Credo Technology Group Holding CRDO

At a glance

One-line thesis. Credo is one of the cleanest AI-connectivity hyper-growth stories in the market — FY26 revenue tripled to $1.34B (+206%), gross margin is 68%, it converts a third of revenue to operating cash, and it sits on $1.44B of cash against ~$21M of debt — but at $265 the stock trades at 36× sales and ~44× FY27E EPS with a 3.20 beta, essentially at both our base case (~$270) and the Street's $269, growth is decelerating from a hyper base, and there is no independent expert thesis in our KB to underwrite the premium — so this is a Watch: a wonderful business we want on a reset, not at fair value.

1. What it is

Credo Technology Group (Nasdaq: CRDO) is a high-speed connectivity company for optical and electrical Ethernet in AI data centers. Its products — integrated circuits, active electrical cables (AECs), and SerDes chiplets, all built on proprietary SerDes/DSP technology, plus SerDes IP licensing — are the plumbing that connects GPUs, switches, and racks. Founded 2008, IPO 2022-01-27; headquartered in San Jose, CA and Cayman-incorporated (KY ISIN — hence a near-zero ~0.7% effective tax rate); CEO William J. Brennan; ~500 employees. Fiscal year ends early May (FY26 = year ended 2026-05-02).

Revenue mix — from IP-and-product mix to a product machine:

• By product: FY26 is reported as a single segment ($1.34B). The FY25 split shows the shape: Product $412M (94%), License $12.5M, Product Engineering Services $12.1M — the IP-licensing lines that were ~25% of revenue in FY23 have shrunk to a rounding error as AEC/optical product shipments exploded.
• By geography (FY26 ship-to): United States $768M (58%) · Hong Kong $378M (28%) · Rest of world $85M (6%) · China $81M (6%) · Taiwan $23M (2%). Note the violent mix shift: FY25 was Hong Kong-led (56%); FY26 is US-led — consistent with US hyperscalers driving the ramp. HK + China still = ~34% of ship-to revenue, a real export-control/geopolitical exposure (§11). The data provides no customer-level split; hyperscaler concentration is known to be high but we cannot quantify it from this file — flagged honestly.

2. The expert thesis — what the panel actually says (traceable)

No expert-panel conviction coverage — this note is fundamentals-driven. The Synthos KB holds only 5 traceable claims on CRDO, and just 1 from an independent voice — and that voice is neutral, not bullish. There is no Visser-grade conviction thesis here, and we will not manufacture one:

• The one independent claim is a warning about price, not a thesis about value. Forward Guidance (skill 1.0, neutral, conviction 60, dated 2026-04-16): "Credo fell 220→80 on nothing fundamental then ripped back — pure trend-following algos plus copper-to-optical narrative, illustrating price dislocation from fundamentals" (forward_guidance-Weratj-dnOI:8de1c1b5fd). Honest weighting: this is a market-structure observation that cuts against paying up today — the same flows that produced +197% in 12 months can run in reverse.
• Everything else is management's own book (half-weighted, CRDO_mgmt, skill 0.5, all 2026-06-01): Q1 FY27 revenue guided to $465–475M, implying continued sequential growth off Q4's $437M (CRDO-earnings-2026Q2:57d4107105, conviction 80); non-GAAP gross margin 67–69%, roughly flat vs Q4's 68.3% (:60a53aa6e1); non-GAAP opex $86–90M, up modestly from $81.7M (:d69726011c); and a strategy thesis that its vertically integrated approach improves GPU utilization, network reliability, and power costs through FY27 (:aa929028f5, conviction 75).

Honest composite note. Breadth 1, stance neutral, plus the company talking its own book. The bull case in §3 is built from the financials and consensus estimates, not from expert conviction — which is exactly why the verdict is calibrated down to Watch despite elite fundamentals.

3. Synthos scores & the Bull / Base / Bear cases

The one-glance judgment — three scores, 0–10, each anchored to real metrics (not probabilities we can't honestly calibrate):

The three cases (our own scenario model — assumptions shown; each target is a ~12–18-month fair value). We deliberately do not attach probabilities: the base case is the expected path, so a weighted blend would just restate it with false precision. The cases bound the range; the scores summarize them.

Synthos fair value = the base case, ~$270 (+2%), with the full $150–$350 span as the honest range. Our base lands essentially on the Street's $269.18 — not by anchoring, but because ~30× FY28E earnings power is a defensible multiple for this growth/deceleration profile and the price already sits there. The asymmetry is the tell: +32% bull vs −44% bear from a fully-priced start. This is a tracked call — the Forecaster Scorecard grades it once it matures.

4. Exponential Potential

Synthos separates compounders (durable high returns on capital) from exponentials (accelerating, multi-baggers-from-here). CRDO is a genuine but decelerating exponential — the mirror image of MRVL's accelerating profile:

• Forward growth: consensus revenue $2.43B FY27E → $3.62B FY28E → $4.55B FY29E → $5.13B FY30E → $6.00B FY31E (~25% CAGR FY27→FY31); EPS $6.10 → $8.94 → $10.39 → $10.85 → $12.50 (~20% CAGR).
• Acceleration (the 2nd derivative) is negative: +205.7% (FY26 actual) → +82.3% (FY27E) → +48.6% (FY28E) → +25.7% (FY29E) → +12.7% (FY30E). Nothing wrong with that — every hyper-grower decelerates — but it means the multiple must compress on schedule, and a 36× EV/sales entry gives no cushion if deceleration runs ahead of schedule.
• Room to run: at $49.5B, Credo is roughly a quarter of MRVL's size against an overlapping AI-connectivity TAM; a 2× from here is arithmetically plausible if FY28-29 estimates land. The constraint is customer concentration, not addressable market.
• Estimate quality caveat: FY27/FY28 rest on 11–13 analysts (solid); FY30 on 1–2 and FY31 on 1–2 — treat the out-years as sketches. The FMP forward EBITDA/EBIT rows are internally inconsistent (FY27 "EBITDA" $126M vs net income $1.14B) and are excluded from our math; we use the revenue and EPS lines only.

Exponential Potential: High (8/10). Big forward growth, small cap, real room — held back from 9 by the negative second derivative and thin out-year coverage.

5. Financials (real numbers — FMP annual/quarterly)

• Revenue: FY26 $1,335.1M, +205.7% (FY25 $436.8M, itself +126.3% on FY24 $193.0M). Two consecutive years of triple-digit growth — the AEC/AI-connectivity ramp landing all at once.
• Quarterly trajectory: Q1 FY26 $223.1M → Q2 $268.0M → Q3 $407.0M → Q4 $437.0M (+157% YoY, +7.4% seq). Note the shape: the explosive sequential jump was Q3 (+52% seq); Q4 and the Q1 FY27 guide ($465–475M, ~+8% seq) show the ramp settling into a steadier climb.
• Margins: gross 68.0% FY26 (TTM identical), operating 33.3%, net 35.4%, EBITDA margin 38.2% TTM. FY24 was still an operating loss (−$37.1M) — the entire margin structure is two years old. Management guides Q1 FY27 non-GAAP GM to 67–69%, i.e. holding.
• Earnings: FY26 net income $472.3M / diluted EPS $2.51 (FY25: $52.2M / $0.29). Tax rate ~0.7% (Cayman) — flagged: any future change in tax status is pure downside to EPS. Income quality is high (0.98) — unlike many hyper-growers, accounting profit ≈ cash profit.
• Cash flow: OCF $464.3M, capex −$57.3M (4.3% of revenue), FCF $407.0M — real conversion, though FCF yield is only ~0.8% at this price. Working capital absorbed $195M (inventory +$174M, receivables +$71M) — deliberate ramp stocking, but 215 days of inventory is a risk if orders pause.
• Balance sheet: cash + short-term investments $1,443M vs total debt $20.6M → net cash $1,144M; current ratio 10.2×; equity $2.06B. Goodwill + intangibles just $122M (5.3% of assets) from a ~$113M FY26 acquisition. Funding note: FY26 included a ~$743M equity issuance — the fortress was partly bought with shareholder dilution (diluted shares 181M → 188M; Q4 running at 192.7M).
• Beat cadence: five straight EPS beats (last: Q4 FY26 $1.16 vs $1.02 est; prior $1.07 vs $0.91, $0.67 vs $0.49, $0.52 vs $0.36, $0.35 vs $0.27) — a strong but crowd-known pattern.

6. Valuation — priced in or room?

There is no honest way to call CRDO cheap on anything trailing: 104× trailing EPS, 36.2× EV/sales, 94.8× EV/EBITDA, 23.8× book, 121.7× FCF. FMP's letter rating is B (overall 3/5), but its valuation sub-scores are the tell: DCF 1/5, P/E 1/5, P/B 1/5 (the 4-5/5 quality scores — ROE, ROA, debt — carry it). The bull case rests entirely on forward compression: ~44× FY27E ($6.10) → ~30× FY28E ($8.94) → ~26× FY29E ($10.39) → ~21× FY31E ($12.50) — the multiple roughly halves in two years at a flat price if estimates land. A reverse read: $265 requires the market to keep paying ~30× on FY28 power and an 82%-growth year to execute cleanly first. Street targets (context): consensus $269.18, median $270, high $350, low $200 — a tight band whose midpoint sits on the current price; even the Street's own numbers say the next 12 months' return is the earnings, not the multiple. Not a value entry; a full-price entry into a decelerating hyper-grower — which is precisely why we Watch rather than buy at $265.

7. Technicals (from the tech block)

• Trend: unambiguously up. $265.55 sits +21% above the 50-DMA ($218.87) and +67% above the 200-DMA ($159.22); MACD +12.3 (positive).
• Location: −12% off the 52-week high ($302.52) — which is also the max drawdown from peak — and +202% off the 52-week low ($87.81). Today's +9.8% pop ($241.91 → $265.55) re-extends the price action.
• Momentum: RSI(14) 53 — neutral; the RSI has reset even as price stayed high, which is constructive.
• Relative strength (the tell): +197% 12-mo vs SPY +21% and QQQ +31%; +162% 3-mo vs SPY +15% / QQQ +24%. Extreme leadership — and per the one independent KB voice, this exact tape round-tripped ~220→~80 within the year on nothing fundamental. Momentum here is a flow phenomenon, not just a fundamentals signal.
• Read: strong but extended. Nothing about this chart says "broken"; everything about it says "late to chase at +21% above the 50-DMA." The lower-risk zones are the rising 50-DMA (~$219) and, in a real flush, the 200-DMA (~$159) — which is our deep-value trigger.

8. Moat & competitive position

Credo's moat is proprietary SerDes/DSP silicon applied where power efficiency matters most — its AECs displace optical links inside the rack at lower power and cost, and the same SerDes core feeds ICs, chiplets, and a licensing line that seeds the technology across the ecosystem. A 68% gross margin on hardware is the empirical evidence that, today, this is differentiated silicon, not a commodity cable. The limits are equally real: AECs face the copper-vs-optical architecture debate (the "copper-to-optical narrative" the KB voice cites as narrative fuel), better-capitalized competitors (Marvell, Broadcom, Astera Labs) surround every socket, and hyperscaler buyers are famous for dual-sourcing anything that works. The FY26 geographic whipsaw (HK-led → US-led in one year) shows how fast the customer mix can swing. ROIC of 21% says the moat is currently earning its keep.

Peer set (FMP-supplied, market cap): Astera Labs $74B, Ciena $61B, Nokia $68B, Ericsson $36B, HPE $57B, Keysight $55B, Super Micro $17.6B, Sandisk $258B, FIS $21B, Pure Storage $28B. Only ALAB and CIEN are close comps; the most relevant names (Marvell, Broadcom) are absent — judge CRDO against the AI-interconnect cohort, not this list. Against ALAB ($74B) Credo is the smaller, cheaper-on-growth pure play.

9. Management, capital allocation & guidance

• Capital allocation: build-the-warchest mode — FY26 raised ~$743M in equity and spent $113M on an acquisition and $57M capex; buybacks trivial ($19M), no dividend. Raising stock at a premium multiple is rational finance but real dilution; the $1.44B cash pile now funds the ramp without leverage.
• Insider activity: routine and small. COO Lam Yat Tung's 2026-07-01/07-05 Form 4s are F-InKind tax-withholding disposals (~3,180 shares each at $259.09/$241.91; still holds ~2.64M shares). CTO Cheng Chi Fung sold ~1,200 shares total on 2026-06-30 at ~$245–256 (indirect; still holds ~5.94M). No discretionary mass selling — founders remain heavily aligned.
• Management guidance (half-weighted, they talk their book): Q1 FY27 revenue $465–475M (+~111% YoY at midpoint), non-GAAP GM 67–69%, non-GAAP opex $86–90M (CRDO-earnings-2026Q2:57d4107105, :60a53aa6e1, :d69726011c). Ingested from the 8-K earnings release; the full analyst Q&A is not on our FMP plan (same gap flagged in prior notes).

10. Catalysts & what to watch

• Next earnings: 2026-09-02 (Q1 FY27; Street EPS $1.16, revenue ~$470M vs guide $465–475M). Key lines: sequential revenue slope (is ~+8% seq the new normal after Q3 FY26's +52%?), gross margin holding ≥67%, and inventory (does the 215-day position start converting?).
• Customer breadth: any disclosure that spreads revenue across more hyperscalers de-risks the single biggest concern; any hint of a top customer digesting does the opposite.
• Copper-vs-optical architecture news: AEC attach rates in next-gen AI racks — the narrative axis the market trades this stock on.
• Estimate revisions: FY27 consensus ($2.43B / $6.10) has to keep rising for the multiple to hold; watch post-print revisions.
• Export-control / China headlines: HK+China ~34% of ship-to revenue.

Thesis tripwires (what would change the call): sequential revenue flat-or-down for two quarters; gross margin below ~65%; an inventory write-down; a disclosed loss/dual-sourcing of a lead AEC customer; or a break and hold below the 200-DMA on volume.

11. Key risks

• Valuation / de-rating (the dominant risk): 36× sales and 104× trailing earnings with a 3.20 beta — the highest-beta profile we cover. The documented ~220→~80 round trip is the template for what "de-rating" means here.
• Customer concentration: hyperscaler-driven orders; the data file has no customer split (flagged), but the FY25→FY26 geographic whipsaw shows how much single-buyer behavior moves the whole company.
• Deceleration risk: consensus already models +82% → +49% → +26%; if deceleration arrives a year early, both the E and the multiple compress simultaneously.
• Competition / architecture: Marvell, Broadcom, Astera Labs in adjacent sockets; a copper-to-optical (or optical-to-copper) architecture swing can strand a product cycle.
• Inventory: 215 days on hand — built for a ramp; toxic in a pause.
• Geopolitical: HK + China ~34% of ship-to revenue; export-control tightening is a live, uninsurable variable.
• Dilution & stock comp: SBC 13.7% of revenue plus a $743M raise — shareholders are funding the ramp; fine while the stock is loved, painful if it isn't.
• Tax normalization: a ~0.7% Cayman effective tax rate flatters EPS; any structural change is a one-way EPS haircut.
• No expert underwriting: one neutral voice in the KB — no independent conviction to lean on when the price action turns.

12. Verdict, position sizing & monitoring

Watch. Credo is, on the numbers, one of the best businesses the momentum screen has ever surfaced: revenue tripled to $1.34B, 68% gross margin, 21% ROIC, 0.98 income quality, $407M FCF, and $1.14B net cash. If the fundamentals were the whole story, this would be a Buy. They aren't: the stock trades at our base-case fair value (~$270) and the Street's ($269), on 36× sales with a 3.20 beta, a negative growth second-derivative, no independent expert thesis in the KB, and a documented history of violent flow-driven round trips. Paying fair value for the highest-beta, most-concentrated name in the pool is not a Synthos trade — being ready when the price action resets is.

• Sizing: none today. On a reset toward the 50-DMA (~$219) with the growth story intact (tripwires green), a satellite ≤1–2% entry is justified; the 200-DMA (~$159, our tier trigger) is the back-up-the-truck level that likely only arrives in a genuine AI-capex scare.
• Monitoring: re-underwrite on the §10 tripwires; formal re-score each earnings print (next 2026-09-02). This verdict is logged as a tracked Synthos call as of 2026-07-06 at $265.55.
• Single biggest risk: hyperscaler order concentration meeting a 36×-sales, 3.2-beta stock — one digestion quarter and the bear case ($150) is in play fast.

Provenance & disclosures

• Traceability: 5 KB claims total — 1 independent voice (forward_guidance, skill 1.0, neutral, 2026-04-16) and 4 management claims (CRDO_mgmt, skill 0.5, 2026-06-01) — all reconciled to real claim_ids (cited inline). There is no expert-panel conviction coverage; this note is fundamentals-driven by design, and kb_net_conviction is left null rather than invented.
• Data as-of: fundamentals 2026-05-02 (FY26 / Q4 FY26, filed 2026-06-15) · estimates & prices 2026-07-06 · last KB claim 2026-06-01. Forward figures are analyst consensus (FMP), labeled as estimates.
• Estimate caveats: FY30/FY31 consensus rests on 1–2 analysts; the FMP forward EBITDA/EBIT rows are internally inconsistent and were excluded — only the revenue and EPS consensus lines are used.
• Customer-data caveat: the data file contains no customer-level revenue split; hyperscaler concentration is discussed qualitatively and flagged as unquantified.
• Management caveat: CRDO_mgmt guidance is management's own book, half-weighted by design; the full earnings Q&A is not on our FMP plan.
• Peer caveat: the FMP-supplied peer list omits the most relevant comps (Marvell, Broadcom); Astera Labs and Ciena are the usable names in it.
• Not investment advice. Independent research, educational and informational only, never personalized. Hypothetical/forward figures are labeled; the only performance numbers Synthos will headline are the live, real-money Flagship's.
• Version: 2026-07-06. Prior versions available via the deep-dive version dropdown ("based on the info at the time").