CRDO Might Be the Most Important AI Company You’ve Never Heard Of — Until Today

Credo Technology just delivered the kind of AI-infrastructure print that forces everyone to revisit their “maybe the AI trade is tired” narrative. Shares of CRDO are up roughly 25% in pre-market trading, pushing to fresh all-time highs and trying to clear the $200 psychological level after the company posted record results and a powerful guidance raise. Coming alongside MongoDB’s blowout quarter, Credo’s move is doubly important: it’s not just another AI software story, it’s a direct read on the plumbing that lets those massive GPU clusters actually talk to each other. If this kind of connectivity name can hold its gains, it goes a long way toward stabilizing sentiment across the broader AI complex.

At a high level, Credo sits in the middle of one of the most critical – and capacity-constrained – parts of the AI stack: high-speed, energy-efficient connectivity inside hyperscale data centers. Its Active Electrical Cables (AECs), optical DSPs, and PHYs are used to wire up AI training and inference clusters, particularly in environments where traditional optics are too expensive or power-hungry. Management is now layering on three new growth pillars – ZeroFlap optics, active LED cables (ALCs), and OmniConnect gearboxes – that extend Credo from “just” AECs into a broader connectivity platform. For investors trying to gauge whether AI infrastructure demand is real, durable, and broad-based, Credo’s numbers are essentially a live traffic report.

The Q2 fiscal 2026 print was about as clean as it gets. Revenue came in at $268 million, up 20% sequentially and a staggering 272% year over year, versus consensus around $235–247 million. Adjusted EPS of $0.67 crushed expectations in the high-$0.40s to low-$0.50s, with non-GAAP gross margin at 67.7%, above the high end of guidance. Non-GAAP operating income reached roughly $124 million, with strong operating leverage as revenue grew more than 20% sequentially while operating expenses were held to mid-single-digit growth. Cash flow followed suit: operating cash flow was about $62 million and free cash flow roughly $38 million, leaving Credo with more than $800 million in cash and short-term investments on the balance sheet.

The demand signal under the hood is very clearly AI-driven. Management highlighted that four hyperscalers each contributed more than 10% of total revenue, with a fifth now entering initial volume production – all heavily tied to the build-out of AI training and inference clusters. AEC remains the fastest-growing segment in the company, displacing optical rack-to-rack connections in many high-bandwidth environments and ramping across multiple lane speeds (25G, 50G, 100G per lane) with an eye toward 200G and PCIe Gen 6. That said, this is not a single-product story anymore: optical DSPs and line-card PHYs are now contributing meaningfully, and IP monetization – as evidenced by the licensing agreement with The Siemon Company for AEC patents – adds another lever.

Guidance and the multi-year outlook are where the AI read-through really tightens. For Q3, Credo guided revenue to $335–345 million, implying roughly 27% sequential growth at the midpoint and far above prior consensus near $247 million. Non-GAAP gross margin is expected in the 64–66% range, with non-GAAP operating expenses of $68–72 million, preserving very healthy incremental margins. Management now expects fiscal 2026 revenue to grow more than 170% year over year to around $1.19 billion, with a non-GAAP net margin of roughly 45% and net income more than quadrupling versus last year. That’s not “AI is slowing” language; that’s “we’re still in the early innings of rack-level AI infrastructure proliferation” language.

Strategically, the story is expanding. Credo now describes five distinct high-growth connectivity pillars: AECs, ICs, ZeroFlap optics, ALCs, and OmniConnect gearbox solutions. Taken together, management believes these products give the company a total addressable market in excess of $10 billion by the end of the decade – more than triple the TAM they were talking about 18 months ago. Street analysts are baking that into their models: price targets have been raised aggressively, with some now assuming mid-$1 billion revenue in FY26 and longer-term EPS power north of $10 if Credo can capture around 50% share in its key segments and sustain mid-60s gross margins.

On the operating environment, CEO Bill Brennan was explicit: these are the strongest quarterly results in Credo’s history, and they reflect the continued build-out of the world’s largest AI training and inference clusters, not a one-off order spike. He underscored diversification – each hyperscaler is effectively its own market, with different ramp profiles and protocols – and pointed out that the company is deliberately manufacturing on older geometry processes to mitigate wafer supply constraints. In Q&A, management leaned into the idea that networking is becoming the key bottleneck in AI systems, and that system-level solutions like ZeroFlap optics and OmniConnect gearboxes give Credo more control over performance, reliability, and margins than a pure component supplier would enjoy.

There are, of course, risks lurking under the glossy growth. Customer concentration is still very real – four hyperscalers account for the bulk of revenue, and ramp patterns are inherently nonlinear. Wafer supply could tighten across the industry as AI demand spreads, and execution risk rises as Credo pushes simultaneously into AEC, optics, ALCs, and gearboxes. Analysts on the call probed these points repeatedly, focusing on sustainability of the growth profile, timing of new product ramps, and the ability to maintain 63–65% gross margins as mix shifts.

But for now, the takeaway is straightforward: Credo just delivered a textbook AI infrastructure beat-and-raise, with numbers and guidance that are entirely inconsistent with the idea that AI capex is rolling over. With the stock up about 25%, pushing above the $200 psychological level and holding the move so far, CRDO – alongside MDB – becomes a key tells for whether investors are ready to re-embrace high-quality AI spend beneficiaries after the recent malaise. If these gains stick through the session and the week, the market will be forced to admit that, at least in the data-center plumbing, the AI cycle is still very much alive.