CRDO › analyze
CRDO · Analyze
Price $164.60 · MktCap $30.9B · ~$1.13B net cash · Beta 3.23 AI-datacenter connectivity: Active Electrical Cables (AECs), SerDes/PAM4 DSP IP, optical DSPs, PCIe retimers. Cayman-domiciled, sells to hyperscalers. FY ends April.
Verdict: WATCH (accumulate on further weakness) · Conviction 6.5/10 · Fair value $160–210
Genuinely one of the best businesses riding the single biggest secular wave in tech — and today it got 18–20% cheaper for the right reasons, not a broken thesis. But extreme customer concentration (top-2 = 61% of revenue), gross-margin compression, a beta of 3.23, and a valuation that still demands beat-and-raise keep this a high-quality/high-risk WATCH, not a table-pound buy. Today's reset opens a window; the concentration math means you want a wider margin of safety before sizing up.
⚠️ Live event — a beat-and-drop, today
Credo reported Q1 FY2027 after the close Sep 1 and the stock fell ~18–20% Sep 2 despite beating: - Revenue $479.0M, +114.7% YoY, +9.6% QoQ (7th straight triple-digit-growth quarter), beat ~$472–473M consensus. Non-GAAP EPS $1.20 vs ~$1.17. A clean top- and bottom-line beat. - The drop was a valuation reset, not a thesis break. Pre-print P/E was ~82x (vs ~47x semi median); the stock had run +71% YTD to a $206.63 close. Guidance was reiterated, not raised (Q2 FY27 rev $525–535M ~in-line; FY27 >85% growth, >$600M optical, ~50% non-GAAP net margin) — and for a stock priced for perfection, "in-line + visible margin compression" was reason enough to unwind. BofA cut its PT $340→$275 (kept Buy), explicitly calling it a "valuation reset."
The read-through: the bar is now beat-AND-raise. An in-line print repeats today.
⚠️ Data traps checked
- EV field ($37.1B) looks inflated — with ~$1.13B net cash, EV should sit below the $30.9B market cap (~$29.7B). I use the net-cash-adjusted figure, not the vendor EV, for all multiples.
- Forward P/E 17.1 (vendor) is on a different/aggressive basis than my ~25x FY27 non-GAAP — do not anchor a site-rendered
fwdtrim to it (pitfall-multiple-trim-inherits-the-broken-vendor-field). Trim expressed as a level below. - Graham IV $25 vs $164 — meaningless for a hypergrowth semi; discarded.
1. Fundamentals — no longer a hope stock; it prints real cash
| Metric | FY2023 | FY2024 | FY2025 | FY2026 | Read |
|---|---|---|---|---|---|
| Revenue | $184M | $193M | $437M | $1.34B | 3yr CAGR 93.5%; +207% in FY26 |
| Gross margin | 58% | 62% | 65% | 68% | Strong; but GAAP GM 68.2%→64.5% QoQ, guided lower ⚠️ |
| Operating inc | −$19M | −$37M | $37M | $445M | 33% op margin — real operating leverage |
| Net income | −$17M | −$28M | $52M | $472M | 35% net margin |
| FCF | −$46M | $17M | $29M | $407M | 30% FCF margin — the inflection |
| Cash / Debt | — | $67M/$14M | $236M/$16M | $1.16B / $25M | Fortress net cash |
| Dil. shares | 147M | 155M | 181M | 188M | ↑ ~8.6%/yr — SBC dilution is the tax |
- Q1 FY27 run-rate ~$1.9B annualized and climbing; FY27 guide implies ~$2.4–2.5B (>85% growth) at ~50% non-GAAP net margin → ~$6.4 non-GAAP EPS.
- The business flipped from cash-burn to $407M FCF in a single year. This is the rare AI-infrastructure name where the profits are already here, not promised.
- Watch-item: gross margin. 68.2%→64.5% GAAP QoQ with more erosion guided — as volume ramps and mix shifts, the >50% net-margin promise leans on volume, not price. Margin is the new tripwire.
Fundamentals grade: A−. Explosive profitable growth, fortress balance sheet, real FCF. Docked only for SBC dilution (~8.6%/yr) and the emerging margin fade.
2. Moat — Narrow (evergreen 5/10)
- The moat is an integrated systems stack, not one chip. Credo owns the SerDes/PAM4 DSP IP and the cable/connector design end-to-end. AECs are the wedge: the value isn't the copper (Amphenol/TE make passive cable) — it's the DSP-in-the-connector that makes copper viable at 100G→224G/lane, with claimed ~1,000x reliability ("ZeroFlap") and ~50% lower power vs optics in-rack. Credo effectively created the AEC category and holds ~75% share.
- Moat sources, ranked: (1) per-platform design-win lock-in — each AEC is co-engineered and qualified to a hyperscaler rack over 12–18 months; (2) integrated iteration speed (owns chip + cable); (3) a reliability data moat. Durability is medium — real switching cost within a platform generation, but re-competed every architecture generation.
- Subscale outside AECs: behind Astera Labs in PCIe retimers (ALAB has PCIe 6 in volume; Credo's "Toucan" just hit PCIe 6.x compliance — challenger, not leader), behind Marvell in optical DSP, and Broadcom/Marvell have deeper SerDes plus switch-ASIC bundling leverage Credo lacks.
- Near-term the moat is strengthening: Credo won 1.6T AEC design-ins on NVIDIA Vera Rubin NVL144 and Kyber NVL576 scale-up — NVLink adjacency, not lock-out. Credo leads the 800G→1.6T AEC replacement cycle.
- Terminal risk: copper→optical/CPO at ≥224G. Physics eventually favors optics/co-packaged optics for in-rack; that's the multi-year disruption vector, and Credo follows on CPO.
3. The core risk — concentration + vertical integration
- Top customer 33% of Q1 revenue; next three 28% / 13% / 10% → top-2 = 61%, top-4 = 84%. Any single hyperscaler capex pause (Amazon / Microsoft / xAI) hits >30% of revenue at once. Not visibly worsening, and management guides toward 3–4 >10% customers — but this is the structural fragility.
- The bear thesis — hyperscalers in-source SerDes or buy from Broadcom — is real but partially mitigated. AEC is a messy systems problem (thermal, mechanical, qualification, RMA) hyperscalers dislike owning; Credo's value is highest at the physical layer, not a reusable IP block. Verdict: a durable bridge, not a permanent toll-road. More resilient than pure-IP licensing, but a single customer's architecture shift resets the model.
4. Sentiment
- Beat top and bottom, sold off on valuation — see live event above. Sell-side stays structurally bullish (mean target ~$277 vs $164 spot), but targets are being trimmed (BofA $340→$275). The gap to target is large; treat the cut targets as the live read.
- Short interest low (~4%); the move is de-rating, not a short attack. Beta 3.23 means this name moves 3x the market — position sizing must respect that.
5. Valuation — the reset brought it back to reasonable
Type: hypergrowth semi, no dividend → Graham/DYT/DDM N/A. Use forward P/E, EV/Rev, growth-adjusted.
| Model | Input | Output |
|---|---|---|
| EV/Rev | ~$29.7B EV ÷ FY27 ~$2.45B | ~12x fwd — reasonable for 85% growth @ 68% GM |
| Fwd P/E (non-GAAP) | ~$6.4 FY27 EPS | ~25x — PEG well under 0.5 |
| PEG | 25x / 85% growth | ~0.3 — cheap if growth holds |
Fair value $160–210 (midpoint ~$185): - 25–35x FY27 non-GAAP EPS ($6.4): $160–224 (25x reflects the concentration discount, 35x a clean-execution premium). - 12–15x FY27 EV/Rev + net cash: ~$160–202.
At $164 the stock sits at the low end of fair after today's reset — no longer the ~$308 "priced for perfection" level, but not a screaming bargain either given the concentration tail. The $277 analyst mean looks stale-high post-print.
Synthesis & verdict
For a hypergrowth semi the framework weights Moat + Sentiment + Valuation, with Fundamentals as the (here, unambiguously strong) foundation. They pull in tension: - Fundamentals (A−) and post-reset Valuation (fair-to-cheap on growth) say this is now an interesting entry, not the frothy $308 name. - Moat (narrow, 5) + concentration (top-2 = 61%) + margin fade + beta 3.23 + "needs beat-and-raise" cap the conviction. One hyperscaler pause is a >30% revenue event, and the stock still punishes in-line prints.
Net: a high-quality, high-volatility AI-infrastructure compounder at a suddenly reasonable price, with a fragile revenue base. That's a WATCH with an accumulate tilt — start small into today's weakness, add aggressively only on a deeper flush (the beta will provide the opportunity), and let the concentration/margin data confirm before sizing up.
Conviction 6.5/10.
- Entry: begin sub-$150; aggressive add sub-$120 (beta 3.23 makes these reachable). Trim into ~40x fwd non-GAAP (~$260–280), where it re-prices for perfection.
- ↑ to BUY (7.5+): concentration broadening to 4+ >10% customers, gross margin stabilizing, another beat-AND-raise.
- ↓ to AVOID: a top-customer capex pause, GM breaking below ~62%, or growth decelerating below the guided 85% with no raise.
A portfolio-specific passage was removed from the public build.