The number is stark: $470 million in quarterly revenue, a year-over-year doubling that most financial media will file under "AI semiconductor momentum." That framing misses the structural signal. Credo Technology isn't just another chip vendor riding the GPU wave. It is the physical layer of a connectivity economy that determines how fast AI clusters — and increasingly, the compute infrastructure that crypto miners and AI-crypto convergence projects depend on — can actually talk to each other.
I've spent the past decade auditing protocol architectures, from Uniswap V2's constant product formula to the liquidity mechanics of DeFi lending markets. The same analytical lens applies here. When a company's revenue doubles in a single quarter, the market wants a narrative. I want the mechanism. And the mechanism behind Credo's growth reveals something about the AI infrastructure buildout that most crypto market participants are not yet pricing.
The Connectivity Bottleneck
Here is the context that matters. Every AI accelerator shipped by NVIDIA or AMD requires two to eight optical modules or active electrical cables to connect to the network fabric. The GPU is the compute engine, but the SerDes, the DSP, the retimer — these are the nervous system. Without them, a cluster of 10,000 GPUs is just a pile of silicon that cannot coordinate.
Credo operates in this connectivity layer. The company is fabless, designing high-speed mixed-signal chips that get manufactured at TSMC on 7nm and 5nm FinFET nodes. Its core intellectual property is self-developed high-speed SerDes IP — 10G through 200G per lane — which is the foundation for 800G and 1.6T optical modules and AI back-end networking.
The product portfolio splits into two engines. The first is Active Electrical Cables (AEC), a category Credo essentially created. AECs replace short-reach optical modules in rack-to-rack and intra-rack connections, using copper with active signal conditioning. The second is optical DSPs, which sit inside 800G and 1.6T optical transceivers. The Q1 earnings report — with optical communications revenue guidance exceeding $600 million for the full year — signals that the optical DSP engine is now firing at scale.
Reading the Dual-Engine Thesis
This is where the analysis gets interesting. The market narrative around Credo has been dominated by the optical DSP story — the idea that the company is taking share from Broadcom and Marvell in AI optical module design wins. The $600 million optical guidance feeds that narrative. But the math suggests something else.
Q1 revenue was $470 million. If optical communications is guided to exceed $600 million for the full year, and if that guidance is not linear — meaning Q1 optical revenue is likely already in the $150-200 million range — then the non-optical business (AEC, IP licensing, and other products) is contributing somewhere in the $270-320 million range per quarter. That implies AEC is a much larger revenue driver than the market narrative acknowledges.
The market is fixated on the optical DSP story because it is the sexier narrative — taking share from Broadcom in AI optical modules. But the AEC business is the quieter, more defensible moat. Credo holds roughly 50% or more of the AEC category as its creator and dominant player. Broadcom and Marvell do not meaningfully compete in AEC. This is a category where Credo has no direct large-scale competitor.
The Competitive Reality
Let me be direct about the competitive landscape. In pure optical DSP, Credo is a third-place player. Broadcom holds roughly 50% of the optical module DSP market, Marvell around 30%, and Credo sits in the 10-15% range. The company is roughly half a generation behind Broadcom in pure DSP technology — Broadcom is shipping 200G/lane SerDes at scale while Credo is transitioning from 112G to 224G per lane.
But this misses the strategic picture. Credo's moat is not in raw process technology — it is in mixed-signal IP accumulation and system-level connectivity solutions. The barrier to entry in this space is 5-8 years of design experience plus extensive customer validation. A new entrant cannot simply hire engineers and replicate Credo's AEC position. The revenue doubling in Q1 suggests its 224G/lane or 800G DSP solutions have passed validation at major hyperscalers and are shipping in volume. That is not a trivial achievement.
The Supply Chain Question
From a supply chain perspective, Credo is a fabless company with moderate dependency on TSMC. The company does not own fabs, so traditional yield analysis does not apply. The risk is concentrated in a single foundry relationship — if TSMC's advanced process capacity becomes constrained or geopolitically compromised, Credo's high-end DSP production would face 6-12 month disruptions. The AEC supply chain is more diversified, relying on standard copper cabling and PCB components.
Geopolitically, Credo is relatively insulated. It is a US company selling primarily to US hyperscalers. Export controls on China have minimal direct impact. The indirect risk comes from China's gallium and germanium export restrictions, which could affect the laser chips used in optical modules upstream — a downstream demand risk for Credo's optical DSPs.
The Contrarian Angle: The Rug Pull Nobody Is Pricing
Here is where I diverge from the consensus. The market is pricing Credo at 20-30 times sales, a valuation that assumes the current growth trajectory extends indefinitely. That is a fragile assumption. Let me enumerate the structural fragilities.
First, customer concentration. Credo's top five customers likely account for over 60% of revenue, with the largest customer — probably Microsoft — representing 30% or more. In the AI connectivity space, hyperscalers have enormous bargaining power. If any single customer decides to develop in-house connectivity solutions or switches suppliers, the revenue impact would be immediate and severe. This is the classic rug pull scenario — not a malicious one, but a structural one. When hyperscaler capex cycles turn, as they did in 2022, the connectivity layer gets hit disproportionately hard.
Second, the competitive response. Broadcom and Marvell are not standing still. Both have the engineering resources and customer relationships to expand into AEC if they choose. Credo's AEC moat is real, but it is not unassailable. A determined Broadcom could compress Credo's pricing power within two to three years.
Third, the inventory cycle. The current AI connectivity market is in a high-demand, inventory-building phase. Hyperscalers are likely building two to three quarters of safety stock. When the AI capex cycle normalizes — and it will — the inventory correction will hit connectivity chip vendors hard. The 2022 data center chip inventory correction is the historical precedent.
The market narrative is fixated on the optical DSP growth story. But the real risk is the AEC business being taken for granted while the optical story gets overhyped. The $600 million optical guidance has created a narrative that the company is becoming an optical DSP powerhouse. The reality is that AEC remains the foundational business, and the optical business — while growing fast — is competing against much larger, better-capitalized incumbents.
The Macro Connection
From a macro perspective, Credo's performance is a confirmation signal for the AI infrastructure buildout thesis. The revenue doubling indicates that hyperscaler AI capex is translating into real, physical infrastructure deployment — not just GPU purchases, but the entire connectivity stack that makes those GPUs useful.
This has direct implications for the crypto market. The AI-crypto convergence narrative — where compute infrastructure becomes a shared resource between AI training and crypto mining — depends on this physical layer being built out. Credo's AEC and optical DSP products are the plumbing of that convergence. When I published my institutional convergence thesis in 2024, I argued that AI computing power markets would converge with crypto mining economics. Credo's earnings are evidence that this convergence is happening at the physical layer.
The Valuation Trap
Let me address the valuation question directly. At 20-30 times sales, Credo is priced for perfection. If FY27 revenue reaches $2 billion — which would require continued acceleration — a 15 times sales multiple implies a $30 billion market cap. The current market cap is likely already in that range. This means the market has already priced in the optical DSP growth story, the AEC moat, and the AI infrastructure buildout.
What is not priced in is the downside scenario: a hyperscaler capex pause, a competitive entry into AEC, or an inventory correction. Any of these would trigger a re-rating that could cut the stock by 50% or more. This is the rug pull that the market is not pricing — not a deliberate one, but a structural one embedded in the customer concentration and valuation assumptions.
Positioning for the Cycle
The question for investors is not whether Credo is a good company — it clearly is. The question is whether the current valuation adequately compensates for the structural risks. My assessment: the market is paying full price for the growth story while discounting the fragility of the customer concentration and the competitive response.
For crypto market participants, the takeaway is broader. The AI infrastructure buildout is real, and it is creating value across the physical layer — from connectivity chips to power infrastructure to data center construction. But the same dynamics that create this value also create fragility. The hyperscaler capex cycle will turn. When it does, the connectivity layer will be hit first and hardest.
The signal from Credo's Q1 earnings is not just about one company. It is about the physical reality of the AI buildout — and the structural fragilities that come with it. The market is pricing the growth. It is not pricing the rug pull. The question is whether you are positioned for the moment when the market is forced to confront the difference.
