had stalled for two reasons. The Committee on Foreign Investment in the United States dragged out its review of G42’s stake, and G42 itself accounted for almost all of Cerebras’ revenue, which spooked institutional investors. Both problems eased in 2026. Customer concentration dropped as new buyers came on board, and the regulatory cloud lifted enough for underwriters to move forward. Investors who passed in 2024 had to chase the deal in 2026.
Trend Insight — The 2024-to-2026 turnaround is a case study in how AI infrastructure economics have shifted. Revenue scaled because inference workloads exploded, not because a single sovereign buyer doubled its order. Once revenue diversifies, regulatory risk shrinks and valuation expands. Other AI hardware companies stuck with concentration issues should be watching this template closely.
The Nvidia Question
Cerebras’ Bet on Inference
The market did not pay 108% above IPO for a training-chip story. It paid for inference. Cerebras has positioned its wafer-scale system as a specialist for the ongoing compute load required when models answer prompts at scale, not the one-time work of training them. That is the segment growing fastest as Claude, ChatGPT, Gemini and dozens of agent products move from beta to billions of daily calls. Inference is where margins compound and where Nvidia’s pricing power is most exposed.
The customer list reads accordingly. OpenAI is now a Cerebras customer through what TechCrunch describes as a “complicated circular-deal relationship.” Amazon Web Services is a customer. G42 remains, alongside Saudi Arabia’s Mohamed bin Zayed University of Artificial Intelligence (MBZUAI). The presence of OpenAI and AWS on the same purchase order is the part that should worry Nvidia. If hyperscalers and frontier labs are willing to validate a second supplier publicly, the era of single-vendor inference is closing.
Trend Insight — Nvidia is not losing share today, but the supply-chain bargaining position is shifting. Every credible alternative IPO raises the cost of single-vendor lock-in. Expect frontier labs to negotiate harder, and expect more inference-first competitors (Groq, SambaNova, Tenstorrent) to file in the next 12 months. The Cerebras debut is a permission slip for the rest of the field.
Related
- OpenAI Spent $4B. Microsoft Just Spent $1B to Beat It.
- Anthropic and Microsoft’s Maia chip deal explained
- 12 Million Tokens for $8 — Yes, Really
- SAP Just Bet the Whole Company on 200 AI Agents
- Claude Opus 4.7 Fast Mode vs GPT-5.5
Sources
- TechCrunch — Cerebras raises $5.5B, then stock pops 108%, in the first huge tech IPO of 2026
- TechCrunch — OpenAI’s cozy partner Cerebras is on track for a blockbuster IPO
- SEC — Cerebras Systems S-1/A Filing (April 2026)
AI Biz Insider · AI Trends EN · aibizinsider.com
댓글 남기기