Earnings Call Highlights: Credo (CRDO) | Q4 FY26 - The Copper Cable Champion Turns "Reliability" into a $600M Optics Business

AEC is still making money, but Credo has already bet everything on the second half of the optics game.

Credo's Q4 FY26 earnings call had numbers so good they felt almost unreal: quarterly revenue of $437M, up 157% year over year - this single quarter exceeded all of FY25's annual revenue. But the real story isn't in the rear-view mirror; it's in the line management drew themselves - in FY27, optics goes from zero to a business of more than $600 million. For a company built on AEC copper cables, that's rewriting its identity.

1. Executive Takeaway: Three Signals to Remember First

First, growth hasn't slowed, but momentum is switching engines. FY26 revenue reached $1.3B, up 206%, with non-GAAP net income surging 5x to $662M. But Credo's FY27 guidance is "80%+ growth for the year", with half of the absolute dollar increase coming from optics - meaning copper's (AEC) high-base growth is handing the baton to optics.

Second, Credo has packaged the industry's biggest pain point as its most expensive product. Throughout the call, CEO Bill Brennan framed "network reliability" as the north star. The core selling point of the ZeroFlap family isn't bandwidth but "the cluster doesn't go down" - in an AI factory where an hour of downtime burns millions of dollars, that's something that commands a triple-digit ASP.

Third, this is guidance that "only takes off in the back half". The first half calls for only mid-single-digit sequential growth; the real hockey stick lands with the optics ramp in the second half. In other words, FY27 success hinges on whether ZeroFlap Optics ramps on schedule in Q3 and Q4 - the biggest execution risk in the report and the variable most worth tracking.

We already broke down how Credo set up this optics storyline last quarter in Earnings Call Highlights: Credo | FY26 Q3; this quarter is the continuation and delivery of that thread.

2. Financials: Still Accelerating Off a High Base, and Earning Even More

First, the quarterly numbers:

Q4 revenue of $437M, up 7% QoQ and 157% YoY, above the top end of guidance.

Non-GAAP gross margin of 68.3%, also above the top end of guidance.

Non-GAAP net income of $226.7M, up 9% QoQ and more than tripling YoY, with net margin reaching 51.9%.

Notably, this quarter's net income was 33% higher than the year-ago quarter's entire revenue - that one sentence best captures Credo's explosive year.

Operations were equally strong.

Q4 operating income was $216.7M with a 49.6% operating margin (flat QoQ); operating expenses were $81.7M (slightly above the top of guidance due to stepped-up R&D).

Cash flow was even stronger: quarterly operating cash flow of $182.2M and free cash flow of $177.5M (CapEx of only $4.8M), with cash at quarter end reaching $1.4 billion.

For the full year, FY26 revenue was $1.3B, up 206%, gross margin 68.1% (up 310 bps), and EPS $3.46, up 392%. Operating margin improved by 2,144 bps in a year - this isn't just revenue doubling but "scale leverage" truly kicking in: revenue tripled while operating expenses grew far more slowly.

Very few semiconductor companies can hold product leadership, margins and operating discipline all at once at this speed.

3. FY27 Guidance: Three Legs of Optics, Each Clearing $100M

This is the core of the call. Management's breakdown of FY27 is very clear:

- Full-year revenue growth of 80%+, gross margin flat with FY26 (about 68%), net margin around 50%. - Operating expenses up about 50%, far below revenue growth - the leverage story continues. - The dollar increase comes roughly "half from optics, half from existing copper (mainly AEC, then retimers)".

The optics half is split into three legs, and management unusually gave quantified floors:

Optics product line	FY27 revenue	Role
Discrete optical DSPs (Robin 100G/lane, Cardinal 200G/lane)	Over $100M	Design wins ramping
Silicon photonics PICs (SiPho PIC, DustPhotonics)	Over $100M	Acquisition just closed last week
ZeroFlap Optics	Over $100M	Main hockey-stick driver
Total	Over $600M	Main FY27 growth engine

The math makes it obvious: each of the three legs is over $100M, yet together they exceed $600M, so one leg is clearly larger. The CEO spelled out the answer later - discrete DSPs and SiPho PICs carry "double-digit dollar" ASPs while ZeroFlap Optics carries a "triple-digit dollar" ASP, so ZeroFlap will be the largest contributor to optics revenue. This echoes the earlier point: Credo isn't selling components, it's selling the system-level value of "reliability".

We covered the full context of Credo's playbook of "extending AEC's reliability DNA into optics" in Credo's session at the Barclays Global Technology Conference; this quarter effectively turned that narrative into guidance numbers.

4. Business Deep Dive: Copper Still Earns, Optics Takes the Baton, Memory Is the Dark Horse

• AEC (Active Electrical Cable): still the core growth engine. ZeroFlap AEC is claimed to be 1,000x more reliable than commercial laser-based optical modules with lower power, and is already the first choice for in-rack and multi-rack links within 7 meters; 100G/lane is in volume, 200G/lane is starting, and PCIe 6.0 AEC is advancing. The CEO said bluntly that apart from xAI, no customer is "fully penetrated" - meaning copper's growth hasn't peaked.

• The three optics legs: FY27 is the inflection point. The DustPhotonics acquisition closed last week, bringing highly differentiated SiPho PIC technology with products spanning 800G/1.6T and a roadmap to 3.2T, and an architecture that uses fewer lasers - which improves reliability, power and cost at once, and also eases the industry-wide laser supply bottleneck. More importantly, this SiPho roadmap leads directly to CPO and NPO, with initial CPO/NPO revenue expected in FY28.

• Retimers: growing at both 100G and 200G/lane, with PCIe 6.0 retimers heating up. Blue Heron, a 200G/lane retimer, supports Ethernet, UALink and Ethernet SAN in one chip and is built for scale-out and emerging scale-up.

• An emerging dark horse - memory: Weaver, the first gearbox in the OmniConnect family, targets memory bandwidth density. The CEO used first customer Positron as an example: it launched an inference engine with 2TB of LPDDR, more than 10x any published spec on the market, and feeding 2TB requires lots of Weaver chips - revenue contribution per GPU can reach $2,000-3,000. This piece only reaches volume in FY28, but the ceiling is not low.

5. Management Language: What's "Confirmed" vs. Merely "Expected"

When reading an earnings call, tone is more honest than numbers.

The CEO's wording on the ZeroFlap supply chain was "we feel very good about our existing supply commitments" - a "confirmed" level, because Credo controls the entire ZeroFlap BOM and places its own capacity bets. But on the 200G/lane ramp, the CFO's wording was noticeably more conservative: "200G/lane revenue in FY27 will be relatively light because the industry isn't really there yet" - that's "expected + wait-and-see", handing the ball back to customers and the maturity of the ecosystem.

Process details also reveal confidence: 12nm is the workhorse for AEC 100G/lane, optical DSP 100G/lane is moving to 7nm, there are 5nm projects in flight, and all 200G/lane products use 3nm. The CEO said flatly that "no one will use 5nm" for the 1.6T market; the mainstream will be 3nm, possibly moving to 2nm later. With 3nm capacity tight across the industry, the subtext is that the supply-chain relationships Credo locked in with resources over the past five years are now a moat.

The timing of the 1.6T transition was full of flexible language: customers generally design two physical ports per GPU, and 1.6T can be "4 lanes x 200G per port" or the more conservative "2 ports x 8 lanes x 100G". Credo wins either way - but this also concedes that the timing of the 200G/lane surge isn't in its own hands. This matches the signal from Semtech's FY27 Q1 earnings call: for the real ramp of 1.6T and 200G/lane, the whole industry is still waiting for the same starting gun.

6. Supply Chain and Customer Clues: Neoclouds Are the New Gold Mine; Copper and Optics Will End Up 50/50

Customer concentration: Q4 had four 10% customers at 34%, 27%, 16% and 10%; the top three are the same as Q3, and the fourth is a new 10% customer. Management expects a more diversified customer base in FY27.

Neoclouds are the most promising trend. The CEO described neoclouds as "perfect customers for Credo" - they move fast, optimize architectures and care intensely about network performance and time to deployment. Asked whether tier-2 clouds could reach 10-20% of revenue, he answered directly that "this group of customers could indeed reach the 20% range combined". That expands the growth narrative from five or six mega hyperscalers to a broader ecosystem.

The long-term copper vs. optics path: the CEO believes the two will eventually reach 50/50, and that "optics may surpass copper in the next few years", because the long-term market for optics (including pluggables) is far larger than for AEC pluggables. This is consistent with the industry's structural shift of "supply bottlenecks moving from downstream modules to upstream components" - we broke down this bottleneck shift in The USD 725B Question (W21), and Credo chose vertical integration (owning the BOM) to tackle this structural problem head-on.

As for scale-up, there's revenue in FY27 but it's only the beginning; FY28 will be more substantial. For how this path is converging from an architecture battle, see How Marvell's FY27 Q1 Defines the Shape of Scale-Up - Credo's approach is "customer by customer, deployment by deployment", without betting on a single architecture.

7. What to Watch: Three Things to Track Closely Over the Next Two Quarters

1. Delivery of the second-half optics ramp: FY27 is a "back-half takeoff" script; if ZeroFlap Optics' hockey stick slips, the whole 80%+ growth story loosens. Q2 guidance (still expected at mid-single-digit sequential growth) will be the first signal, with the real test in Q3.

2. 3nm capacity and 200G/lane timing: Credo says it's "ready now", but the starting gun for 200G/lane is in customers' hands. Watch whether 3nm capacity can support that steep second-half ramp, and the actual pull-in pace of customers' 1.6T platforms.

3. DustPhotonics integration and customer concentration: SiPho PIC revenue contribution, the FY28 path for CPO/NPO, and whether neocloud share moves toward 20% as the CEO suggested will decide whether this "copper-to-optics" pivot really holds.

At its core, this report shows a company that has taken "copper cable reliability" to the extreme now trying to transplant the same value proposition into optics - and to turn optics into a $600M business within a year. If the numbers deliver, it's no longer just an AEC company but a system-level player in the AI cluster connectivity layer.

This article is for technology and industry trend analysis only and does not constitute investment advice.