The Cerebras ticker illuminated the screen above the Nasdaq MarketSite in Times Square on Thursday afternoon, May 14. A company that had priced its shares at $185 the previous evening opened at $350. That is not typical, nor is it a typo. Before anyone had even finished their morning coffee, the AI chipmaker priced its initial public offering (IPO) at $185 per share and opened at $350, an 89% increase. It was difficult to avoid seeing something akin to panic in the purchases while looking at the order book. A lot of people who wanted to enter just couldn’t.
Even by themselves, the raw numbers are startling. In its Nasdaq debut on Thursday, Cerebras Systems surged 68%, closing at $311.07 after selling shares at $185, significantly above the company’s anticipated range. The company raised $5.55 billion by selling 30 million shares, making it the biggest tech company IPO since Uber’s 2019 launch. Demand is more than twenty times greater than supply. Reading the chatter on the trading floor gives the impression that this was not at all about Cerebras. Before the door closed, the goal was to gain any unadulterated exposure to AI chips.
The beginning of the story is what gives it its texture. Ten years ago, the company’s first office was located on the second floor of Foundation Capital in Menlo Park, according to a post made by an early backer. An office manager had created the floor plan by hand. The founders are now billionaires. The stakes held by CTO Sean Lie and CEO Andrew Feldman are valued at $1.7 billion and $3.2 billion, respectively. For that kind of arc, ten years is quick, but for those who endured the long nights it required, it is slow. Both of these statements are accurate.
The technology that powers all of this is genuinely strange, but in a good way. The majority of AI is powered by racks of interconnected GPUs. In contrast, Cerebras developed a single, massive processor. Wafer-scale AI processors, which are about the size of a dinner plate and have about four trillion transistors etched onto a single piece of silicon, are designed by Cerebras. The wager is that bottlenecks will be eliminated along with the spaces between thousands of tiny chips. It’s not a slide-deck argument; it’s an actual engineering one. It’s a different and more difficult question whether it can compete commercially with Nvidia.
Then Friday arrived. Following the initial spike, CBRS began to see some profit-taking, with a 10.1% decline in share price over the course of one day. Like a fever breaking, that pullback felt almost healthy. We might have learned more on the second day than on the first. By Monday, the stock had recovered and was rising once more due to rumors of fast-track index inclusion. Since then, the chart has resembled a rollercoaster with large swings, numerous day-trader fingerprints, and no stable price.
Finding skepticism is not difficult, and it is not irrational. Cerebras’s revenue increased by 76% to about $510 million last year. This is a significant increase, but it is small compared to the company’s peak valuation of almost $100 billion. Following the opening, analysts who had previously deemed the $185 IPO price fair ceased to do so. Additionally, a significant portion of last year’s revenue came from a single university in the United Arab Emirates. The OpenAI and Amazon agreements are significant because they imply that the base is expanding.
Investors appear to think that paying for the future now is worthwhile. Perhaps it is. However, the Arm and Snowflake debuts both made a big splash before going silent for a while, and seasoned LinkedIn users were already cautioning about this pattern. Whether Cerebras becomes the long-lasting second name in AI silicon or just the loudest IPO of a contentious year is still up in the air. Observing this develop, it is evident that the market’s appetite hasn’t peaked and that the company’s ability to grow into the number it was given before the market shifts its mind will be the next challenge.
