Cerebras Unlocks 171 Million Shares. Math Says Run.

Cerebras went public on May 14th at $185 per share, raising $6.4 billion on the back of 34.5 million shares. The company makes AI chips. Investors showed up. Then they did the math on the lockup schedule and the colour drained from the room.

Here is the problem distilled to its essence: roughly 171 million shares—five times the size of the IPO float itself—become eligible to trade over the next four months. Not all at once, which would be merciful. In waves. Waves that start in September and keep arriving until November 9th, when the lockup expires completely.

This is not conjecture. This is the prospectus. The staggered release begins with 14.6 million shares becoming eligible every two weeks starting in September. Then three more tranches land between September 30th and October 28th—about 19.4 million shares each time. By August, up to 84 million Class B shares convert and become sellable, which is almost three times the company's entire current public float. Let that sink in. The supply about to hit the market is roughly the size of three times what exists today.

To understand what this means, you need to understand the denominator. SpaceX faced a 319-million-share release in 2023. That sounds enormous until you realise SpaceX's total float is roughly 13 billion shares. The relative dilution was noise. Cerebras's 84 million share release represents a fundamentally different creature. It is a sledgehammer aimed at a smaller target.

The timing compounds the problem. Retail investors and momentum traders have been operating in a low-supply environment since the May IPO. Every bid has moved the price because there simply were not enough shares to sell against rising demand. That dynamic ends in September. The market will discover what happens when supply finally meets an abundance of shares seeking exit velocity.

Valuation adds another layer of concern. Cerebras trades at roughly 87 times 2026 sales targets based on current growth projections. That is not a number you want to be holding when a five-fold supply shock arrives and sentiment shifts from scarcity to abundance. The stock is priced for perfection and execution at scale.

There is, admittedly, a backlog. Cerebras claims to have up to $24.6 billion in customer commitments. That number should matter. It probably will matter. But here is the wrinkle: roughly 80 percent of that backlog depends on OpenAI. One customer. One concentration risk. One reason that backlog, however large, may not prove as sticky as marketing materials suggest.

The staggered unlock schedule exists for a reason. It avoids the catastrophic single-cliff event that can crater a stock overnight. Wise founders and their lawyers learned long ago that you do not release 171 million shares on the same Tuesday. You spread them out. You hope the market absorbs them incrementally. You hope early waves do not spook holders for later waves.

But hope is not a strategy. And the fundamental math remains unchanged: there are 171 million shares about to become available against a public float of 34.5 million. The ratio is 5 to 1. In the real world, that is called dilution. In the trading world, it is called an opportunity to short. In retail investor forums, it will soon be called a warning flag that arrived too late.

Cerebras has genuine technology. It has customer interest. It has a valuation that prices in most of the upside already. What it does not have, starting in September, is supply scarcity. That is not conjecture. That is the calendar. And the calendar, unlike earnings guidance, never lies.