Nobody wants to wait. As AI becomes useful, everyone wants it fast.
In the long arc of Silicon Valley ambition, Thursday marked a rare convergence of timing, technology, and personal reinvention: Cerebras Systems debuted on the Nasdaq with shares surging nearly 68 percent, crowning the company with a $67 billion valuation and its CEO, Andrew Feldman, with a fortune of $3.2 billion. The company's unconventional bet — building AI chips the size of a sheet of paper, optimized for the speed of inference rather than the brute labor of training — found its moment precisely because the world has grown impatient for artificial intelligence that actually works in real time. What markets rewarded on Thursday was not merely a product, but a philosophy: that sometimes the path others abandoned holds the most value.
- Cerebras stock surged 68% on its first day of trading, making it the largest tech IPO of the year and instantly reshaping the competitive landscape around Nvidia's dominance in AI chips.
- CEO Andrew Feldman, a Stanford-raised serial entrepreneur with a complicated past that includes a federal guilty plea, arrived at billionaire status through decades of iteration, failure, and an unconventional hardware bet most engineers had avoided.
- The company's oversized chips — once considered impractical due to power and heat risks — have become a sought-after solution for AI inference speed, attracting a $20 billion partnership with OpenAI and a distribution deal with Amazon AWS.
- A compensation structure approved in February signals the board's conviction that a $250 billion valuation is within reach, with Feldman's stock grants potentially exceeding $6 billion if that target is met.
- With the majority of last year's revenue tied to customers in the UAE, Cerebras' growth story is as geographically surprising as it is technologically bold, suggesting the AI chip race is being run on unexpected terrain.
On Thursday, Cerebras Systems made its Nasdaq debut to a market that had clearly been waiting. Shares jumped nearly 68 percent in a single session — the largest technology IPO of the year — valuing the AI chip maker at roughly $67 billion and elevating its 56-year-old CEO and cofounder, Andrew Feldman, to a personal fortune of $3.2 billion. "It's quite a day, isn't it?" he told Bloomberg Television, measured but unmistakably pleased.
Feldman's path to this moment was long and uneven. Raised on the Stanford campus by two professors, he drifted toward business rather than academia, joined a networking hardware startup as its first non-engineer, and sold it at 27 for $280 million. His next company went public during the telecom boom, though a 2007 document revealed a federal guilty plea tied to accounting violations during his time there — a chapter he moved past with probation and a fine. He went on to sell SeaMicro to AMD in 2012, and then began what would become Cerebras, founded in 2015 with four former colleagues in the offices of venture firm Foundation Capital.
The company's core bet was audacious: build chips the size of a sheet of printer paper, far larger than anything Nvidia or its peers were producing. Engineers had long avoided this path because larger chips demand more power, and excess power destroys silicon. Cerebras solved the problem — and when ChatGPT arrived in late 2022 and ignited global demand for fast AI inference, the bet looked prescient. "As AI becomes useful, everyone wants it to be fast," Feldman said. "Nobody wants to wait."
The partnerships followed. OpenAI committed to acquiring more than 750 megawatts of Cerebras processing capacity through 2028, a deal valued near $20 billion. Amazon announced it would offer Cerebras chips through AWS by year's end. Meanwhile, company filings revealed that a significant share of last year's revenue came from customers in the United Arab Emirates — an unexpected geography for a company now at the center of the AI hardware race.
Feldman holds 4.6 percent of Cerebras, making him its largest individual shareholder. Cofounder Sean Lie holds 2.5 percent, a stake now worth $1.6 billion. In February, the board approved compensation packages for both men that could yield Feldman more than $6 billion — and Lie more than $3.6 billion — if the company reaches a $250 billion valuation. The board, it seems, believes Thursday was not a ceiling but a starting point. Through it all, Feldman has maintained one ritual from the early days: firing up a smoker and barbecuing for his team, a tradition that followed Cerebras from a venture firm's back lot to a company now worth tens of billions.
On Thursday, Cerebras Systems made its debut on the Nasdaq, and the market responded with unmistakable enthusiasm. Shares of the artificial intelligence chip manufacturer jumped nearly 68 percent in a single day, the largest technology IPO of the year, valuing the company at roughly $67 billion. For Andrew Feldman, the company's 56-year-old CEO and cofounder, the moment represented something unprecedented in his long career as a Silicon Valley entrepreneur: a fortune officially calculated at $3.2 billion.
Feldman had already sold three companies and taken another public before Cerebras. He grew up on the Stanford campus—his parents were professors there—and had spent decades building and exiting ventures in the networking and server hardware spaces. But nothing in that résumé compared to what happened Thursday morning. "It's quite a day, isn't it?" he told Bloomberg Television, his tone measured but unmistakably pleased. "This is one of the largest technology IPOs in history, and also the largest IPO of a semiconductor company. We're extremely proud."
The timing was no accident. Since OpenAI released ChatGPT in November 2022, AI chip makers have become the most coveted stocks on the market. Nvidia has dominated the landscape, growing into the world's largest company by market capitalization and making its cofounder Jensen Huang the seventh richest person on Earth with a fortune exceeding $186 billion. But Cerebras, founded in 2015, has carved out a different niche. While Nvidia and others built their empires on smaller, more traditional chips, Cerebras developed something radically larger—a processor roughly the size of a sheet of printer paper. For decades, engineers had avoided this path because larger chips require more power, and that power can damage the silicon itself. Cerebras solved the problem, and the market wanted what they had built.
The company's chips have proven especially popular for inference—the process of running AI models rather than training them from scratch. As artificial intelligence becomes more practical and widespread, speed matters enormously. "There's extraordinary demand for fast inference right now," Feldman explained. "As AI becomes useful, everyone wants it to be fast. Nobody wants to wait." That simple observation captured the entire market dynamic. OpenAI announced in February that it would acquire more than 750 megawatts of Cerebras processing capacity through 2028, a partnership valued at roughly $20 billion. Amazon followed in March, announcing it would offer Cerebras chips through AWS by year's end.
Feldman's path to this moment was neither straight nor unblemished. The son of Stanford professors, he initially considered academia before switching to business school. A roommate asked him to write a financial plan for a networking hardware company, and he joined as the first non-engineer employee. Eighteen months later, at 27, the company sold for $280 million. His next venture, Riverstone Networks, went public in 2000 during the telecom boom, but a 2007 document revealed that Feldman had pleaded guilty to a federal crime related to accounting controls violations while serving as vice president of marketing. He received three years of probation and a $5,000 fine. He moved on to SeaMicro, a data center server design company that AMD acquired in 2012 for $334 million.
Cerebras began in the offices of Foundation Capital, a venture firm that still owns 7 percent of the company. Feldman started the project with four colleagues from SeaMicro, and he brought with him an unusual tradition: he would fire up a smoker behind Foundation's offices and barbecue for the team. As Cerebras grew, the tradition persisted. For a recent employee gathering, he smoked several hundred kilograms of ribs and chicken, a detail he shared on LinkedIn with evident pride.
The early years were difficult. Several of Cerebras's first products failed to gain traction. But ChatGPT's arrival changed everything. Demand for AI chips exploded, and suddenly Cerebras's larger-chip approach looked prescient rather than reckless. According to company filings, a significant majority of last year's revenue came from customers in the United Arab Emirates, suggesting the company had found its initial foothold in unexpected places.
Feldman holds 4.6 percent of Cerebras as the largest individual shareholder, while cofounder Sean Lie owns 2.5 percent, a stake now worth $1.6 billion. In February, the board approved aggressive new compensation packages for both men. If the company reaches a market valuation of $250 billion—an ambitious but not impossible target given current momentum—Feldman's stock grants could be worth more than $6 billion. Lie's would exceed $3.6 billion. The message was clear: the board believed Cerebras was only at the beginning of its ascent.
Notable Quotes
This is one of the largest technology IPOs in history, and also the largest IPO of a semiconductor company. We're extremely proud.— Andrew Feldman, CEO of Cerebras, on Bloomberg Television
There's extraordinary demand for fast inference right now. As AI becomes useful, everyone wants it to be fast. Nobody wants to wait.— Andrew Feldman, explaining the market opportunity for Cerebras chips
The Hearth Conversation Another angle on the story
What made Cerebras's chip design so different from what Nvidia and others were already doing?
They went bigger. Everyone else had been constrained by the physics—larger chips need more power, and that power generates heat that can destroy the silicon. Cerebras figured out how to make it work. A chip the size of a sheet of paper sounds simple until you realize no one else had managed it.
And that size advantage matters most for inference, not training?
Exactly. Training is expensive and rare—you do it once, usually. But inference happens constantly, millions of times a day. If you can run inference faster, you save money and time at scale. That's where the real market is.
Feldman had already made money multiple times before. What was different about this moment for him?
Scale. He'd sold companies, taken one public, but this is different. A $67 billion valuation, a 68 percent pop on day one. He's not just wealthy now—he's in a different category entirely. And the compensation package suggests the board thinks he's just getting started.
The accounting crime from 2007—does that linger?
It's in the record, so it's part of his story. But markets move forward. By the time Cerebras went public, investors were focused on the technology and the partnerships with OpenAI and Amazon. The past doesn't disappear, but it gets smaller in the rearview mirror.
What does the $250 billion valuation target tell you?
That the board thinks AI infrastructure is going to be enormous. Bigger than most people currently imagine. If they hit that number, Feldman's additional grants could be worth $6 billion. That's not a conservative bet. That's a belief in exponential growth.