Phase Space AI

Valuation

Credo Technology Group [CRDO]

Credo Technology Group Holding Ltd [CRDO] — Valuation Analysis (re-underwrite)

Investing Hub Research | 2026-07-29 | framework v1.5.1 | supersedes archive/CRDO_Valuation_2026-07-29_superseded.md

Price $177.45 (Alpaca SIP, 2026-07-29 close). The prior document used $208.14 (07-27); CRDO has fallen 14.7% in two sessions.

Re-underwrite trigger: CRDO was one of 21 of 84 covered names carrying a terminal EBIT margin below its own trailing actual operating margin — here 17.7pp below. That defect suppressed the name's rank. This document rebuilds the parameter from the company's own demonstrated economics.

---

1. WHAT CHANGED

---

2. DATA VERIFICATION — CRDO IS CLEAN

av_vs_edgar.py (6 quarters, 52/53-week calendar tolerance applied): VERIFIED, disagree=0/6, zero invariant violations.

Spot checks against EDGAR primary (CIK 0001807794), fiscal year ended 2026-05-02 — AV normalises this to 2026-04-30:

Defect found on this name — AV ebit ≠ operatingIncome:

Had ebit been used, CRDO's TTM operating margin would read ~35.6% instead of the true 33.33% — a 2.3pp overstatement. All margins in this document are operatingIncome / totalRevenue.

Splits: none since 2015. Per-share figures are on one basis.

---

3. ACCOUNTING QUALITY — THE AAOI TEST, AND CRDO PASSES IT DECISIVELY

The AAOI precedent in this same sector: a private stocking distributor at 53% of revenue, 71.6% of receivables, ~264-day DSO, named in none of 18 quarterly releases. That is the pattern to look for.

3.1 Receivables and DSO — improving sharply, the opposite of the AAOI tell

DSO has fallen from 147.1 days to 48.7 days while revenue grew 632%. Receivables in absolute dollars are lower than three quarters ago ($233.4m vs $245.2m) against revenue up 63%. A receivable growing slower than revenue is the inverse of the single most reliable tell in the brief. This is high-quality, cash-converting growth.

Inventory days rose from ~147 to 164.7 — a modest build, consistent with staging for a ramp against 205.7% revenue growth, and the one item to keep monitoring.

3.2 Is the growth organic, recurring, repeatable?

Verdict: PASS. The growth is organic, cash-converting and driven by operating leverage on a stable gross margin.

3.3 Customer concentration — extremely high, but fully and voluntarily disclosed

CRDO discloses concentration on two bases, which is the opposite of AAOI's concealment.

(a) Contracting-party basis (audited financial-statement footnote):

(b) End-customer basis — a supplemental, voluntary disclosure, explicitly given "to provide further insight into our end customer concentration, rather than based on the contracting parties":

Also disclosed: top 10 customers ≈ 90% of FY2026 revenue; two customers >10%. Amazon is named in the 10-K via the Customer Warrant — so at least one hyperscaler counterparty is identified, not merely implied.

The AAOI comparison, made numerically:

CRDO does not replicate the AAOI pathology. The concentration is genuinely severe and is the name's principal fundamental risk, but it is disclosed, the receivables are collected in seven weeks, and the largest customer's AR share is falling faster than its revenue share. Management states: "We expect that as our products are more widely adopted and as our number of customers increase, customer concentration will decrease" — and FY2026 is the first year that actually happened (A: 67%→49%).

3.4 Design wins — implied, not contracted. Stated explicitly.

The 10-K is unusually candid, and this answers the question directly:

"We consider a design win to occur when a customer notifies us that it has selected our products or technology… While not legally enforceable contractual obligations, we believe design wins are an important step towards the adoption of our products."

"qualification of a product by a customer does not assure any sales of the product to that customer… despite achieving a design win, the customer may determine not to proceed with a contemplated project and cancel the project with little notice."

Conclusion: CRDO's design wins carry NO contractual force. This is a material and instructive contrast with MU, whose SCAs are take-or-pay with binding volume commitments and price floors. Two names in the same AI-infrastructure supply chain, re-underwritten together, with opposite revenue- security structures — CRDO's backlog is a forecast, MU's is an obligation. CRDO carries no disclosed RPO of consequence.

3.5 Mention-frequency over time (required core metric)

Prepared remarks / Q&A, 7 quarters:

Reading: aec is the dominant topic in every single quarter and is analyst-pulled, not management-pushed — Q&A mentions exceed prepared mentions in 6 of 7 quarters (12–23 in Q&A vs 2–10 prepared). That is the stronger form of evidence: the market is interrogating AEC rather than management promoting it. pcie spiked in 25Q3 (13/13) as the PCIe retimer line was introduced and has since normalised. 1.6t re-emerges in 26Q3 (1/5, analyst-led) — the next speed-node transition beginning to be priced. Product mix is broadening (AEC + SerDes + optical DSP + PCIe retimers), which is the mechanism by which customer concentration falls.

---

4. COMPANY STATE

STATE C — scaling but economically observable.

Test evidence: - Positive and stable gross margin: 57.7% → 68.0% across four fiscal years, tight band of 63–68% for the last eight quarters. Not a State-D "unstable or deeply negative" profile. - Visible expense scaling: R&D 20.9% of revenue and SG&A 13.8% in FY2026, both falling as revenue scales, with absolute dollars rising in a controlled way. - Now decisively profitable: FY2026 operating income $445.0m (33.33% margin), net income $472.3m, diluted EPS $2.51.

Why not State A: history is far too short and the growth rate (205.7% in FY2026) is nowhere near "low-variance across ≥5 years." Why not State B: margin is not driven by an exogenous commodity cycle — gross margin has been flat while revenue tripled, which is the opposite of cyclical price-driven margin. Why not State D: the economics are fully observable and profitable.

Consequence: the terminal margin must be BUILT via the opex bridge, and the instrument is the two-dimensional expectations surface.

---

5. TERMINAL MARGIN — THE PARAMETER THAT WAS WRONG

5.1 Demonstrated record

Quarterly operating-margin path: −24.2% → −11.7% → +19.4% → +20.4% → +27.2% → +29.4% → +36.8% → +35.7%. Eight consecutive quarters of expansion on a flat gross margin.

5.2 The defect, quantified

The clamp is the identical bug that cut AVGO's real 43.4% to 15.6%: max(own, industry_median) then capped at an industry 75th percentile. A cap can only ever flatter a weak operator and penalise an exceptional one — and it fired on a business whose demonstrated operating margin was already double the cap.

Effect on the answer, holding everything else fixed (exit 15.0x, WACC 10%, 5 years):

The clamp alone inflated the required CAGR by 27.3pp — from 48.8% to 76.1%. That is the magnitude by which this single parameter suppressed the name's rank.

5.3 The opex bridge — m_EBIT,T = m_gross,T − R&D − SG&A − other

Hard constraint satisfied: 38.0% ≤ 67.0%. ✓ Large headroom — 29pp of revenue remains for all operating cost, so the bridge is not merely passing the literal ceiling test but is economically comfortable.

Bridge validated against actuals: FY2026 → 68.0% − 20.9% − 13.8% = 33.3% vs actual 33.33%. Residual 0.03pp. The construction reproduces the reported margin essentially exactly.

The terminal margin is ABOVE the trailing actual (+4.7pp). Justification: this is a State-C business in the middle of a demonstrated operating-leverage ramp — the latest quarter is already 35.71%, i.e. 2.4pp above the FY average, and the trend has been monotone for eight quarters on a flat gross margin. Terminalising at 38.0% extends a demonstrated trajectory by 2.3pp beyond the most recent observed quarter. My bridge would support 40.0% (67 − 17 − 10); I use 38.0% as the more conservative point. This is not peak-extrapolation — it is the mid-point of a still-improving, mechanically understood series.

---

6. NET CASH — REBUILT

EV is LEASE-INCLUSIVE. CRDO has no financial debt whatsoever — the only debt-like items are operating leases.

Prior memo: $673.3m. Understated by $744.5m — a factor of 2.11x. The likely cause is omission of short-term investments and use of a stale period; the prior figure corresponds to neither the FY2026 nor the FY2025 balance sheet as filed.

Diluted shares: FY2026 weighted-average diluted was 188.232m, but that averages a year over which the count grew. Current basis: 192.681m basic outstanding (10-K) + 9.694m dilutive share-based awards = 202.4m. The prior memo's 192.0m used basic outstanding only and understated diluted shares by 5.4%.

---

7. THE IMPLIED-PATH TEST — THE VALUATION CRITERIA

Fixed inputs, all named: spot $177.45 · diluted shares 202.4m · net cash +$1,417.8m · TTM revenue $1,335.1m · horizon 5 years · WACC 10.0% · terminal EBIT margin 38.0% · exit multiple 15.0x EBIT · tax 15% (Cayman domicile, low effective rate).

EV implied by today's price: $34,498m (25.8x TTM revenue).

7.1 Exit multiple — derived, not asserted

EV_T/EBIT_T = (1−t)(1−g/ROIC)/(WACC−g):

Base 15.0x. The prior memo's 27.0x sits above every point in this range and would require g≈6.5%+ with a sub-9.5% WACC — it was asserted from traded comparables rather than derived. Note valuation.md rule 4: a high terminal growth rate, a high margin, low reinvestment and a high exit multiple double-count the same quality.

7.2 Result

VALUATION CRITERIA: PASS, +44.8pp.

The price requires a 48.8% five-year revenue CAGR. CRDO has delivered 93.6% over three years and 205.7% in the last twelve months. The price does not require acceleration — it requires roughly half the growth rate the company has already demonstrated.

Implied FY2031 revenue at 48.8%: ~$9.7bn, from $1.335bn. That is a demanding but not absurd endpoint — consensus already has FY2028 at $3.68bn, which is ahead of the required path's FY2028 point.

Consensus corroboration (AV EARNINGS_ESTIMATES):

Consensus FY2028 growth of 49.8% is almost exactly the required 48.8% CAGR — the Street's two-year-out estimate independently brackets the price-implied path. Revisions are net upward on both years.

7.3 Sensitivity over the exit multiple

The name passes at every exit multiple in the range, including the most punitive 10.0x. The conclusion is robust to the parameter that carries the most variance — which is the strongest form this test can take.

---

8. TWELVE-MONTH TARGET

8.1 Multiple anchor — own history, admissible

CRDO's own P/S history, monthly from Alpaca SIP closes against TTM revenue, n=55 months from 2022-01:

This series IS admissible — unlike MU's, it does not span a regime change. Gross margin has been 57.7–68.0% throughout, a 10pp band, so a sales multiple is comparable across the window. n=55 monthly observations with real dispersion (6.69x–57.70x).

8.2 The target

In twelve months (July 2027), CRDO's trailing-TTM revenue will be approximately FY2027 (fiscal year ends May 2027) = consensus $2,455.1m. Applying the own-history median trailing P/S to the revenue that will then be trailing is dimensionally correct:

12-MONTH TARGET: $268 — +51.1% ABOVE SPOT.

Basis: own-history median trailing P/S of 22.11x (current 25.61x = 65th percentile) applied to FY2027E consensus revenue of $2,455.1m. Note this uses the median, i.e. it assumes multiple compression from today's 65th percentile down to the 50th — the target's upside comes entirely from the 83.9% consensus revenue growth, not from re-rating.

Named events inside 12 months: 1. FQ1 FY2027 results (~early Sept 2026) — consensus revenue $387.6m / EPS $0.94. 2. FQ2 FY2027 (~Dec 2026) — consensus $433.3m / EPS $1.03. 3. 1.6T / next-node AEC and optical DSP ramp — the 26Q3 re-emergence of 1.6t in analyst questions (1/5) marks the transition beginning. 4. Customer-concentration trajectory — whether Customer A falls below 40% of revenue, continuing the 67%→49% path. This is the key de-risking milestone. 5. FY2027 10-K (~June 2027) — updated concentration and AR tables.

---

9. DOWNSIDE CASE WITH A NAMED CAUSE (MEASURED)

Named cause: Customer A (49% of revenue) in-sources its AEC/retimer content or dual-sources it to a larger merchant supplier, with no contractual protection to prevent it.

This is the specific, evidenced risk — and §3.4 establishes there is no contractual barrier: design wins are "not legally enforceable contractual obligations," and the 10-K warns a customer "may determine not to proceed with a contemplated project and cancel the project with little notice." CRDO's competitors are "substantially larger" with "greater financial, technical, marketing, distribution" resources — Marvell and Broadcom both compete directly in SerDes and AEC.

Mechanism and quantification: loss of half of Customer A's volume (~25% of revenue) with the remaining book growing 40% would produce roughly flat revenue for a year. Because opex is largely fixed and R&D is 20.9% of revenue, operating margin compresses violently — the FY2025 comparison is instructive: at $436.8m of revenue the operating margin was 8.70%, not 33%. On a flat-revenue, margin-compressed year at ~15% operating margin and a p25 exit of 10x, the required CAGR framing inverts and the equity is worth roughly $85–100/share, −45% to −52% vs spot.

Mitigants, stated fairly: DSO of 48.7 days means little cash is at risk in a rupture; net cash of $1,417.8m (8.0% of market cap) funds a transition; product breadth is increasing; and Customer A's share is already falling (67%→49%).

Type: MEASURED. Logged and scored; it does not reject the name.

---

10. CRITERIA SUMMARY

Evidence grade: A−. Accounting quality and disclosure are grade A (EDGAR-exact, dual-basis concentration, improving DSO). The terminal margin is grade A− (bridged from verified actuals, reproduces the reported margin to 0.03pp, extends a demonstrated trend by 2.3pp). The 12-month multiple anchor is grade B (own-history series admissible, n=55, but only 55 months long).

No position verdict is issued. The book decides. The concentration level — one customer at 49% of revenue with no contractual protection — is a position-size constraint, not an admission constraint (valuation.md rule 5).

---

Prepared 2026-07-29. Data: Alpha Vantage normalized statements (24 quarterly / 7 annual periods) cross-verified against SEC EDGAR companyfacts and the FY2026 10-K (CIK 0001807794, filed 2026-06-15); Alpaca SIP prices; 7 quarters of earnings-call transcripts with speaker/title tags. Publication out of scope this run.