Stay connected with KayaToday—follow us on Instagram and Facebook for the latest news and reviews delivered straight to you.
On Thursday morning, a ten-year-old chip company walked onto the Nasdaq floor and raised $5.55 billion in a single day. By the closing bell, Cerebras Systems was worth $95 billion — the largest initial public offering by a US technology company since Uber went public in 2019, at a time when the IPO market had all but seized up. The stock opened at $350, nearly double its offering price of $185, peaked at $386, and then drifted back to close at $311.07. By Friday it had fallen a further 10 per cent. The arc of those two trading sessions contains, in miniature, most of what needs to be understood about where the AI investment cycle currently stands.
Cerebras makes a chip unlike any other in commercial production. Its Wafer Scale Engine 3 is, quite literally, the size of an entire silicon wafer — 58 times larger than a standard GPU — and the company claims it can run AI inference workloads up to 15 times faster than Nvidia’s equivalent products. The architecture is genuine, the engineering ambition is real, and the company’s founders spent a decade bringing an idea that most of the semiconductor industry considered impractical to the point of commercial viability. That journey deserves recognition. It should not, however, be confused with the question of whether the valuation assigned to it on Thursday is sustainable.
The Numbers Behind the Number
Revenue at Cerebras jumped 76 per cent last year to $510 million, and the company generated net income of $88 million, swinging from a loss of $481.6 million a year earlier. Those figures represent genuine progress. But they require careful reading. The GAAP net income of $238 million includes a one-time $363 million non-cash gain; the real operational loss widened to $76 million. A company reporting an operating loss on $510 million in revenue, priced at a market capitalisation of $95 billion, is trading at a price-to-sales ratio of approximately 95, compared to Nvidia’s current ratio of around 25, and CoreWeave’s of 15 at its own recent IPO. The market is, in other words, pricing Cerebras as though it will achieve $3 billion to $4 billion in revenue within three to four years and sustain profitability throughout. That is a demanding projection for any company. It is a particularly demanding one for a company whose customer base remains heavily concentrated.
The Mohamed bin Zayed University of Artificial Intelligence in the UAE accounted for 62 per cent of Cerebras’s revenue in 2025, with G42 contributing a further 24 per cent. The company’s prospectus describes that concentration plainly. What changed the investment narrative was not the diversification of customers but the arrival of a single new anchor: OpenAI struck a $20 billion multi-year deal with Cerebras in early 2026 for computing capacity and related services. That contract, combined with an AWS deployment agreement announced in March, gave the IPO roadshow the storyline it needed. Whether those relationships translate into the revenue trajectory the valuation demands is now the central question for every institutional investor holding the stock.
What This IPO Is Really About
The Cerebras debut matters beyond the company itself for two reasons. First, it is the opening act of what analysts expect to be the most consequential AI IPO cycle since the dot-com era. Elon Musk’s SpaceX, which merged with AI company xAI in February, is gearing up for a share sale, and model developers OpenAI and Anthropic could hit the market later this year — all three valued at or near trillion-dollar levels in private markets. The reception that Cerebras received this week — 20 times oversubscribed during the roadshow, priced above its marketed range, and surging 68 per cent on debut — will calibrate the ambition and pricing of every offering that follows. In that sense, Thursday’s session was as much a market-making event as a company-making one.
Second, the IPO illuminates the structural dynamics of the AI chip market with unusual clarity. Nvidia’s Vera Rubin architecture boasts 336 billion transistors and claims a fivefold performance leap over Blackwell; AMD’s MI400 has already caught up to 320 billion transistors; Google’s TPU v6, Amazon’s Trainium 3, and Microsoft’s Maia 2 are all being developed by mass-scale manufacturers. Against that competitive backdrop, Cerebras is not positioning itself as a replacement for Nvidia but as a specialist in the inference layer — the part of the AI stack that handles live interactions with users rather than initial model training. That is a real and growing market. It is also a narrower one than the general framing of “AI chip company” implies, and the valuation attached to it on Thursday assigned very little discount to that distinction.
A Venture Capital Reckoning
For Silicon Valley’s venture community, Thursday represented something more personal: a long-awaited validation of patience. Benchmark, which co-led Cerebras’s Series A funding round in 2016, is sitting on shares worth $5.5 billion at Thursday’s close. That is a meaningful return on a decade-long bet. But the broader context is sobering. There were only 31 tech IPOs in 2025, down from 121 four years earlier. The IPO market has been effectively closed for most of the period since 2022, starving venture funds of the exits they need to return capital to their limited partners and raise new funds. Cerebras’s debut will relieve some of that pressure and may encourage other late-stage companies to accelerate their listing timelines. It will not resolve the underlying question of whether the AI infrastructure market can sustain valuations built on growth projections that stretch to 2028 and beyond.
That question will be answered not on the trading floor but in the quarterly revenue disclosures of the companies now shouldering those expectations. For Cerebras, the first report as a public company will arrive before the summer is out. Wall Street, for once, will be paying very close attention to a chip company whose most important product is not yet its largest source of revenue.
Read Also: The Semiconductor Cold War Has Arrived — And Every Nation Must Choose a Side
