Earnings Call Highlights: Credo (CRDO) | FY26 Q3

1. Summary | Credo keeps setting revenue records as Zero-Flat Optics ignites growth ahead of schedule Credo delivered an exceptionally strong FY26 Q3, with quarterly revenue topping $407.0M, up more than 200% YoY. Management stressed that the company is in the fastest expansion phase in its history (6x growth in two years), driven by AI infrastructure's extreme demands for reliability and power efficiency. The biggest highlight of the call: volume production of Zero-Flat Optics was pulled in from the second half of FY27 to FY27 Q1, showing how urgently hyperscalers need to solve link flaps in optical links. Going forward, the "three growth arrows" will be AEC, ZF Optics, and OmniConnect for inference workloads.

   
   

2. Financials (non-GAAP)

Item	This quarter (FY26 Q3)	QoQ / YoY
Revenue	$407.0M	+52% QoQ / >200% YoY
Gross Margin	68.6%	+92 bps QoQ
Operating Margin	49.6%	+327 bps QoQ
Net Income	$208.8M	+63% QoQ
OpEx	$77.4M	+35% QoQ (higher R&D investment)
CapEx	$26.5M	Mainly production mask sets
Cash and equivalents	$1.3B	Includes ATM equity raise and strong free cash flow

• Revenue mix: Product revenue makes up the vast majority; IP licensing revenue is no longer broken out separately. The top three customers (all hyperscalers) accounted for 39%, 32%, and 17% of revenue, 88% combined.

3. Outlook (FY26 Q4 / FY27) •

   • Q4 revenue guidance: $425M - $435M ($430M midpoint), with steady growth expected. •

   • Q4 gross margin estimate: 64% - 66% (a more conservative forecast given product mix changes). •

   • FY27 full-year outlook: revenue expected to grow >50% YoY, driven by the continued ramp of 800G AEC and contributions from ZF Optics.

4. Product Updates: AEC (Active Electrical Cables) & Systems •

   • Market position: AEC has become the de facto standard for intra-/inter-rack interconnects within 7 meters, replacing low-end optical modules. •

   • Technology progress: 100G/lane continues to ramp, and 1.6T (200G/lane) AEC is ready, supporting Ethernet, UALink, and ESUN protocols. •

   • New areas: PCIe Gen 6 AEC is sampling, with volume production expected in the first half of FY2027.

   • Optical (Zero-Flat Optics & ALC) •

     • ZF Optics: Solves the link-flap problem of conventional laser-based modules, improving reliability 1,000x. First customer TensorWave has shipped, and three more customers (including a hyperscaler) are in qualification. •

     • ALC (Active LED Cables): Combines Hyperlume's microLED technology to fill the 7–30 meter reach gap. Sampling expected in FY27, volume production in FY28.

   • IC (DSP & Retimers) •

     • Blue Heron: A 200G per lane retimer designed specifically for AI scale-out. •

     • OmniConnect (Weaver): A breakthrough gearbox product that lets XPUs connect over VSR to DDR memory instead of expensive HBM, sharply lowering inference cost; volume production expected in FY28.

5. Looking Ahead •

   1. Heterogeneous interconnect trend: Credo believes future data centers won't rely on a single technology but on a heterogeneous mix of short-reach copper (AEC), mid-reach ALC, and long-reach ZF Optics. •

   2. Vertical integration advantage: The acquisition of Comira adds Layer 2 protocol, error correction, and security IP, letting Credo offer complete solutions from SerDes up to the system level. •

   3. Supply chain positioning: Credo has secured wafer capacity from 12nm to 3nm and built FY27 inventory of key components such as lasers, enough to support raised revenue expectations.

6. Q&A Highlights

   • Question 1: Why was ZF Optics production pulled in?

     • Management: Customers place enormous weight on "time to stability" for AI clusters. ZF Optics' real-time telemetry can predict link degradation in advance, something conventional optical modules can't do, and demand has exceeded expectations.

   • Question 2: Will CPO replace AEC?

     • Management: Right now there is more noise than signal around CPO. Until NPO/CPO reach "bulletproof reliability," market share will stay in the single digits. Credo prefers to deliver CPO-like low-power benefits through ALC first.

   • Question 3: Is revenue too concentrated?

     • Management: Although the top four customers contribute heavily, Credo has broken into a fifth hyperscaler and begun winning neocloud orders (e.g., TensorWave), so the customer base is diversifying.

     
     

7. Risks and Watch Points

   • Customer concentration: The top three customers account for nearly 90%, so a CapEx adjustment at any one of them could swing revenue sharply.

   • Gross margin volatility: As the mix shifts from pure ICs to system products that include optical components, gross margin may be unstable quarter to quarter.

   • The 200G transition: Demand for 1.6T is clear, but aligning customer adoption timelines (e.g., NVIDIA's Rubin platform) with Credo's AEC remains a challenge.

8. Bottom Line

   • In one sentence: Credo is in the "high-growth sweet spot" of moving from a single product (AEC) to a comprehensive AI connectivity platform.

   • Core moat: Building on SerDes leadership, Credo extends up into software telemetry and system design, solving the most painful problem in AI clusters: stability.

   • Investment takeaway: Strong operating leverage (revenue up 50% while OpEx rises only modestly), plus long-term engines like ZF Optics and OmniConnect arriving early, leave room for continued upward revisions to mid-term valuation.