Cerebras Systems just had the kind of IPO debut that makes venture capitalists and retail traders lose their minds. The chip startup priced shares at $185 on Wednesday evening, then watched them open for public trading at $385 the next morning. That’s a 108% pop in a single session, and by mid-day, shares were still hanging around $330.
A company that couldn’t go public a year ago just entered the market at a fully-diluted valuation of $56.4 billion. Co-founder CEO Andrew Feldman’s personal stake is now worth nearly $1.9 billion. Co-founder and CTO Sean Lie’s holdings crossed the $1 billion mark. This is what happens when the market decides you’ve cracked something important.
But here’s what makes this story interesting: it’s not really about a spectacular product reveal or a sudden earnings miracle. It’s about how quickly sentiment shifts when geopolitical obstacles disappear and fundamentals improve enough to matter.
When Foreign Investment Becomes a Regulatory Minefield
A year ago, Cerebras looked dead in the water. The company had filed to go public in 2024, but the Committee on Foreign Investment in the United States (CFIUS) threw the process into endless review purgatory. The culprit was a major investment from Abu Dhabi-based Group 42, which spooked regulators worried about foreign capital flowing into sensitive technology infrastructure.
The timing was particularly brutal because Group 42 wasn’t just an investor. It was also Cerebras’s largest customer, accounting for almost all of the company’s revenues. For potential public market investors, that’s a massive red flag. You’re not looking at a diversified business with multiple revenue streams. You’re looking at a dependency relationship. Throw CFIUS concerns into that mix and institutional investors ran for the exits.
So the IPO plans got shelved. Quietly. The kind of shelved where nobody talks about it anymore because it’s too embarrassing.
The Turnaround Nobody Expected
Then something shifted. By April, Cerebras was ready to try again, and this time the numbers told a radically different story.
The company reported roughly double its revenues: $510 million in 2025, a 76% year-over-year jump. More importantly, it had diversified its customer base. Group 42 was still there, but now OpenAI, Amazon Web Services, and Saudi Arabia’s Mohamed bin Zayed University of Artificial Intelligence were in the mix too. The dependency problem wasn’t solved, but it was diluted enough to pass scrutiny.
The profitability swing was even more dramatic. Cerebras swung from a $500 million net loss in 2024 to a $237.8 million net profit in 2025. That’s the kind of reversal that changes how people price risk.
The CFIUS blockade also seems to have loosened. Whether that’s because regulations evolved, political winds shifted, or Cerebras simply restructured its foreign relationships isn’t entirely clear. But the regulatory fog lifted enough for this IPO to actually happen.
What Cerebras Actually Does
Let’s be clear about what Cerebras sells. It’s not general-purpose chips. Cerebras designed its processor from scratch, purpose-built for AI inference. That’s the computational grunt work that happens after a model is trained, when it’s actually answering your prompts.
Inference is where the money is starting to accumulate. Training is the expensive upfront cost. Inference is the ongoing operating cost that scales with every user interaction. As AI companies race to deploy models at scale, they’re suddenly very interested in chips optimized specifically for inference workloads. Cerebras positioned itself directly in that lane.
Whether the company can actually compete with Nvidia’s stranglehold on the AI chip market is a different question entirely. Nvidia isn’t sitting still, and it has decades of GPU architecture expertise plus an ecosystem advantage that’s nearly impossible to replicate. But the market’s willingness to value Cerebras at $56 billion suggests investors believe there’s room for multiple players in this space, at least in specific niches.
The Retail Investor Euphoria Problem
Here’s the part that deserves some skepticism. Cerebras opened at $385 and settled around $330 mid-day. That’s a gorgeous debut, the kind of performance that makes headlines and gets retail traders hyped. But it’s also the kind of performance that raises obvious questions about whether the IPO was priced conservatively enough to leave money on the table or whether the stock is overheated going into its first full week of trading.
The company increased the offering size to 30 million shares to capture more of that demand. Smart move for existing shareholders. Good luck for anyone buying at $330 hoping for another 108% run.
Valuations in AI infrastructure are notoriously hard to pin down right now. Cerebras could be a legitimate competitor to Nvidia in a specific inference niche, or it could be a $56 billion bet on a market that compresses and consolidates over the next few years. The IPO performance tells you what retail investors think today, not what the company will be worth once the dust settles.
The fact that a company with genuine geopolitical complications managed to go public at a massive valuation suggests one thing clearly: AI chip ambitions are serious enough now that investors are willing to look past almost anything. That’s either confidence in an exploding market or a collective fever dream. Time will tell which one.
