Credo and the copper window
A complete-suite interconnect vendor, roughly ninety percent sold to ten buyers — and what one quarter did to the price of that book.
The company the wires made visible
In July 2025, Elon Musk posted photos from inside xAI's Colossus 2 data center in Memphis — and what the pictures showed off was not the accelerator racks but the wiring behind them: thousands of neatly organized purple cables [3]. Those cables are Credo's signature product, selling for roughly $300 to $500 each by one industry analyst's estimate, and the company behind them — a 17-year-old Silicon Valley semiconductor firm — spent most of the AI boom unmentioned beside its leaders while its stock more than doubled in 2025 on top of a 245% rise in 2024 [3]. This is a deep look at what that company actually holds, and at what one earnings report did to the price of it.
Where the copper window sits
Credo's core product category, the active electrical cable, occupies a physics-defined window in the AI data center. The company's own filing positions AECs as "a high-performance alternative to short, thick DACs and high power, high-cost Active Optical Cables" at 100G through an emerging 1.6T [1]:
reach ──────────────────────────────────────────────────────▶
DAC AEC (Credo) AOC / optics
passive copper retimed copper lasers + fiber
cheapest, shortest "high-performance longest reach,
reach; thick and alternative" to both highest power
hard to route neighbours [1] and cost
~in-rack rack-scale across rows/sites
The window exists because passive copper runs out of reach and optics still costs power and money at short range. Credo's stated edge in holding it is cost: silicon that reaches "similar performance to leading competitors' products but at a lower cost and more highly available legacy node," the n-1 advantage in the company's own words [1]. Around the cables sits a full stack — seventh-generation optical PAM4 DSPs, ZeroFlap optical transceivers, PCIe retimers, SerDes chiplets and licensed SerDes IP — which is what licenses the filing's claim to be "the only company in our industry offering a complete suite of high-performance connectivity solutions" [1].
The customer book
The buyers are the hyperscale few. Reporting on the company names Amazon, Microsoft and xAI among its largest AEC customers, with adoption broadening to include Meta [3]; the filing discloses the shape of that book without the names: top-10 customers were approximately 90% of fiscal 2026 revenue, and two customers each accounted for 10% or more [1]. The model is deliberately direct — Credo engages the end users, who then require their OEMs, ODMs and module makers to design Credo in — and the Oracle partnership that produced ZeroFlap Optics, addressing the reliability failures known as Link Flap, is that model working in public [1].
The flanks are guarded by much larger names. Broadcom, Marvell and Astera Labs are the filing's named principal competitors, competing "with respect to some, but not all" of the solutions — Broadcom and Marvell against the optical DSP line, Astera most directly in PCIe retimers [1]. Nobody is named against the whole suite; everybody is named against a piece of it.
The quarter
The first quarter of fiscal 2027, ended August 1, 2026, was a growth quarter on every stated line: revenue of $479.0 million, up 9.6% sequentially and 114.7% year over year; non-GAAP gross margin of 68.0%; non-GAAP net income of $236.3 million; $764.3 million in cash and short-term investments [2]. The scale shift is easiest to see against the company's own recent past — full fiscal-year 2025 revenue was $436.8 million [3]:
FY2025, entire year $436.8M ████████████████████
Q1 FY2027, one quarter $479.0M ██████████████████████
One quarter now exceeds what the whole company earned two fiscal years ago. The guide extended the growth — Q2 revenue of $525 to $535 million — but held the margin: non-GAAP gross margin of 67.0% to 69.0%, a band whose midpoint is exactly the 68.0% just delivered [2].
What the tape did with it
The report beat the frozen consensus on both lines this site records — EPS of $1.20 against a $1.17 estimate [6]. The market's answer was the most violent of this earnings wave. The closing prices tell it plainly [5]:
2026-08-31 close $226.19
2026-09-01 close $206.63 −8.7% report published after this close
2026-09-02 close $165.22 −20.0% first post-report session
across the two sessions: −27.0%
Intraday coverage on September 2 was headlined at an 11% drop and deepened from there [4]; the session finished down 20.0% — arithmetic from the two cited closes. A beat met with a one-fifth repricing is not noise; it is information about what the price had been carrying.
The transmission mechanism
The sell-off coverage named its reasons, and they are all in the release's own arithmetic. GAAP gross margin fell to 64.5% from 67.4% a year earlier even as non-GAAP gross margin rose; GAAP operating expenses more than doubled to $188.4 million from $89.6 million, outpacing revenue growth; share-based compensation rose to $88.0 million from $35.5 million; GAAP operating margin slipped to 25.2% from 27.2% [4]. Add the guide's flat non-GAAP margin band [2], and the quarter's message reads: hypergrowth continues, profitability's trajectory does not steepen.
Why that message moved the price by a fifth is about the price, not the quarter. The stock entered the print having doubled in 2025 after rising 245% in 2024 [3] — an entry level at which the marginal fact is no longer whether revenue grows, but whether margins expand while it does. When the guide answered "held, not expanded," the revenue beat was already in the price and the margin path was not. That is a description of how expectation-laden prices process mixed reports — the same report at a different entry price is a different day — not a view on where the price goes next.
What the record will answer next
Four things the cited record already names, worth re-reading next quarter: whether the non-GAAP margin band holds as the portfolio tilts toward optics — the CEO's own framing spans "from millimeters to kilometers, with solutions across optics and copper" [4]; whether GAAP operating expenses keep outpacing revenue [4]; whether either 10%-plus customer grows further as a share of revenue [1]; and whether the named rivals start being named against the suite rather than its pieces [1]. The reaction row on this site's earnings record will carry the measured price answer as the sessions accumulate [6].
Sources
- [1] Credo FY2026 10-K (filed 2026-06-15) — Item 1 Business: products, competition, customer concentration www.sec.gov ↗
- [2] Credo Q1 FY2027 financial results — 8-K Exhibit 99.1 (2026-09-01) www.sec.gov ↗
- [3] CNBC, '$500 purple cables put Credo in middle of the AI boom' (2025-10-17) — cable economics, customer roster, share history www.cnbc.com ↗
- [4] The Motley Fool via Yahoo Finance, 'Credo Stock Drops 11% Despite Beating on Revenue and EPS' (2026-09-02) — GAAP margin, operating-expense and share-based-compensation detail finance.yahoo.com ↗
- [5] Yahoo Finance — CRDO daily closing prices (2026-08-31 through 2026-09-02) finance.yahoo.com ↗
- [6] TradingPilot earnings record for the quarter — reported EPS vs the frozen consensus, with the cited review tradingpilot.ai ↗