Cerebras did not just give Andrew Feldman a paper fortune. Its Nasdaq debut showed how aggressively public investors are now pricing access to AI compute.
Cerebras Systems arrived on the Nasdaq with the kind of demand that tells you something bigger than one founder's net worth. The Sunnyvale AI chipmaker priced 30 million Class A shares at $185 on May 13, opened at $350 the next day, and then closed its offering on May 15 after underwriters exercised their full option, lifting gross proceeds to about $6.38 billion.
That is a serious amount of capital for any public debut. It is even more telling because Cerebras is not selling a consumer app, a workflow tool, or another software layer sitting on top of the AI boom. It is selling infrastructure. More specifically, it is selling the possibility that customers can get high-performance AI compute without being trapped inside the Nvidia supply chain.
Bloomberg reported that co-founder and CEO Andrew Feldman's stake was worth about $3.2 billion after the debut, which makes for a neat headline. But the richer story is the public market signal. Investors are now willing to value AI hardware scarcity in a way that looks much closer to software-style growth pricing than old semiconductor caution.
According to Reuters, Cerebras raised $5.55 billion in the base offering at a fully diluted valuation of about $56.43 billion, after investors placed orders for more than 20 times the shares available. That demand came after the company had already lifted its range from $115 to $125, then to $150 to $160, before finally pricing at $185.
The market then did what hot IPO markets do. Shares opened roughly 89% above the offer price, touched even higher levels during the first day, and closed Thursday at $311.07, still up more than 68% from the IPO price. By Friday morning, the stock was already giving back part of that move, which is normal for a deal where the opening trade ran far ahead of the allocation price.
For startup operators and investors, that volatility matters less than the reason the deal worked. AI companies are still hungry for chips, power, networking, and data center capacity. The software narrative is not gone, but the bottleneck has moved closer to the physical world. If you control compute, or can credibly offer it at scale, public markets are listening.
Cerebras has a cleaner story than many hardware companies because its product is easy to understand at the strategic level. Its Wafer-Scale Engine uses almost an entire silicon wafer as one processor, rather than cutting the wafer into many smaller chips. The company says its WSE-3 has 4 trillion transistors and 900,000 AI-optimized cores, with an inference platform that can run much faster than leading GPU-based systems on certain open-source models.
That does not mean Cerebras has solved the whole market. Nvidia's advantage is not just chip performance. It is software, developer habit, supply relationships, networking, and a customer base that already knows how to buy and deploy its systems. A real Nvidia alternative has to compete with an ecosystem, not only a processor.
The Valuation Comes With A Concentration Problem
Cerebras also carries the kind of risk that investors tend to forgive during a hot debut and remember later. Its revenue grew to $510 million in 2025 from $290.3 million a year earlier, and Nasdaq's newsroom noted that the company swung to $237.8 million in net income after a large prior-year loss. That profit figure needs context. The filing also showed a non-GAAP net loss after one-time items, so investors should not read the IPO story as proof that the core business is already comfortably profitable.
The concentration issue is just as important. A large share of Cerebras' business has depended on a small group of major customers and government-linked AI infrastructure projects. A company can look very different when a few buyers are purchasing large systems, especially if those purchases are tied to national AI ambitions, cloud capacity commitments, or multi-year infrastructure buildouts.
The newer customer story is more promising. Cerebras has pointed to major relationships with OpenAI and Amazon Web Services, both of which give the company more credibility than a pure challenger narrative would. These names matter because AI infrastructure buyers do not casually swap out the core systems behind model training and inference. If large customers are willing to test or deploy Cerebras at meaningful scale, other buyers will pay attention.
But big-name deals can cut both ways. They validate the technology, while also raising the bar for execution. Cerebras must deliver capacity, performance, reliability, and economics in an environment where customers are sophisticated and alternatives are improving. The IPO gave the company more capital to build. It did not remove the need to prove that demand can survive outside a narrow set of flagship accounts.
That is why the IPO should be read as a market test, not a victory lap. Public investors just showed that they will reward AI infrastructure companies when the scarcity is obvious and the customer names are credible. Now they will want evidence that the business can scale beyond a few headline relationships.
The next thing to watch is not whether Feldman's fortune moves up or down with the stock price. It is whether Cerebras can turn a spectacular debut into durable public-company numbers. If it can, the IPO market may open wider for AI infrastructure startups. If it cannot, this week will look less like a new rule and more like a reminder that scarcity can create demand faster than fundamentals can catch up.
Also read: Figure's robot livestream raises the bar for humanoid proof • YouTube makes deepfake protection a mainstream account feature • Cisco shows that record AI revenue does not protect jobs
