Cerebras can sell speed. Public investors still have to buy customer mix.

That was the paper's June 2 point when it said Cerebras still came down to customer concentration. Thursday's follow-up sharpens the question. The problem is not whether the company's wafer-scale systems are technically impressive. It is whether OpenAI, AWS, and other relationships diversify the business or merely swap one concentration risk for another.

The Motley Fool's analysis of the IPO laid out the uncomfortable arithmetic. It described the offering range, valuation, revenue, a price-to-sales figure near 95 times, and the fact that 86 percent of revenue came from two UAE customers. It also noted the OpenAI caveat: a relationship can excite the market without immediately solving the concentration table. [1]

Cerebras's own homepage supplies the opposite tone. It showcases IPO links, OpenAI and AWS references, customer testimonials, and performance claims. Those materials are useful for understanding the sales pitch. They are not a filing-quality revenue table. [2]

This is where the X/MSM split helps rather than distracts. X can turn Cerebras into either a clean AI-chip winner or a clean short. Mainstream finance can write the familiar IPO story: fast-growing company, famous customers, large market, expensive stock. The harder reading is less glamorous. It asks whether a customer list has become a diversified revenue base.

There is a difference between having OpenAI in the story and having OpenAI solve the story. There is a difference between AWS as a validation partner and AWS as a revenue counterweight. There is also a difference between backlog talk and recognized revenue.

The market likes to price AI infrastructure as if all demand is equal. It is not. A dollar from one customer with unusual geography, financing, or strategic ties is not the same as a dollar from a broad base of ordinary buyers. Cerebras may grow through that risk. But the risk is the point.

If the next company document does not make customer concentration clearer, the IPO will remain an AI-chip narrative attached to a very old public-market problem.

-- THEO KAPLAN, San Francisco