There's a company in San Jose selling $500 purple cables, and it just tripled its revenue to $1.34 billion in a single year. By some measures it's still early. That sentence should make you pay attention.

Credo Technology Group — ticker CRDO — is what happens when you find a niche so essential that no AI data center on earth can function without it, then you dominate that niche so completely that competitors are still figuring out how you did it. The company's active electrical cables — those famous purple cables — connect GPUs, switches, and servers inside hyperscale AI clusters. They're cheaper than fiber, use less power, and don't drop connections the way optical links do. Credo owns roughly 73% of the AEC market. When you need to move data between 100,000 GPUs without losing a single bit, you call Credo.

The numbers are genuinely absurd. Full-year FY2026 revenue hit $1.34 billion — up 206% year-over-year. The Q4 quarter alone, $437 million, surpassed the company's entire FY2025 revenue. Gross margins sit at 68%. Non-GAAP operating margin: 47.8%. The company is essentially debt-free with $1.4 billion in net cash. This is not a startup burning money to chase growth. This is a machine.

And yet the stock — trading around $268 as of this morning — sits roughly 13% below its 52-week high of $308.67. The pullback happened for no fundamental reason. No earnings miss, no guidance cut, no customer loss. Just the market deciding that a stock up roughly 86% year-to-date needed a breather. If you've been waiting for an entry point into Credo, this might be your window.

Three major analysts certainly think so. Between June 22 and June 23, Stifel raised its target to $350, Bank of America went to $340, and Evercore ISI initiated coverage with an Outperform and a $325 target. Three of the biggest names on the Street, all converging on the same thesis within 48 hours: Credo's connectivity story has legs well beyond copper cables. Stifel called it a "vertically integrated, system-level connectivity strategy" after multi-day meetings with management. BofA labeled Credo a key AI capex winner with a long runway through 2030. These aren't cautious nods. These are conviction calls.

The conviction comes from DustPhotonics. Credo closed its $1.3 billion acquisition of the Israeli silicon photonics company in May, and it changes the entire growth narrative. Before DustPhotonics, Credo was primarily a copper-cable company with some optical DSPs on the side. Now it has a fully vertically integrated optical stack: Credo's own SerDes IP feeding into Credo's own DSPs, sending signals through DustPhotonics' silicon photonic integrated circuits. The combined optical portfolio is expected to generate $600 million or more in FY2027 — three revenue legs (optical DSPs, silicon photonics PICs, ZeroFlap optics) each doing $100 million-plus individually.

CEO Bill Brennan told CNBC this AI buildout is "global in nature" — and the numbers back him up. Every hyperscaler — Amazon, Microsoft, xAI, Meta — is building clusters that need Credo's connectivity fabric. The AEC market alone is projected to hit $4 billion by 2028. And Credo's guide for FY2027 calls for 80%+ revenue growth. Not 20%. Not 30%. Eighty percent. On top of a year where revenue already tripled. Forward EPS estimates sit at $8.91, which puts the forward P/E at roughly 30x — expensive, sure, but a lot more reasonable than the 107x trailing multiple suggests.

There are risks — and you should know them. Customer concentration is extreme: the top two customers account for 71% of revenue combined, with the single biggest at 39%. Losing either would hurt badly. Insider selling totaling $76.9 million from the CEO, CFO, CTO, and legal chief in mid-June is worth noting — though execs at high-growth companies sell shares all the time, and Brennan still owns significant skin in the game. At 107x trailing earnings, the stock prices in years of perfection with zero room for execution missteps or a slowdown in AI capex.

But here's the thing about picks-and-shovels plays: they don't depend on which AI model wins. Doesn't matter if it's OpenAI, Anthropic, xAI, or Google DeepMind. Every training cluster needs connectivity. Every rack of GPUs needs cables. Every data center needs retimers and DSPs to keep signals clean at 1.6T speeds. Credo is the plumbing underneath the entire AI revolution — and if FY2027 guidance is anywhere close to accurate, it's plumbing that's about to get a whole lot bigger.

Disclosure: The Signal holds no position in CRDO. Positions may change. This is not financial advice.