- What Does Cerebras Even Do?
- The Cerebras-OpenAI Deal That Changed Everything
- Cerebras vs Nvidia: "Or" Is the Wrong Word
- Should You Buy CBRS Now? The Honest Answer
- The Bigger Picture for the Semiconductor Market
A chip the size of a dinner plate just walked onto Wall Street and the street gave it a standing ovation. On May 14, 2026, Cerebras Systems (Nasdaq: CBRS) made its public debut, opening at $350, nearly double its IPO price of $185, made an intraday high of $385 and finally closing the day up 68% at $311.07. CBRS Stock is further up 5.35% in after hours trading. According to Yahoo Finance and Bloomberg, the company raised $5.55 billion making it the largest US tech IPO since Uber went public in 2019. Institutional demand was so fierce it was 20x oversubscribed. And in the backdrop? Nvidia's stock also had a good day, rising 4%. So no, the story here isn't a fight. It's something far more interesting.
Let's break down what Cerebras actually builds, why this IPO caught fire the way it did, and what you, as an investor, should do with all this information right now.
What Does Cerebras Even Do?
Think of a regular AI chip like a Mumbai local train system where many small trains (chips) run on parallel tracks, connected by thousands of bridges. Data has to cross those bridges constantly, and every crossing takes time. That's how Nvidia's GPUs work: brilliant, powerful, but built around the idea of many smaller chips talking to each other at high speed.
Cerebras threw that playbook out. Their Wafer Scale Engine 3 (WSE-3) is one single chip; the entire silicon wafer left intact, not sliced into smaller pieces. According to the company's SEC filings, the WSE-3 packs 4 trillion transistors and 900,000 cores onto a chip that is 58 times larger than Nvidia's B200 GPU.
The result: data never has to leave the chip. No bridges. No wait. Just raw speed.
Cerebras claims this makes their system up to 21 times faster for AI inference tasks; that's the step where a trained AI model actually answers your question, compared to Nvidia's DGX B200, at 32% lower total cost of ownership.
The Cerebras-OpenAI Deal That Changed Everything
Cerebras isn't just a cool chip company. It's a chip company with receipts.
In January 2026, OpenAI signed a multi-year compute deal with Cerebras valued at over $20 billion, as per Bloomberg. And earlier this year, OpenAI launched its first AI model running entirely on Cerebras hardware and not Nvidia.
Amazon Web Services and Meta Platforms are also listed as customers in the IPO prospectus. When the world's most valuable AI companies vote with their wallets, the market pays attention.
Revenue in 2025 hit $510 million, up 76% year-over-year, as per the company's SEC filing. Net income swung from a loss of nearly $500 million in 2024 to a profit of $237.8 million in 2025. That's quite a dramatic turnaround. It’s worth noting that this is the GAAP net income which includes a massive one-time accounting gain of $363 million. This gain came from restructuring a commercial agreement with G42 (the UAE firm). The adjusted net income for cerebras comes out to be $87.9 million.
Cerebras vs Nvidia: "Or" Is the Wrong Word
Here's the most important thing to understand: this is not a zero-sum game.
Nvidia still controls roughly 80-90% of AI accelerator market share, as per industry estimates. Its fiscal year 2026 revenue was $215.9 billion; that's 423 times Cerebras' revenue. And Nvidia's CUDA software ecosystem, built over 15+ years, keeps developers deeply anchored to its hardware.
Cerebras is not attacking all of Nvidia's market. It's attacking a specific, fast-growing slice: high-speed inference at scale. Think of Nvidia as the general-purpose power grid and Cerebras as the high-speed EV charging lane for AI applications that need instant responses.
The VanEck Semiconductor ETF is already up 58% in 2026. The market for specialized AI silicon is expanding fast enough for multiple winners. So, it’s not a question of Cerebras “or” Nvidia, but rather a question of Cerebras “and” Nvidia and how much of each of them.
Should You Buy CBRS Now? The Honest Answer
No, not immediately at least, and here's why.
At its IPO close, Cerebras was trading at roughly a $66-95 billion market cap, depending on dilution basis, as per Bloomberg. With $510 million in revenue, that's a price-to-sales multiple of over 100x. You're not paying for what Cerebras earns today. You're paying for what analysts expect it to earn in 2028-2030.
There are also real risks. The Mohamed bin Zayed University of Artificial Intelligence in the UAE accounted for 62% of Cerebras' 2025 revenue, as per its SEC prospectus. Customer concentration at this level is a flag. Additionally, Nvidia's next-generation Blackwell chips are designed to close the inference gap, and rivals like Groq and AMD's MI300X are fighting for the same niche.
History also works against the FOMO buyer. Post-IPO pops are frequently followed by a "quiet period hangover", a 15-30 day drift when initial excitement cools and analyst lockups expire. Investors who bought Uber or Rivian at their opening highs waited years to see those prices again.
The smarter play: watch the first earnings report. Watch whether OpenAI and AWS ramp up usage. And if the stock dips 20-30% in the weeks ahead, as many IPOs do, that's likely a better entry point for a long-term position.
The Bigger Picture for the Semiconductor Market
Cerebras' debut signals something broader. The AI chip market is splitting into two distinct lanes: general-purpose compute (Nvidia, AMD) and specialized inference hardware (Cerebras, Groq, custom silicon from Google and Amazon).
Investors who understand this bifurcation and make a position across both, will likely do better than those treating it as a binary choice.
The real question isn't whether Cerebras kills Nvidia. It's whether the AI inference market grows large enough for Cerebras to justify a $95 billion valuation by the time its 2028 revenue arrives. Given the OpenAI deal, the AWS partnership, and the structural shift from training to inference workloads, the answer isn't obviously no.
But at today's prices? You'd be buying a perfect future. And perfect futures are rarely on sale for long.