Here's a number that should stop you mid-scroll: Credo Technology's return on invested capital sat at 14 percent a year ago. On Simon's own screen it now reads 85.9 percent (other trackers land lower, 46 to 73 percent depending on how invested capital gets defined, still elite either way). That's not a rounding error. That's a business that changed shape in twelve months.

Here's the number that should make you check position size before you check the chart: three customers still explain roughly 88 percent of Credo's revenue, and the AI-interconnect category Credo dominates is one technology generation away from a format it doesn't fully own yet.

Neither fact cancels the other out. A stock already up 70 percent this year, riding the exact AI-infrastructure wave every headline is chasing, can be a genuinely improving business and a genuinely risky entry point in the same sentence. The question isn't whether Credo makes something the industry needs right now. It obviously does. The question is whether 86 percent ROIC is a moat compounding in real time, or a temporary window before either a customer or a technology shift closes it.

What's actually driving the number

Pull the trend back and the acceleration is real, not a one-quarter fluke. Fiscal 2026 revenue hit $1.3 billion, up 206 percent year over year. The fourth quarter alone did $437.0 million, up 7.4 percent quarter over quarter and 157.0 percent year over year, with GAAP gross margin at 68.2 percent and GAAP net income of $169.1 million. Guidance for the current quarter sits at $465 to $475 million, non-GAAP gross margin 67 to 69 percent.

That combination, tripling revenue while gross margin holds near 68 percent, is what turns a 14 percent ROIC into an 86 percent one. Most companies get growth or margin expansion when they scale this fast. Credo is getting both, and the capital base barely has to grow to support it: capex was $4.8 million in the fourth quarter against $182.2 million of operating cash flow. This is a fabless, IP-licensing-shaped business. It doesn't need factories to grow. It needs design wins.

The design win it already has: Active Electrical Cables, the copper cabling that connects GPUs to switches inside AI clusters. Picture a physical cable, thick as a garden hose, running rack to rack inside a hyperscale data center. Engineers in the industry just call it the purple cable, priced around $500 a unit, and Credo pioneered the category. It holds roughly 73 percent share of it today. AECs use up to 50 percent less power than optical transceivers doing the same job. In a data center where power is the actual bottleneck, not chip supply, that's not a marginal edge. In March 2026, Credo settled patent infringement suits against TE Connectivity, Molex, Amphenol, and Volex. Settling, not losing outright, but settling still means those companies are now paying to compete rather than competing for free. That's a moat with a receipt attached.

Is the moat widening, or is the format about to change under it

TD Cowen is positive into the upcoming print, with a $300 price target and a model that has Credo's networking revenue compounding roughly 62 percent a year out to 2030. Their case rests on diversification: Credo moving beyond AECs into optical DSPs, ZeroFlap optical transceivers, OmniConnect-based gearboxes, and Active LED Cables. Street models call for only about 10 percent quarter over quarter growth in the October quarter. TD Cowen thinks that bar is beatable.

Read that diversification list again, though, and notice what it actually says. Credo is racing into optical because AECs have a physical range limit, and the next generation of AI clusters is moving toward interconnect speeds where copper starts to lose to optical on distance. The 1.6T optical transition is the single biggest threat to the category Credo currently owns 73 percent of. TD Cowen's growth thesis and Credo's biggest structural risk are the same sentence, told from two different angles. Say both sentences out loud and they rhyme uncomfortably. Winning the optical socket the way Credo won the AEC socket is the whole ballgame from here, and it hasn't happened yet. It's a bet on the same team that won the last format war, not a repeat win already banked.

Competitively, Credo is still the smaller player at the table. Broadcom, Marvell, and Astera Labs all sit in the same interconnect stack with materially more balance sheet and R&D scale. Credo's edge so far has been focus and being first. That's real. It's also the kind of edge that erodes the moment a bigger competitor decides the category is worth their attention. Optical is exactly the kind of higher-complexity category where scale starts to matter more.

The concentration question that hasn't gone away, only softened

A year ago this was a one-customer story. Amazon alone accounted for 67 percent of fiscal 2025 revenue and 61 percent of the fourth quarter of that year. That's the kind of number that ends a growth story overnight if the one customer changes its mind.

It has genuinely diversified since. By the third quarter of fiscal 2026, the top three customers stood at 39 percent, 32 percent, and 17 percent of revenue. Three hyperscalers each clearing double digits, instead of one clearing two-thirds. Progress, real progress. But do the arithmetic on those three numbers together: 88 percent of Credo's revenue still sits with three buyers, and the top 10 customers combined make up roughly 90 percent of the fiscal 2026 total. This isn't a diversified customer base yet. It's a slightly-less-concentrated one, three phone calls away from a very different quarter instead of one.

There's something unsettled in the air any time 88 percent of a company's revenue depends on three phone calls going the right way. None of those three customers has said anything today that threatens Credo's business. But every one of them, being a hyperscaler with its own chip and interconnect ambitions, is the exact kind of customer that eventually builds in-house once volume justifies the R&D. That's not a Credo-specific risk. It's the standing risk of selling critical infrastructure to companies with more engineering resources than you have.

Five scenarios, roughly four years out to 2030

Super Bull (10 percent): the optical transition goes Credo's way, AEC share holds, revenue compounds 65 percent a year, the market re-rates the stock to 25 times sales on durable-moat conviction. Implied price: roughly $1,547, more than 6x from here.

Bull (25 percent): TD Cowen's own model plays out close to as written, revenue compounds 62 percent a year, the multiple settles to 18 times sales as growth is still exceptional but no longer a total surprise. Implied price: roughly $1,030, up about 320 percent.

Base case (30 percent): growth decelerates to 40 percent a year as the 1.6T optical shift takes real share from AECs even as Credo wins some of that optical business itself, and the multiple compresses to 12 times sales, still a premium but a normal one. Implied price: roughly $366, up about 50 percent, or roughly 10 percent a year.

Bear (25 percent): one of the three concentrated customers meaningfully cuts orders or brings interconnect design in-house, growth slows to 20 percent a year, the multiple compresses hard to 8 times sales on the concentration scare. Implied price: roughly $126, down about 48 percent from here.

Super Bear (10 percent): two things go wrong at once, a customer walks and the optical transition genuinely displaces AEC volume faster than Credo's own optical ramp can replace it. Growth stalls, the multiple craters to 5 times sales. Implied price: roughly $36, down about 85 percent, real capital impairment.

Blend those five and the probability-weighted price lands near $557 by 2030, an expected return of roughly 21 percent a year. That's a genuinely good number. It rests on a 30 percent-weighted base case where growth still decelerates by more than half from today's rate. It also carries a 35 percent combined weight on outcomes where the concentration risk or the technology-transition risk actually bites.

The scorecard

Same ten questions, every stock, so the answer is a score, not a feeling.

Essential or nice-to-have: 8/10. AI clusters do not scale without high-speed interconnect, though the specific format (copper today, more optical tomorrow) is in flux.

Current moat: 8/10. 73 percent AEC share, foundational patents just validated in court via the March 2026 settlements, a real and measurable power advantage.

Moat expanding: 5/10. The diversification into optical DSPs and Active LED Cables is the correct move, but it's Credo defending against its own category's obsolescence, not yet a proven second win.

Balance sheet strength: 9/10. $1.4 billion in cash and short-term investments, $177.5 million of free cash flow in a single quarter, capex under $5 million a quarter. This is an asset-light, high-margin model.

EPS and earnings accelerating: 9/10. Revenue up 206 percent for the fiscal year, GAAP net income of $169.1 million in the fourth quarter alone.

Net margins increasing: 9/10. Gross margin above 68 percent, non-GAAP net margin north of 50 percent.

ROIC profile: 7/10. Genuinely elite on every source checked, ranging from the high 40s to 86 percent depending on how invested capital is defined for a business this asset-light. The trend, 14 percent to 86 percent in a year, is the real story. The wide range across sources costs it a point.

Reinvestment rate: 6/10. Different shape than a capital-heavy business. Reinvestment here is R&D and opex, not factories, so it doesn't show up as a rising capex line the way it would elsewhere. Guided opex is climbing (86 to 90 million next quarter) as the optical push ramps.

Capital return: 1/10. No dividend, no buyback. Every dollar stays in the business, which is correct at this stage but scores as zero return of capital today.

Valuation: 2/10. $46.6 billion market cap on $1.34 billion of trailing revenue, a price-to-sales ratio near 34 times, the kind of multiple that leaves a rich, hard-to-digest aftertaste even for buyers who believe the growth story. Among the most expensively priced large-cap semiconductor names trading today.

Overall score: 64/100. A real business with a real moat, priced for a lot of that moat to keep expanding cleanly, carried by three customers who could each independently change the story.

What would change the answer

Weekly RSI breaking below 30 would be the technical signal a real pullback has arrived, the kind that usually accompanies a concentration scare or a guide-down tied to the optical transition. Today it sits neutral at 58.4, daily RSI at 51.7. No entry signal exists right now under the house rule, regardless of how the fundamentals read. There's nothing to decide today, only a number to watch. This one goes back on the desk the moment weekly RSI breaks 30, same as every other name on the list, so watching that one number is most of the strategy.

Genuine further customer diversification, the top three sliding from 88 percent combined toward something closer to 60 or 65 percent, would resolve the single biggest unknown in this thesis without anyone needing to guess at it.

A confirmed optical design win at real scale, not just a product announcement, would answer whether Credo's moat is a franchise or a first-mover head start with an expiration date.

Not investment advice

This is not investment advice, just Simon's own research process shared publicly. Nothing here is a recommendation to buy or sell any security. Do your own research, size positions to your own risk tolerance, and talk to a licensed advisor before making decisions with real money.

Most coverage on a stock like this stops at the ROIC headline and the price target. You got the concentration math, the five-year scenario spread, and the entry rule that says no, not yet, even with fundamentals this strong. That's the whole method, the same four checks run before any name reaches this list, Credo included. If you want the full process, start free at https://www.strategysprints.com.

Ready to accelerate now? Book a Discovery Call: https://calendly.com/strategysprint/discovery-call

Happy hunting. Simon & The Sprinters 🐬⚡️🐆