Credo Technology Posts Record Quarter, Unveils New Silicon Photonics Wins and Scale-Up Optics Push Beyond $600 Million Optical Target
Q1 FY2027 earnings call, September 1, 2026
Credo Technology delivered another blowout quarter, with revenue hitting a record $479 million, up 10% sequentially and more than doubling year-over-year, marking the company's seventh consecutive quarter of triple-digit growth. But the more consequential disclosures came from CEO Bill Brennan's granular breakdown of where the next legs of growth are coming from: two new undisclosed silicon photonics design wins from the DustPhotonics acquisition, a fresh entry into the Open CPX consortium for scale-up optical networking, and early detail on OmniConnect, a memory-fanout product line targeting AI inference that Brennan says can represent "thousands of dollars of Credo content per GPU."
Non-GAAP gross margin came in at 68%, with net income of $236.3 million, up 140% year-over-year and a 49.3% net margin. CFO Dan Fleming guided Q2 revenue to $525 million to $535 million and reiterated that fiscal 2027 will show more than 85% year-over-year revenue growth, anchored by over $600 million in optical revenue, split roughly evenly across ZeroFlap Optics, silicon photonics PICs, and optical DSPs, each expected to contribute more than $100 million.
DustPhotonics Integration Already Paying Off With Two New Design Wins
The most notable new data point from the call was confirmation that the DustPhotonics silicon photonics team, acquired earlier this year, has already landed design wins with what Brennan described as "2 major players" for next-generation ramps beginning in fiscal 2028, with some contribution possibly starting late in fiscal 2027. Importantly, these initial wins do not include Credo's own DSP, meaning there is unrealized upside if Credo can cross-sell the combined DSP-plus-PIC package over time. Brennan called the pace of traction "amazing," adding that the small Dust team's rapid customer engagement is "evidence of the fact that their technical solution is absolutely leading edge in the market." Analyst Christopher Rolland of Susquehanna pressed for detail on the economics of DSP-PIC integration, and management confirmed the co-marketing opportunity remains a future catalyst rather than a current contributor.
Credo Joins Open CPX Consortium, Targets Near-Package Optics for Scale-Up
Credo disclosed it has joined the Open CPX MSA consortium, positioning itself for the emerging near-package optics, or NPO, market tied to scale-up AI architectures that require roughly 10x the connection density of today's scale-out networks. Brennan explained the company's approach as deliberately agnostic across form factors: "We'll basically look at what our customers are asking us and what our customers are driving towards." He confirmed confirmed NPO design wins are already in place, with ramps expected to begin in fiscal 2028, and said Credo will pursue both component sales and full system-level optical solutions, leading with its silicon photonics PIC before eventually offering the complete optical engine. This adds a third major optical growth vector alongside DSPs and ZeroFlap Optics, and follows Credo's broader thesis that AI networks are becoming increasingly heterogeneous across copper and optical mediums rather than converging on a single technology.
OmniConnect Opens a New Multibillion-Dollar Inference Opportunity
Management gave its most detailed explanation to date of OmniConnect, a SerDes-and-gearbox architecture targeting the memory fanout bottleneck that increasingly constrains AI inference performance. The product combines a low-power, small-footprint SerDes with reach of up to 10 inches and a gearbox called Weaver that bridges XPU interfaces to LPDDR memory. Brennan said the design is "future enabled," since a shift from LPDDR5 to LPDDR6 memory would only require swapping the gearbox rather than a full chip redesign. The first named customer is Positron, whose forthcoming 2-terabyte memory-capacity product Credo says can dramatically improve throughput for frontier models by fitting more of the model directly into memory. Brennan noted Credo's roadmap can achieve bandwidth "very comparable" to HBM5 while avoiding the reliability issues associated with packaging XPUs and HBM together. Revenue from OmniConnect is expected to begin in fiscal 2028, and management has previously described the opportunity as worth billions of dollars annually.
PILOT Telemetry Emerges as a Deeper Moat Than Previously Understood
Responding to a question on whether Credo's growing base of network telemetry data constitutes a genuine competitive advantage, Brennan offered an unusually detailed technical explanation of the PILOT software platform underpinning ZeroFlap Optics. The system monitors six link directions between XPUs and switches in real time, tracking metrics like eye height and SNR post-histograms to detect degrading link quality before failure, what Brennan called "a check engine light" rather than a simple red-or-green status indicator. More striking were two specific capabilities disclosed for the first time: the ability to detect faint electrostatic discharge damage on a transceiver before it causes failure, and detection of dust contamination on fiber connections that can cause multipath interference. Brennan framed the financial stakes bluntly, noting that faster cluster bring-up alone can be worth "hundreds of millions or even $1 billion" given how much idle GPU capacity costs per month, while the ultimate goal is pushing network uptime to 99% or higher.
Customer Concentration Remains High Even as Diversification Continues
CFO Dan Fleming disclosed that Credo's largest customer represented 33% of quarter revenue, followed by 28%, 13%, and 10% from the next three, with the fourth 10%-plus customer different from the prior quarter's fourth customer. Management continues to expect three to four customers above the 10% threshold in coming quarters, and emphasized that hyperscaler spending is broadening across product lines rather than remaining concentrated in AECs alone. Credo's optical DSP and retimer businesses both posted record revenue in the quarter, with retimer growth driven by the Screaming Eagle and newer Blue Heron products at 100-gig and 200-gig-per-lane, respectively.
Active Electrical Cables Still the Core, But Growth Rate Naturally Decelerating
Responding to a modeling question from Bank of America's Vivek Arya on the split between AEC and optical growth, Brennan acknowledged that AEC growth will look slower in relative terms simply because of the size of the base it is growing from, after more than doubling from fiscal 2024 to 2025 and more than tripling from fiscal 2025 to 2026. He characterized the AEC opportunity as part of a broader pluggable transceiver market forecast to grow from 60 million units in calendar 2026 to 175 million units by 2030, according to third-party market forecasters cited on the call. Credo is also developing Active LED Cables, a third pluggable category using micro-emitter technology capable of reaching 30 meters, with initial revenue targeted for fiscal 2028 and a planned OCP demonstration in October.