muji (CMF_muji) | 2026-02-22 | HHHYPERGROWTH
Note: This is based on the Q3 FY26 PRELIMINARY release only. Full results + transcript: March 2, 2026 after market.
long CRDO ~10%
$404-408M (prelim) vs $335-345M guide.
Let me put that in context. Management guided $340M midpoint. They printed 406Mmidpoint.That′sa * *+66M beat, ~19% above the top of the range**. On a run rate basis that's a company guide of ~1.36Bannualizedthatprintedat 1.62B.
Revenue trajectory: | | Q3_FY25 | Q4_FY25 | Q1_FY26 | Q2_FY26 | Q3_FY26 (prelim) | | | Jan-25 | Apr-25 | Jul-25 | Oct-25 | Jan-26 | |---|---|---|---|---|---| | Revenue ($m) | 135 | 170 | 223 | 268 | 406 | | YoY % | — | +180% | +274% | +272% | +201% ^^ | | QoQ % | — | +26% | +31% | +20% | +51% !! | | Beat vs guide | +$10M | +15M|+28M | +28M|+66M !!! | | GM [GAAP] | 64.1% | 67.4% | 67.4% | 67.5% | TBD (Mar 2) | | Op margin [GAAP] | (10.5%) | 36.8% | 27.2% | 29.4% | TBD (Mar 2) |
That beat streak: +10 → +15 → +28 → +28 → +66 !!!
Five consecutive quarters of sandbagging, each wider than the last. But +$66M is not sandbagging. That's a fundamental revision to demand visibility. Something changed materially in Q3.
FY26 guide raised to >200% YoY. That implies ~$1.3B+ for the year. With Q3 prelim at $406M and three quarters done at 135+170+223+268 = $796M... wait, that's only the first three. Let me recalculate: Q1+Q2+Q3 = 223+268+$406 = $897M. FY26 ends April 2026. So Q4 needs to bring it to whatever >200% implies off FY25 base.
FY25 revenue was approximately 223+170 (Q4)+earlier quarters. From the data: FY25 TTM was roughly 435 − 440M. > 2001.31B. With $897M in first three quarters, Q4 needs to deliver $410M+ for the math to work at the low end. Yet Q4 guidance is "mid-single-digit sequential." $406M * 1.05 = $426M. That works.
WAIT. Reread: "mid-single-digit sequential" off 406M= 426-430M. That's actually a solid quarter — just looks bad versus the +51% QoQ they just delivered.
Here's the thing... Credo is one of the purest "building block" stories in my portfolio.
What Credo actually is: Full-stack SerDes IP and active electrical cable (AEC) company. Their IP is embedded inside hyperscaler custom ASICs (Google TPUs, Microsoft Maia, Meta MTIA, etc.). When hyperscalers build AI training clusters at scale, CRDO's SerDes drives every high-speed interconnect in the rack. This is not a peripheral product — it is silicon IP embedded in the critical path of AI infrastructure.
Scale in Platform: ^^
Architecture: This is picks-and-shovels for the AI connectivity layer. Every GPU-to-GPU, GPU-to-storage, GPU-to-network path at hyperscaler scale runs through high-speed SerDes. As cluster sizes scale from 8K to 32K to 100K+ GPUs, interconnect density compounds faster than compute. CRDO rides that leverage.
GTM model: B2B semiconductor, not self-serve. But the IP licensing model — where CRDO IP gets designed into custom silicon — creates multi-year revenue visibility once a design win is secured. This is why the guidance methodology is structurally conservative: management literally cannot know design win pull-through timing until wafers ship.
Crowdsourced intelligence: No. But competitive moat from density of design wins and cumulative SerDes IP development creates a data-loop equivalent — more design wins = more silicon learnings = better next-gen IP.
The market read "mid-single-digit sequential" as a deceleration signal. Technologist take:
Management has sandbagged EVERY guide — guide accuracy history: +7%, +9%, +13%, +12%, +19% beats. The pattern is systematic, not isolated.
Q3's +$66M beat was itself a positive revision — if Q3 came in that far above their own internal view, they're not able to predict their own demand curve. Why trust Q4 guide?
Mid-single-digit off 406M= 426M. That's NOT bad. That's $1.32B-ish for FY26, which confirms the >200% YoY raise.
WHAT I WANT TO KNOW ON MARCH 2: Has the 5th hyperscaler crossed 10% of revenue? What is the mix shift between AEC, IP licensing, and ALCs? Any color on FY27? Is the Q4 guide truly conservative or a genuine air pocket before the next ramp?
Customer concentration risk remains the primary concern. Top 1 customer: 42% of revenue. Top 3: 82%. Four customers >10%. This is CRDO's structural vulnerability — a design win loss or inventory digestion at Microsoft would crater a quarter. Until revenue diversifies below 60% for top 3, this stays as a risk flag.
Profitability swing: From -14.5Moperatingincome(Q1FY25)to+60.7M (Q1 FY26) to +$78.8M (Q2 FY26). This is operating leverage working exactly as it should in a semiconductor IP business.
Balance sheet: $813.6M cash, zero debt. Cash up 88% in 6 months. Something capital-raise happened — confirm on March 2 call. FCF margin: 23% and expanding.
SBC: $45.3M in Q2 FY26 on $268M revenue = 16.9%. Elevated but declining as % of revenue as scale grows. Watch for absolute dollar trajectory.
Margins [GAAP]: GM 67.5% !! — semiconductor IP businesses should target 70%+. At 67.5% and rising, on a scaling revenue base, CRDO is approaching the inflection where GMs compound with scale. IMHO this reaches 70%+ by FY27.
Rule of 40: ~201% YoY + 29.4% operating margin = north of 230. Not a typo.
Tier 1 — Revenue growth >75%? Check (+201% YoY). Accelerating QoQ? Check (+51% QoQ). Margins expanding? Check. Balance sheet strong? Check. Platform defensible? Check.
CRDO is a Tier 1 position by every muji criterion. The ONLY question is portfolio sizing given the concentration risk (both customer concentration AND single-product exposure to AI capex cycle timing).
Promises tracked:
Language shift to watch: Any softness in "design win funnel" language = red flag. Any commentary on inventory buffers at hyperscalers = risk signal.
Atlas rates conviction 4/5 and flags the "67.5% GAAP vs 45% non-GAAP FY guide" confusion as a margin label issue to resolve on March 2. I think that's right — it reads like the 45% refers to operating margin guide for FY26, not gross margin.
Where I push further than Atlas:
The beat magnitude changes the narrative. +$66M at 19% above the top end is not a guidance methodology quirk. Something structurally changed in demand. Management needs to explain this on March 2.
The $813M cash raise is unexplained. Cash went from $431M → 813MintwoquarterswhileOCFwas 100M. There's a $270M+ unexplained capital event. This needs resolution.
FY27 setup is the real question. Q4 mid-single-digit sequential sounds like a pause before the next hyperscaler ramp. If March 2 includes any FY27 color with continued growth, this is a sustained Tier 1 story.
Thesis: STRENGTHENING.
CRDO is delivering extraordinary results at an extraordinary growth rate with expanding margins and a dominant competitive position in the most important secular theme in tech (AI compute scaling). The Q3 prelim beat is one of the largest beat-to-guide ratios I've seen from any company at this revenue scale.
I am biased to add on any post-March 2 weakness if:
Risk I'm watching: Customer concentration at top 3 = 82%. If one hyperscaler pulls back capex (Microsoft, Google, or Meta), a quarter gets ugly fast. The serdes IP stickiness is the defense.
Position: Long CRDO ~10%. High conviction. Waiting for March 2 for full picture before final sizing decision.
-muji