What happens when a single cable fails inside a 100,000-GPU training cluster?

A single link flap, a momentary loss of signal, could corrupt the entire training run. The cluster would have to reload from the last checkpoint. Millions of dollars in compute time would instantly be gone. This issue changed everything about how hyperscalers thought about connectivity inside their data centers.

That problem is what made Credo Technology Group (CRDO) a serious business and a compelling investment case. It also explains why the stock went from $29 to $213 in roughly two years, and why, after pulling back 45% from that peak to around $116 today, the question worth asking is whether the business thesis has changed or whether the market is simply exhausted from the run.

I think it's the latter. I'm initiating coverage with a Buy rating and a 12-month price target of $165, representing roughly 42% upside from current levels. This is a position for investors who understand that owning a dominant infrastructure supplier in the early innings of a multi-year capital cycle carries real risk, but also the kind of return profile that makes the risk worth bearing.

What Credo Actually Sells

Credo designs and sells the semiconductors that live inside active electrical cables (AECs) - the purple cables connecting GPUs to switches inside AI data centers. The company is fabless, meaning it designs the chips and outsources manufacturing, keeping capital requirements low and gross margins high.

Its flagship HiWire AEC product line integrates Credo's proprietary SerDes (serializer/deserializer) technology directly into the cable assembly, giving it active signal processing capability that passive copper cables lack entirely.

Credo's AECs deliver up to 1,000 times better reliability than optical alternatives while consuming roughly half the power. In a world where data center operators are rationing electricity and paying premium prices for it, that power profile matters as much as the reliability story.

Beyond AECs, Credo sells optical PAM4 digital signal processors (DSPs) and PCIe retimers - chips that handle signal integrity for optical modules and high-speed server interfaces. It also licenses its SerDes intellectual property to other semiconductor firms, providing a high-margin recurring revenue stream that reinforces the company's technical moat.

In Q3 FY2026, reported March 2, 2026, revenue hit $407 million up an absolutely astounding 201.5% year over year - with product revenue dominated almost entirely by AECs.

Three Reasons to Own This Stock Right Now

The first reason is structural. Credo has what amounts to a de facto standard position in AEC connectivity at the exact moment that standard is becoming universally mandated.

Hyperscalers building out AI clusters at 100,000 GPU scale and beyond don't have a realistic alternative that combines Credo's reliability profile with its power efficiency.

The active electrical cable market is projected to reach $4 billion by 2028, and Credo commands approximately 88% of that market today. When a standard gets entrenched in infrastructure, it tends to stay entrenched. The switching costs are engineering time, recertification, and the operational risk of introducing new failure modes into mission-critical systems. None of those are trivial.

The second reason is the customer diversification story, which the market is underappreciating. Twelve months ago, a single hyperscaler believed to be Amazon Web Services, represented the majority of Credo's revenue. That concentration was a legitimate concern. By Q3 FY2026, four hyperscalers each contributed more than 10% of total revenue, and a fifth began contributing initial revenue. The largest single customer's share has been declining as new customers ramp. This is the business maturing in real time, and the stock price has not caught up to the structural improvement.

The third reason is the financial model itself.

Credo's Q3 gross margin came in at 68.5% on $407 million in revenue. Non-GAAP EPS of $1.07 beat the Street's $0.94 estimate by 14%. The company holds $1.3 billion in cash with near-zero debt.

Q4 guidance of $425-435 million implies continued sequential growth into the fiscal year-end. And Q4 FY2026 guidance came in above consensus when it was issued. This is a business operating with real leverage, that has revenue doubling while gross margins hold and net income accelerates faster than the top line.

Looking further out, the product roadmap extends the runway.

Credo plans to sample its next-generation active LED cables (ALCs) to lead customers in FY2027, with revenue contribution beginning in FY2028. The acquisition of CoMira Solutions, announced in March 2026 on undisclosed terms, adds optical capabilities that round out the product portfolio heading into the 1.6 terabit per second connectivity transition.

Competition and the Capital Cycle

Credo's most direct comparable is Astera Labs (ALAB), which also sells connectivity semiconductors into AI data center infrastructure. Astera focuses more heavily on PCIe and CXL connectivity while Credo is concentrated in intra-cluster GPU-to-GPU connectivity through AECs. They are adjacent, and should not be considered head-to-head.

Traditional transceiver vendors like Coherent and Lumentum serve the optical module market that AECs partly displace, creating an indirect competitive dynamic. In-house development by hyperscalers is the longer-term risk worth watching; Amazon and Google both have semiconductor design operations capable of attempting to replicate this technology, though the engineering and certification timeline to match Credo's reliability record would be measured in years, not quarters.

The macro context is supportive in a specific way, at least for now.

Amazon, Microsoft, Meta, and Google have all announced increased data center capital expenditure in 2025 and 2026 regardless of the broader economic uncertainty created by rising energy costs and the current geopolitical environment. AI infrastructure spending has proven remarkably durable even as consumer-facing technology budgets tighten. If anything, the Strait of Hormuz situation and associated energy price volatility makes Credo's power-efficiency story more compelling, not less.

Cheap, Fair, or Expensive?

At $116.50, Credo trades at roughly 65x trailing GAAP earnings and approximately 27x forward revenue on consensus FY2027 estimates, which themselves call for growth moderation after the explosive FY2026.

Those multiples look demanding until you compare them to the underlying business trajectory. The forward earnings multiple compresses significantly as the analyst community projects FY2027 EPS growth to continue even as revenue growth decelerates from triple digits to something more sustainable.

The 12-month median analyst price target sits around $176-$187 across 14-16 analysts, implying 50% upside from current levels. That consensus includes recent target reductions from BofA (to $160) and Susquehanna (to $170) following the Q3 results, so it is not stale optimism.

The stock's 52-week range of $29.09 to $213.80 tells you that this is a name that prices in scenarios aggressively in both directions. The current price near $116 reflects genuine concern about customer concentration, growth deceleration in FY2027, and the broader sell-off in AI infrastructure names since February.

I find the valuation reasonable relative to the competitive position and financial model, not a screaming bargain but not overextended either.

What Could Go Wrong

Customer concentration remains the central risk even as it improves. Two customers, believed to be AWS and Microsoft, still represent the majority of revenue. A single hyperscaler decision to pause its AI cluster build-out, reduce AEC orders to work down inventory, or attempt to develop alternative connectivity solutions in-house could crater Credo's revenue in a single quarter. This is not a theoretical risk; it is the precise scenario that the market is pricing when the stock trades at a 45% discount to its peak.

The 1.6T transition creates both opportunity and execution risk. As data center speeds move from 800G to 1.6T per port, every component in the signal chain needs to be re-engineered and re-certified. Credo must maintain its reliability and power-efficiency advantages at the new speed standard, while competitors have a natural re-entry point whenever the technology resets. Missing the 1.6T qualification at a key hyperscaler would be a meaningful setback.

Insider selling deserves acknowledgment. Credo's insiders, led by CTO Chi Fung Cheng, have executed 365 open-market sales over the past six months with zero purchases. That is a pattern worth monitoring, though it is common at high-growth companies where equity compensation is the primary form of long-term compensation.

Finally, FY2027 growth deceleration, even if the business remains healthy, could weigh on the stock if the deceleration is sharper than expected.

The company's own guidance language of 'mid-single-digit sequential growth' heading into the fiscal year-end implies that the era of 200% YoY comparisons is transitioning. How the market reprices this during the transition is genuinely uncertain.

My Take

I'm buying CRDO in the $110-$125 range. The business has earned its competitive position. You don't reach 88% market share in a mission-critical application without genuinely superior engineering. The stock has pulled back enough that you're no longer paying for a perfect execution scenario, but you are still paying for a company expected to grow earnings substantially over the next two years.

My 12-month price target is $165, based on roughly 35x FY2027 consensus earnings estimates. That multiple assumes the business demonstrates further customer diversification and executes on the 1.6T product cycle without a meaningful miss. If those conditions hold, the target is achievable.

If customer concentration worsens or a major hyperscaler signals reduced AEC spending, I would reassess at around $90, which represents the next meaningful support level and where the FY2027 earnings multiple compresses to the low 20s.

What would change the thesis entirely is a formal announcement from a major hyperscaler that it is developing an in-house AEC alternative, or a consecutive quarter of gross margin deterioration below 60%. Neither appears imminent. Until one of those events materializes, the asymmetry here favors ownership.

Time will tell, and we will continue to write updates on this investment case.

The Earnout Investor provides analysis and research but DOES NOT provide individual financial advice. Jamie Dejter may have a position in some of the stocks mentioned. All content is for informational purposes only. The Earnout Investor is not a registered investment, legal, or tax advisor, or a broker/dealer. Trading any asset involves risk and could result in significant capital losses. Please, do your own research before acquiring stocks.