The Numbers Nobody Led With

Cerebras opened at $350 per share on May 14. That's 89% above the $185 IPO price, according to Reuters via BNN Bloomberg. By close, it settled at $311.07 — still a 68% single-day pop, per SSBCrack News.

Fully diluted valuation hit $106.75 billion at open. For context, only two tech companies in U.S. history have ever closed their IPO day above $100 billion: Alibaba and Facebook. Cerebras came close but didn't quite clear the bar, per CNBC.

The offering was oversubscribed more than 20 times, according to Reuters. David Yakobovitch, general partner at DataPower Capital, told SSBCrack News that level of demand surprised even experienced investors.

What Changed From Three Months Ago

Cerebras was valued at $23 billion just three months before the IPO, after closing a $1 billion funding round — SSBCrack News reported that figure. By IPO day, the valuation had more than doubled to $56.4 billion at pricing. Then the market pushed it past $95 billion on debut.

The company originally planned to price between $115 and $125 per share. It got bumped — twice — before landing at $185. This level of adjustment reflects a company with more demand than it anticipated.

The Near-Death Story Everyone Buried

Most mainstream outlets focused on the celebration. TechCrunch went deeper.

In 2019, Cerebras was burning $8 million per month and had already incinerated nearly $200 million trying to solve one engineering problem: packaging a wafer-scale chip that was 58 times larger than conventional chips and consuming 40 times more power than anything previously built.

CEO Andrew Feldman told TechCrunch directly: "We destroyed an enormous number of chips" — and an enormous amount of cash along with them. Every few weeks, Feldman had to walk into a board meeting and report another failure.

There were no vendors. No heat sinks built to spec. No manufacturing partners who'd done this before. The team had to invent a machine that could drive 40 screws simultaneously without cracking the wafer. The company had to solve hard physics problems through trial and error, backed by massive investor capital.

The chip the company eventually produced packs hundreds of thousands of compute cores onto a single processor the size of a dinner plate — a fundamentally different architecture than Nvidia's GPU clusters.

The Business Model Pivot Nobody Mentions

Alvaro Filpo, co-founder of Fabrica Ventures, told SSBCrack News something the celebratory coverage glossed over: Cerebras initially got the business model wrong. The company pushed expensive hardware sales. It didn't work.

The pivot to selling cloud access tokens for its chips changed the equation. That's the model that got OpenAI and Amazon in the door as customers — the same customers who replaced the concentration risk that almost killed the company's IPO attempt last year.

In 2024, more than 85% of Cerebras' revenue came from a single customer: G42, a UAE-based AI firm. That partnership triggered a national security review by the Committee on Foreign Investment in the United States. The committee eventually cleared it, but the damage to the first IPO attempt was done. The company had to pull its listing.

Now, with Amazon and OpenAI on the customer list, the concentration risk is reduced — though not eliminated.

The Crowding-Out Problem

Cerebras' success may actually hurt the rest of the IPO pipeline.

Sam Lessin, partner at Slow Ventures, told CNBC Thursday: "It's very hard to care about anything other than the $3 trillion potential IPOs that, in theory, are going to happen in the next year."

He's talking about SpaceX, OpenAI, and Anthropic. SpaceX is expected to file its IPO prospectus as soon as next week, per CNBC.

Nick Smith, senior equity research analyst at Renaissance Capital, told SSBCrack News that Cerebras timed its entry strategically — because once SpaceX and OpenAI hit the market, they will dwarf everything else and suck up every available investor dollar.

Smaller tech companies with legitimate businesses — not AI-branded ones — are getting crushed for attention. The venture ecosystem has been largely frozen since early 2022.

What Mainstream Coverage Got Wrong

CNBC and BNN Bloomberg focused heavily on the pop and the AI mania framing. But nobody emphasized the near-failure story — the $200 million in burned cash, the $8 million monthly burn rate, the years of engineering dead ends.

The reason this matters: the technology is defensible because of it. Nvidia can't replicate a wafer-scale chip overnight. The competitive advantage isn't a patent portfolio — it's a decade of brutal, expensive engineering lessons that competitors would have to repeat from scratch.

The valuation may or may not be justified at $95 billion. Coinpaper noted that analysts considered the $185 IPO price reasonable against 2028 projected financials — meaning the current market price prices in years of execution that hasn't happened yet.

What This Means for Investors

If you own AI-adjacent stocks, the Cerebras IPO signals the market still has appetite — but that appetite is increasingly concentrating at the very top. SpaceX and OpenAI IPOs will be capital events unlike anything the market has seen. Everything else will have to fight for scraps.

For retail investors watching CBRS trade at $311 on day one: remember it was a $23 billion company three months ago. The math on that move deserves more skepticism than the headlines gave it.