CRDO

Credo had an extraordinary year and the stock is down about 40% anyway. Revenue for the fiscal year that ended in May was $1.34 billion — more than triple the year before — and profit jumped to $472 million from about $52 million. The last quarter alone grew 157% from a year earlier. And unlike most of the AI hardware companies we've looked at, Credo keeps a lot of what it earns: about 68 cents of gross profit on every dollar of sales. That's the difference between designing something and assembling it. Management expects growth above 80% again this year. Meanwhile the stock fell from a high of $308 to around $177 in late July, dropping 22% in a single week. What Credo makes is the plumbing that moves data around inside a data center. When you fill a building with AI chips, those chips have to talk to each other constantly and at enormous speed, and the cables and chips that make that work reliably are a real engineering problem. Credo's signature product is a copper cable system that doesn't drop connections — a genuine headache-solver for anyone running thousands of chips at once. It's also moving into optical, meaning fiber-based connections, which is where the industry is headed for longer distances and higher speeds. Two risks are worth being clear-eyed about. Credo's ten largest customers account for roughly 90% of its sales, and two of them are each more than 10% on their own. If one of those customers switches to a competitor or decides to build the part themselves, it's an enormous hole. And insiders sold about $113 million of stock over the last three months with zero purchases — the same pattern we've now flagged at Celestica and Lam Research. As for the crash itself: nothing broke. This was the whole AI trade selling off at once, and Credo swings harder than most because it's small and crowded. If you own it, the number we'd watch is the optical business — management is targeting more than $600 million there this year. If optical works, Credo isn't just a copper cable company with two customers anymore. If it doesn't, that concentration problem is the whole story. Next report is September 9. (written with the help of Claude)