Credo created the market for active electrical cables — the copper links between AI servers and their network — tripled revenue last year and keeps 68 cents of every sales dollar after making the product. Charge the roughly $450 million a year it pays in stock and the 15% minimum tax its own 10-K expects, and the cash flows are worth $113–169 a share. After a 29% rise in seven sessions, with no new guidance, no new customer and the same uncontracted second half still to deliver, $194 needs either that neither cost is real or that the bull case is the central one. That is a strong business, overvalued. Our band is $113–169.
The disagreement with the market is two costs, and at this price it takes both of them. On one revenue path — fiscal 2028 at the $3.65bn consensus, fiscal 2031 at $5.5bn, 10% above management’s own strategic plan — the model returns $135 with share-based pay and a 15% tax charged, $161 with share-based pay waived, $157 with the tax waived and $187 with both waived. At $150 either waiver closed the gap to the price; at $194 even both together leave it 4% short. Charged in full, the price needs the bull case’s revenue path — fiscal 2031 near $7.9bn — or an 8.0% discount rate. Sell-side targets of $245–310 are 31–39× fiscal 2028 non-GAAP earnings; the same earnings after the two charges are about $6.24 a share, and the shares trade at 31× that.
Fiscal 2027 is 19% delivered, 21% guided and 59% required. First-quarter revenue of $479m and a $525–535m second-quarter guide leave $1.46bn to be delivered in the second half for the “more than 85%” growth outlook — a 45% step up half on half, carried by an optical business that must go from roughly $0.15–0.20bn in the first half to $0.40–0.45bn in the second. Contracted remaining performance obligations are $4.2m. The company has beaten its own guide in every quarter since the ramp began; that record, not a backlog, is the evidence.
Concentration has improved from extreme to merely high, and the direction is right. The largest end customer was 86% of revenue in the January 2025 quarter; it is 33% now, with four end customers at or above 10% (33 / 28 / 13 / 10%) and five hyperscalers engaged. On the contracting-party basis the filings use, one customer is still 43% of revenue and 57% of receivables. The bear case — 30% weight, $62 — is a pause at one of those accounts in fiscal 2028, and it is why the probability-weighted value of $136 barely exceeds the base despite two upside tails.
Run the model in Section 02 backwards with every cost charged and $194.35 needs one of three things. Every year’s growth increment 1.76 times the base — fiscal 2031 revenue of about $7.9bn, above the $7.5bn top revenue hurdle of the chief executive’s own special award, and fiscal 2036 near $11.1bn. Or the base path discounted at 8.0% instead of 10.0%, below the rate we would use for Broadcom. Or gross margin 17 points higher in every year, which no business adding module revenue earns. Those are three ways of saying the same thing: that the fiscal 2028 product cycle — 1.6T cables, silicon-photonics chips, near-package optics, Active LED Cables, OmniConnect — delivers in full, with no digestion year and no customer pause. That is close to our bull case of $207, which we weight at 20%, with a 5% super-bull tail above it.
The other route to the price is to stop charging the two costs. With share-based pay treated as free and the tax rate held near today’s 1.5%, the base path is worth $187, within 4% of the price. That is how the market quotes this company — on non-GAAP earnings that leave out the first and assume almost none of the second — and it is why the gap between our value and the price is mostly two lines of the income statement rather than a different view of the business. We charge both because both are real: share-based pay has grown the share count 9.8% in fifteen months, and the company’s own 10-K expects the 15% minimum tax from fiscal 2028. At $150.39 on 15 September either waiver alone closed the gap; at $194 it takes both, and a little more.
What separates those two prices is a 29% move in seven sessions on a sector-wide rally, not a change in the business: no new guidance, no new customer, the same $1.46bn second half still to be delivered on $4.2m of contracted backlog.
The “more than 85%” growth outlook implies fiscal 2027 revenue of about $2.47bn. Delivered, guided and still to be earned, drawn to scale:
The second half implies a third quarter near $690m and a fourth near $770m, against a prior best quarter-on-quarter step of +52% in the January 2026 quarter and +7–10% in the three quarters since. Remaining performance obligations at 1 August 2026 were $4.2m, all due within twelve months. Management refers to “unfulfilled backlog” when explaining inventory but discloses no number, and the company does not report revenue by product line, so the optical ramp — from roughly $0.15–0.20bn in the first half to $0.40–0.45bn in the second — is inferred from the $600m-plus optical target.
The valuation reduces to whether fiscal 2028 is another “outsized” year, as management says, with no customer pause — the consensus $3.65bn, up 47%. The dated tests are the hyperscalers’ 2027 capital-expenditure commentary in late October; the second-quarter print around 1 December, where the third-quarter guide needs to be near $690m for the year to land; and the fiscal 2028 outlook around 1 June 2027. At $194 the price leaves nothing for any of the three to disappoint.
At $194.35 the shares trade above every cell of the sensitivity grid — 15% above the band’s top, 44% above the base case and 42% above the probability-weighted value — so the verdict needs no boundary argument. On peer multiples the shares are merely fair: 22–28× fiscal 2028 non-GAAP earnings gives $174–221, which brackets the price. On their own cash flows they are not, and when the two disagree the rating follows the cash flows. The shares are 37% below the 22 June peak of $308.67; that drawdown is a statement about June, not about the price.
FY2026–36 revenue CAGR 18.7% · FY2036 operating margin after share-based pay 43.0% · FY2036 free cash flow $2.6B · terminal value is 60% of enterprise value.
This is an unlevered discounted cash flow on a ten-year explicit path, fiscal 2027 to 2036, with share-based compensation charged as a cost. Free cash flow to the firm is non-GAAP operating profit less share-based pay, after tax, plus depreciation and amortisation at 2% of revenue, less capital expenditure at 3%, less working capital at 25% of each year’s incremental revenue. Fiscal 2027 also carries a $100m outflow for refundable capacity deposits — $102.4m was committed after the quarter — which recycle through purchases thereafter. The valuation date is 24 September 2026: the first quarter’s $83m of free cash flow is already in the cash balance and is netted out of the year, the remaining three quarters are discounted at 0.4 years, and each later year at its mid-point. Terminal value is a Gordon perpetuity at 3% on fiscal 2036 cash flow; on base settings it is 60% of enterprise value and 14.7 times fiscal 2036 free cash flow. Amortisation of acquired intangibles is left out: it is not cash, and its tax shield is of no consequence.
The four revenue settings are the thesis. Fiscal 2027 at $2.48bn is the guide taken at face value, because it has been beaten every quarter. Fiscal 2028 at $3.65bn is the post-print consensus. Fiscal 2031 at $5.53bn is 10% above the $5.0bn strategic plan the board disclosed with the chief executive’s special award in June; fiscal 2036 at $7.39bn is growth fading from 8% to 4%, an 18.7% ten-year compound rate from fiscal 2026. Between the settings the path keeps the shape of whichever scenario is loaded, so each card reproduces its published value exactly and the sliders bend it from there. Moving fiscal 2031 alone by $1bn is worth about $5 a share; moving fiscal 2036 by $1bn, about $12. The endpoints matter more than the near years, because the terminal value is struck on the last one.
Three cost settings, each a judgement the market makes differently. Terminal gross margin of 65%, from 68% today, prices the shift toward optical modules and photonic chips, which earn less than the cables; the setting phases in over five years and leaves fiscal 2027 at the guide. Terminal operating expenses of 16% of revenue sit just below today’s 17%. The share-based pay setting scales the whole path in proportion: the base runs from 18% of revenue this year (about $446m) through 14%, 12% and 10% to 6% terminal (about $443m), so the dollar charge stays near $450m a year for a decade. Set it to zero and the model gives $161 — the non-GAAP convention. Each point of terminal share-based pay is worth about $4 a share, and each point of terminal gross margin or operating expense about $3.
Tax is 1.5% in fiscal 2027 and 15% from fiscal 2028 in every scenario, because the company’s 10-K says it expects the OECD’s 15% global minimum tax to apply from that year; the effective rate was 0.7% in fiscal 2026 through a Cayman Islands holding structure. Holding the rate at 1.5% gives $157; holding it there and treating share-based pay as free gives $187. The 10% discount rate is an unlevered, net-cash rate. The shares’ regression beta of about 3.2 is a volatility artefact; a fundamental beta of 1.3–1.5 with a 4.3% risk-free rate and a 4–4.5% equity premium gives 9.5–11%. The rate sits between the roughly 9% appropriate to Marvell or Broadcom and the 11–12% we use for speculative small caps, and it is the same in every scenario, so concentration and cycle risk are carried once, in the cash flows and the weights. Each 25 basis points is worth about $5 a share.
Share count and net cash. 191.2m shares: 187.95m basic at 25 August 2026 (10-Q cover), 1.68m options by the treasury method at a $2.34 average strike, and 1.54m for the $300m DustPhotonics earn-out, which is settled in shares and counted at the current price. Unvested RSUs (8.3m) and PSUs (3.1m) are not added, because their cost is already in the share-based pay line; adding them as well would count the same dilution twice. Net cash $754m: $764.3m of cash and investments at 1 August 2026, less a $10m contingent liability; no debt.
The number above is not a fact about Credo. It is what these assumptions produce, and the two that decide it are not about the business at all: whether share-based pay is a cost, and whether the minimum tax arrives. The third is a judgement about the industry — whether the hyperscalers keep raising connectivity spend through fiscal 2029–2030 without a pause. The rating survives every cell of the sensitivity grid and either cost waived on its own. It does not survive both waived together with a slightly better revenue path, or the bull case becoming the central one — which is the choice the price asks you to make.
Each card loads its published path into the model above and prints that path’s discounted value: $62, $135, $207 and $314. These are the published scenario values; the page model is the model that produced them. All four use the same 10% rate and 3% terminal growth, so the differences between them are operating paths only. The bear carries a 30% weight because the base case credits fiscal 2028 consensus and a second-half ramp that has no contracted backlog behind it; when a base case leans on uncontracted growth, its failure branch has to carry real mass. Inside the bear sits a harsher sub-case — the largest end customer exiting in fiscal 2028, taking revenue to $2.0bn that year — worth $55. Weighting the four at 30 / 45 / 20 / 5 gives $136, barely above the base: the bear’s mass offsets two upside tails.
| Assumption | Bear | Base | Bull | Super-bull |
|---|---|---|---|---|
| Revenue FY2027 / 28 / 29 | $2.35 / 2.70 / 2.90bn | $2.48 / 3.65 / 4.45bn | $2.60 / 4.00 / 5.40bn | $2.65 / 4.40 / 6.60bn |
| Revenue FY2030 / 31 | $2.75 / 3.00bn | $5.07 / 5.53bn | $6.50 / 7.50bn | $8.80 / 10.80bn |
| Revenue growth FY2032–36 | 6 / 5 / 4 / 3 / 3% | 8 / 7 / 6 / 5 / 4% | 12 / 10 / 8 / 6 / 5% | 16 / 11 / 8 / 6 / 5% |
| Revenue FY2036 | $3.69bn | $7.39bn | $11.0bn | $16.7bn |
| Non-GAAP gross margin | 67% → 63% | 68% → 65% by FY2032 | 68.5% → 66% | 68.5% → 67% |
| Non-GAAP operating expenses, % of revenue | 17.5 → 21 → 19% | 17 → 15 → 16% | 16.5 → 13 → 15% | 16.5 → 12 → 14.5% |
| Share-based pay, % of revenue | 18.5% → 7% | 18% → 6% | 17.5% → 5% | 17.5% → 5% |
| Terminal operating margin after share-based pay | 37.0% | 43.0% | 46.0% | 47.5% |
| Tax rate | 1.5% in fiscal 2027; 15% from fiscal 2028 in every scenario (OECD Pillar Two, per the 10-K) | |||
| Capex / depreciation and amortisation | 3.0% / 2.0% of revenue. Fiscal 2026 capex was 4.3% including mask sets; the first quarter of fiscal 2027 was 1.5% | |||
| Working capital | 25% of incremental revenue. Fiscal 2026 actual 26%; receivables about 55 days, inventory about 168 days, payables about 55 days | |||
| Capacity deposits | $100m extra outflow in fiscal 2027; none thereafter, as refundable deposits recycle through purchases | |||
| Discount rate / terminal growth | 10.0% / 3.0% | 10.0% / 3.0% | 10.0% / 3.0% | 10.0% / 3.0% |
| Discounted value per share | $62 | $135 | $207 | $314 |
| Weight | 30% | 45% | 20% | 5% |
| $m, FY to ~1 May | FY27 | FY28 | FY29 | FY30 | FY31 | FY32 | FY33 | FY34 | FY35 | FY36 |
|---|---|---|---|---|---|---|---|---|---|---|
| Revenue | 2,480 | 3,650 | 4,450 | 5,070 | 5,530 | 5,970 | 6,390 | 6,770 | 7,110 | 7,390 |
| Growth | +86% | +47% | +22% | +14% | +9% | +8% | +7% | +6% | +5% | +4% |
| Gross profit (non-GAAP) | 1,686 | 2,446 | 2,959 | 3,346 | 3,622 | 3,880 | 4,154 | 4,400 | 4,622 | 4,804 |
| Operating income (non-GAAP) | 1,265 | 1,880 | 2,292 | 2,586 | 2,765 | 2,925 | 3,131 | 3,317 | 3,484 | 3,621 |
| margin | 51.0% | 51.5% | 51.5% | 51.0% | 50.0% | 49.0% | 49.0% | 49.0% | 49.0% | 49.0% |
| Share-based compensation | (446) | (511) | (534) | (507) | (498) | (478) | (479) | (474) | (462) | (443) |
| Operating income after share-based pay | 818 | 1,369 | 1,758 | 2,079 | 2,267 | 2,448 | 2,652 | 2,843 | 3,022 | 3,178 |
| margin | 33.0% | 37.5% | 39.5% | 41.0% | 41.0% | 41.0% | 41.5% | 42.0% | 42.5% | 43.0% |
| After-tax operating profit | 806 | 1,163 | 1,494 | 1,767 | 1,927 | 2,081 | 2,254 | 2,417 | 2,568 | 2,701 |
| Unlevered free cash flow | 395 | 834 | 1,250 | 1,561 | 1,757 | 1,911 | 2,085 | 2,254 | 2,412 | 2,557 |
| FCF margin | 15.9% | 22.9% | 28.1% | 30.8% | 31.8% | 32.0% | 32.6% | 33.3% | 33.9% | 34.6% |
Valuation date 24 September 2026. Fiscal 2027’s first-quarter free cash flow of $83m is already in the cash balance and is netted out of the year’s $395m. Present value of fiscal 2027–36 flows $10.1bn; terminal value $15.0bn, 60% of enterprise value and 14.7 times fiscal 2036 free cash flow; enterprise value $25.1bn; plus $0.75bn of net cash, equity $25.8bn; $135 a share on 191.2m shares.
The base sits above management’s own plan from fiscal 2029, and that is deliberate. The board’s strategic plan, disclosed with the chief executive’s special award in June 2026, reaches $5.0bn of revenue in fiscal 2031. Run on the plan exactly — fiscal 2029 $4.3bn, 2030 $4.7bn, 2031 $5.0bn — the model gives $121. We used consensus and a smoother deceleration because the plan was set before the latest quarter’s beat, because compensation-plan targets are set to be reached, and because this management finished fiscal 2026 about 60% above where the year’s consensus began. The counter-argument is on the record too: consensus and the plan cannot both be right, and one of them implies growth falling to about 11% a year immediately after fiscal 2028.
Share-based pay is charged at its accounting cost, and that cost is sticky in dollars. It was $183m in fiscal 2026, $88m in the first quarter of fiscal 2027 alone, and is guided to about $104m in the second; unrecognised expense is about $1.27bn. The base lets it fall from 18% of revenue to 6%, which holds the dollar charge near $450m a year rather than letting it grow with the company. Grants made at $155–227 in the June quarter are expensed at those values whatever the share price does afterwards, which is one reason the known weaknesses below show the halved path as well.
The bear is a customer event, not a cycle alone, and the bull is anchored to a document. Four end customers are 84% of revenue and there is no contracted backlog, so the failure branch is a pause at one of them in the year consensus has at +47%; a fiscal 2030 digestion year then follows the pattern of every capex-driven semiconductor cycle. The bull takes fiscal 2031 to $7.5bn because that is the top revenue hurdle in the chief executive’s own award — the range of outcomes the board considered plausible when it set the pay.
The grid holds the base operating path fixed and varies the discount rate against terminal growth. Every cell is the published fair-value band, $113–169.
| WACC ↓ / terminal growth → | 2.5% | 3.0% | 3.5% |
|---|---|---|---|
| 9.0% | $151 | $159 | $169 |
| 9.5% | $139 | $146 | $154 |
| 10.0% base | $129 | $135 | $141 |
| 10.5% | $121 | $125 | $131 |
| 11.0% | $113 | $117 | $122 |
Every cell of the grid is below the $194.35 price; the most generous, 9.0% and 3.5%, is 13% below it. The grid is reproducible from the model above by moving the last two sliders.
Modal base case $135. Fair-value band $113–169, the WACC 9–11% by growth 2.5–3.5% grid, which is what anchors the rating; the price is 15% above its top. Probability-weighted value $136 across the four scenarios at our own weights. The two anchors agree, and the price is more than 40% above each.
| Hold everything else at base and change | What the price requires |
|---|---|
| Revenue path | Every year’s growth increment 1.76× the base: fiscal 2031 $7.9bn, 57% above the strategic plan and above the $7.5bn top hurdle of the CEO award; fiscal 2036 $11.1bn; a 23.6% ten-year compound rate. Our bull case |
| Discount rate | 8.0% instead of 10.0% — below what we would use for Broadcom, for a company with four customers and purchase-order revenue |
| Share-based pay | Treat all of it as free, the non-GAAP convention: $161. Not enough on its own; with a 9% rate as well, $189 |
| Tax | Effective rate stays near 1.5% instead of stepping to 15% in fiscal 2028: $157 |
| Both conventions | Share-based pay free and tax held at 1.5%: $187, within 4% of the price. This is the closest single reading to how the market quotes the stock |
| Margins | Gross margin 17 points higher in every year (terminal 82%; operating margin after share-based pay 60%) — not a plausible path for a business adding module revenue |
At $150 on 15 September the market was pricing this company on non-GAAP earnings at a large-cap discount rate, and either convention alone closed the gap to the model. After the rally neither does, and both together get within 4%. So $194 requires one of two beliefs: that share-based pay and the coming tax are not costs to shareholders, or — with both charged — that fiscal 2028–2031 go right in full, with no digestion year and no customer pause, which is our bull case as the central expectation or an 8% discount rate. We think both costs are real, and the price pays nothing for the possibility that the second belief is wrong. The disagreement is more than 40% against both anchors and is not bounded by the band. That is an overvalued verdict.
| Risk | Priced in | How |
|---|---|---|
| Customer concentration / order cancellation | Scenario weights | Bear (30%) is a one-customer pause; the customer-loss sub-case ($55) sits inside it. Not in the discount rate, not in an exit multiple |
| Second-half fiscal 2027 shortfall | Bear cash flows | Bear fiscal 2027 $2.35bn; the base takes the guide at face value because the guide has been beaten every quarter |
| Capex-cycle digestion | Cash flows (both) | Base decelerates to +14% in fiscal 2030 and +9% in 2031; the bear has an outright −5% year. Not also in the discount rate |
| Optical mix dilutes gross margin | Base cash flows | 68% to 65% by fiscal 2032; bear 63% |
| Dilution / share-based pay | Cash flows | Charged as a cost every year; unvested awards therefore excluded from the share count, so dilution is not counted twice |
| Pillar Two tax | Cash flows | 15% from fiscal 2028 in all scenarios |
| Competition / price erosion at 1.6T | Cash flows and weights | Base fades non-GAAP operating margin from 51.5% to 49%; bear to 44%. Partly in the bear weight |
| Systematic risk, size, volatility | WACC | 10.0% — the only thing the discount rate carries |
| Upside: Active LED Cables, OmniConnect, near-package optics, module premium | Bull and super-bull weights | Zero in the base; held to the same evidence standard as forward-looking risk |
Against its two direct connectivity peers Credo is the cheapest name on every forward multiple, and it has the widest gap in the group between the profit it is quoted on and the profit it reports. All prices are 24 September 2026 intraday quotes, taken within four minutes of each other.
| Metric | Credo | Marvell | Astera Labs | Broadcom | MACOM | Lumentum |
|---|---|---|---|---|---|---|
| Share price 24 Sep 2026 intraday | $194.35 | $260.49 | $365.23 | $350.81 | $278.00 | $934.14 |
| Market cap · EV | $36.5bn · $35.8bn | $228bn · $231bn | $63.4bn · $62.1bn | $1,674bn · $1,711bn | $21.3bn · $21.0bn | $84bn · ≈$84bn |
| Latest quarter · revenue growth y/y | Jul-26 · +115% | Jul-26 · +37% | Jun-26 · +104% | Jul-26 · +86% | Jun-26 · +36% | Jun-26 · +109% |
| Next quarter guided growth y/y | +98% | >+50% | ≈+140% | +93% | ≈+60% | >+130% |
| Non-GAAP gross margin | 68.0% | 58.9% | 73.7% | 75.0% | 59.7% | 50.4% |
| Non-GAAP operating margin | 48.2% | 36.6% | 39.1% | 67.9% | 31.5% | 36.6% |
| GAAP operating margin | 25.2% | 16.8% | 22.7% | 53.9% | 22.5% | 27.8% |
| Gap: non-GAAP less GAAP points | 23.0 | 19.8 | 16.4 | 14.0 | 9.0 | 8.8 |
| Share-based pay / revenue latest quarter | 18.4% | 11.9% | 16.3% | 6.8% | 7.2% | n/d |
| Free-cash-flow margin | 30% (FY26) | 19% (H1 FY27) | 19% (H1 26) | 46% (Q3 FY26) | 17% (9M FY26) | 10% (FY26) |
| Revenue — current FY · next FY | $2.48bn · $3.65bn | $12.0bn · $18.0bn | ≈$1.87bn · ≈$2.7bn | $105.9bn · ≈$165bn | $1.32bn · ≈$1.7bn | ≈$6.5bn · — |
| Non-GAAP EPS — current FY · next FY | $5.10 · $7.90 | ≈$4.15 · ≈$6.80 | ≈$3.90 · ≈$5.80 | ≈$11.60 · $21.10 | $5.51 · ≈$8.80 | ≈$21.00 · — |
| EV / revenue current · next FY | 14.4× · 9.8× | 19.3× · 12.9× | 33.2× · 23.0× | 16.2× · 10.4× | 15.9× · 12.3× | 12.9× · — |
| P/E non-GAAP current · next FY | 38.1× · 24.6× | 63× · 38× | 94× · 63× | 30× · 17× | 50× · 32× | 44× · — |
| Next-FY EPS growth · PEG | +55% · 0.45 | +64% · 0.60 | +49% · 1.29 | +82% · 0.20 | +60% · 0.53 | ≈+150% · 0.30* |
Prices are 24 September 2026 intraday quotes taken between 13:08 and 13:12 New York time. Between the 15 and 22 September closes the group re-rated sharply — Credo +28%, Astera +44%, Marvell +18%, MACOM +17%, Lumentum +13%, Broadcom +7% — on a broad AI-infrastructure rally, with no change to any company’s guidance. Fiscal years: Credo April (FY27/FY28), Marvell January (FY27/FY28), Astera calendar (2026/2027), Broadcom October–November (FY26/FY27), MACOM September–October (FY26/FY27), Lumentum June (FY27 is its current year; no next-year basis shown). Operating figures are from each company’s latest results release: Credo 1 Sep 2026 (Q1 FY27); Marvell 27 Aug 2026 (Q2 FY27); Astera Labs 4 Aug 2026 (Q2 2026); Broadcom 2 Sep 2026 (Q3 FY26); MACOM 6 Aug 2026 (Q3 FY26); Lumentum 11 Aug 2026 (Q4 FY26). Figures marked ≈ are derived from company guidance and outlooks rather than a published consensus: Marvell’s revenue is its own ~$12bn / ~$18bn outlook; Astera adds one modelled quarter to first-half actuals and the third-quarter guide, with 2027 at aggregator growth rates; Broadcom fiscal 2026 is nine months reported plus the fourth-quarter guide, fiscal 2027 revenue its $115bn AI outlook plus run-rate non-AI and software, and fiscal 2027 EPS a post-results broker estimate ($21.12); MACOM fiscal 2026 is nine months plus the fourth-quarter guide, fiscal 2027 at management’s “mid-20%” growth; Lumentum fiscal 2027 assumes the guided $1.25bn first quarter rising to management’s ~$2bn quarterly target by the fourth. Credo consensus about $2.48bn / $5.10 and $3.65bn / $7.90. Marvell’s enterprise value treats its $2bn Series A convertible preferred as debt-like; Lumentum’s is shown at market capitalisation because its remaining convertible debt was not verified. *Lumentum’s growth and PEG are current-year.
Against its two direct connectivity peers, Credo is the cheapest name on every forward measure even after its rally: 24.6× next-year earnings against 38× at Marvell and 63× at Astera Labs; 9.8× next-year sales against 12.9× and 23.0×; a PEG of 0.45 against 0.60 and 1.29 — while its latest-quarter growth of 115% is three times Marvell’s and above Astera’s, and its 48% non-GAAP operating margin is the highest of the three. Three months ago the reverse was true: at $308 the shares were on about 23 times fiscal 2027 sales and 60 times earnings, the richest in the group. Against Broadcom the picture is different. Broadcom trades at 17 times its next fiscal year on an 82% EPS growth outlook it says supply is secured for, and at a 46% free-cash-flow margin — a PEG of 0.20 that says the market is pricing its two-year AI roadmap at a heavy discount. Credo is not cheaper than that; it is cheaper than the companies it actually competes with for the same sockets. Relative cheapness is a statement about Marvell’s and Astera’s multiples, not about Credo’s cash flows.
Three things explain the discount and all three are in the table. First, the gap between quoted and reported profit: 23 points between Credo’s non-GAAP and GAAP operating margin, the widest in the group, because share-based pay runs at 18% of revenue (Astera 16%, Marvell 12%, Broadcom 7%) and the DustPhotonics amortisation has just started. Second, concentration and the purchase-order model: four end customers at 84% of revenue against Broadcom’s broader base and Marvell’s mix of custom, switching and storage; Astera shares the concentration problem and is priced higher regardless. Third, a full year reiterated rather than raised after a beat, which told the market the second half has to do the work. Optical peers are priced very differently: Lumentum, on a 130%-plus growth guide, is on 12.9 times sales and 44 times earnings for a year that has only just begun, with a 10% cash-flow margin and a 50% gross margin — the market pays for growth there without demanding the cash.
Is the discount compensating for the risks identified elsewhere in this report? Less than it was a week ago. A PEG of 0.45 against 0.6–1.3 for Marvell and Astera still says the market prices some probability that fiscal 2028 consensus is wrong, but the absolute level — 31 times fiscal 2028 earnings after share-based pay and tax, for a company with no contracted backlog — no longer leaves room for a 30% bear weight. That is why the model does not add a concentration premium to the discount rate: the risk is carried once, in the bear scenario and its weight. What expands the multiple: a third-quarter guide near $690m in December, first-half optical revenue disclosed at or above $200m, a fifth end customer above 10%, and evidence that fiscal 2028 gross margin holds 67%-plus as modules scale. What contracts it: a second-half shortfall, a hyperscaler capex pause in the October reporting round, a competitive loss at 1.6T, or continued cost growth above guidance.
The peer discount does not make the shares undervalued on our framework. Every company in the table is priced on non-GAAP earnings that treat share-based pay as free, and the two direct peers on growth outlooks — Marvell’s $18bn, Astera’s Scorpio ramp — that are as uncontracted as Credo’s second half. A relative discount to an expensive group is consistent with overvaluation on absolute cash flows, and Broadcom, the one peer on a lower forward multiple, has a 46% cash-flow margin and a supply-secured two-year roadmap that Credo cannot match. On peer multiples the shares are fairly priced; on their own cash flows they are not. When the two regimes disagree the rating follows the cash flows, and the disagreement is disclosed rather than averaged away.
Intrinsic value is not a price target and this report does not publish one. But in twelve months the market will price CRDO on fiscal 2028 earnings, and the multiple regime it applies is worth being explicit about. At 22–28 times the $7.90 fiscal 2028 non-GAAP consensus the shares would trade at $174–221, a range that brackets today’s price. This is a market-regime estimate, not a fair-value estimate, and it is not the anchor of the rating: it prices the company on the convention the model rejects, at a multiple the direct peers currently exceed.
The base sits above management’s plan from fiscal 2029. Re-run on the $5.0bn fiscal 2031 strategic plan exactly (fiscal 2029 $4.3bn, 2030 $4.7bn, 2031 $5.0bn) the value is $121. Share-based pay is the swing factor, and it is measured at grant-date fair value. Grants made at $155–227 in the June quarter are expensed at those values even though the shares traded at $150 in mid-September; if the true economic cost is the dilution at today’s price, the charge overstates it. Halving the path gives $148; ignoring it gives $161; ignoring it and holding the tax at 1.5% gives $187. Working capital at 25% of incremental revenue is a heavy drag in fiscal 2027–29; at 15% the value is $138, a small change because the drag reverses as growth slows. Product-line revenue is not disclosed, so the optical and cable split, and therefore the mix-driven gross-margin path, is inferred from the $600m optical target and management’s “more than 90%” cable commentary. Peer forward figures are partly derived from company outlooks, and hyperscaler capex is a single, correlated driver of every scenario here. Upside variants: gross margin held at 68% throughout gives $144; fiscal 2028 at $4.0bn with the same growth rates thereafter gives $147; both together give $157. None reaches the price. And the market data are intraday: the $194.35 price was captured during the 24 September session at 13:08 New York time, and the close will differ.
Credo reported after the close on 1 September. Revenue of $479m beat the company’s own $465–475m guide and a $472m consensus; non-GAAP EPS of $1.20 beat $1.17. The shares fell 20% the next session. When a beat on every headline costs a fifth of the company, the price is telling you what it assumed.
| Metric | Q1 FY27 | Q4 FY26 | Change | Why it matters |
|---|---|---|---|---|
| Revenue | $479.0m | $437.0m | +9.6% q/q · +114.7% y/y | Seventh straight quarter above 100% year on year; above the guided range |
| Gross margin — GAAP / non-GAAP | 64.5% / 68.0% | 68.2% / 68.3% | −3.7 / −0.3 pts | The GAAP gap is $11.0m of acquired-intangible amortisation in cost of revenue; the underlying margin is steady |
| Operating expenses — GAAP / non-GAAP | $188.4m / $95.2m | $142.2m / $81.7m | +32% / +16% | Non-GAAP above the $86–90m guide on R&D headcount and design activity |
| Operating margin — GAAP / non-GAAP | 25.2% / 48.2% | 35.7% / 49.6% | −10.5 / −1.4 pts | The 23-point gap is share-based pay, amortisation and deal costs |
| Net income — GAAP / non-GAAP | $129.4m / $236.3m | $169.1m / $226.7m | −23% / +4% | Non-GAAP net margin 49.3%; GAAP hit by $88m of share-based pay |
| Diluted EPS — GAAP / non-GAAP | $0.67 / $1.20 | $0.88 / $1.16 | −24% / +3% | Beat the $1.17 consensus; 197.5m non-GAAP diluted shares |
| Share-based compensation | $88.0m (18.4%) | $49.7m (11.4%) | +77% | New grants (2.1m RSUs at $227; 1.9m PSUs) and acquisition awards; second-quarter guide about $104m |
| Operating cash flow / free cash flow | $90.2m / $82.9m | $182.2m / $177.5m | −50% / −53% | $151m working-capital outflow; capex $7.3m |
| Inventory | $313.1m | $250.8m | +25% | About 168 days of first-quarter cost of revenue; built “to support unfulfilled backlog and new product ramps” |
| Cash and investments | $764.3m | $1,443.3m | −$679m | DustPhotonics: $770m cash plus $169m of shares plus $310m contingent |
| Top four end customers | 33 / 28 / 13 / 10% | 4 at ≥10% | Same top three | The fourth is a returning customer; a year ago 33 / 35 / 20% |
| Q2 FY27 revenue guide | $525–535m | $465–475m Q1 guide | +10.7% q/q at mid | Above the $515m consensus; non-GAAP gross margin 67–69%; operating expenses $100–105m; about 200m diluted shares |
Source: first-quarter fiscal 2027 results release and call, 1 September 2026, and the Form 10-Q filed 2 September 2026. Consensus figures as reported after the print and not independently verified.
One: growth is still compounding, and it now has four legs. Revenue rose 115% with “substantial year-over-year growth across four domestic customers”. Active electrical cables, optical DSPs and retimers each set records.
Two: the full year was reiterated, not raised — and the shape got steeper. “More than 85%” growth (about $2.47bn) and “more than $600 million” of optical revenue were reaffirmed; the second-quarter guide was above consensus but implies the second half must be 45% larger than the first. Management calls this “an inflection in the second half”.
Three: the optical business has broadened from a chip to a stack. First revenue from silicon-photonics chips three months after the DustPhotonics close; two photonic-chip design wins “with major players” for fiscal 2028, for the chip only, with Credo’s DSP as “upside”; membership of the OpenCPX consortium for near-package optics; ZeroFlap transceivers in production. Management: “AECs continue to grow. Optics is growing faster.”
Four: costs stepped up faster than sales. Non-GAAP operating expenses of $95.2m were above the $86–90m guide, up 16% on the quarter and 75% on the year; GAAP operating expenses rose 32% on the quarter to $188m on $82m of share-based pay in operating expenses and $10m of deal costs; GAAP gross margin fell 3.7 points to 64.5% on acquired-intangible amortisation. Non-GAAP net margin still held at 49%.
Five: cash conversion fell while the balance sheet reshaped. Operating cash flow of $90m was down $92m on the quarter on a $151m working-capital build — inventory +$62m, receivables +$55m, capacity deposits +$23m — and cash fell $679m to $764m after the acquisition. Goodwill is now $986m, a third of total assets.
Fiscal 2027: revenue growth above 85%; optical above $600m, with three lines each above $100m; non-GAAP gross margin “broadly consistent” with fiscal 2026’s 68%; non-GAAP operating expenses up about 55%; non-GAAP net margin “in the vicinity of 50%”; guidance assumes the current tariff regime. The tone was confident and long-horizon — “we are really playing the long game”, fiscal 2027 “a stepping stone”, “outsized growth again” in fiscal 2028 “and the years that follow” — with reliability as the recurring “north star”. The chief executive was more expansive on fiscal 2028 products than on second-half mechanics; the chief financial officer stayed with numbers and did not quantify first-half optical revenue.
Stifel and Barclays pressed on the optical mix and the credibility of the module ramp under supply constraints. Bank of America noted that if optics is $600m, implied second-half cable growth is modest; management neither confirmed nor rejected it beyond “we see growth across the board” and “clearly a slower growth overall” for cables. TD Cowen asked about platform as well as customer concentration; Mizuho asked for a fiscal 2028 framework and received a market-size answer rather than a number; Susquehanna established that the two photonic-chip wins exclude Credo’s DSP. Nobody asked about the chief executive’s special award, insider sales or the minimum tax.
| Positive surprises | Negative surprises |
|---|---|
| Revenue and EPS above guide and consensus; second-quarter guide above consensus | Non-GAAP operating expenses above the guided range; GAAP operating expenses +32% on the quarter |
| First photonic-chip revenue one quarter after the close; two fiscal 2028 design wins | GAAP gross margin 64.5%, the first print below 65% in six quarters |
| OpenCPX membership opens scale-up optics | Cash conversion 38% of non-GAAP net income |
| Fourth end customer back above 10% | Full-year outlook reiterated rather than raised after a $7m beat |
What to watch next. The second-quarter print, expected around 1 December 2026, is less about the $530m than about the third-quarter guide: the fiscal-year outlook needs roughly $690m. Watch, in order, the optical revenue disclosed or implied for the first half; non-GAAP gross margin holding 67–69% as modules scale; inventory and receivable days; any change in the three or four customers above 10%; and whether operating expenses keep running above guide.
Credo designs the chips and cables that carry data between the processors, switches and memory inside an AI data centre. Its flagship is the purple ZeroFlap Active Electrical Cable — a copper cable with Credo’s own signal-processing silicon in each plug, carrying 400G, 800G and soon 1.6T Ethernet over the one-to-seven-metre runs from a GPU server to its first switch. These cables replace both passive copper, which is too lossy and too thick at these speeds, and optical cables, which cost more, use more power and are prone to the intermittent failures — “link flaps” — that stall large AI training jobs. Credo sells the complete cable: silicon, firmware, manufacturing, test and qualification are all its own.
Around that core sit four smaller lines: optical DSPs, the signal-processing chip inside a third party’s optical transceiver, now in their seventh generation and shipping at 50G, 100G and, from later this fiscal year, 200G per lane; silicon-photonics chips, acquired with DustPhotonics in May 2026; ZeroFlap optical transceivers, Credo’s own complete optical module, developed with Oracle and sold on reliability and telemetry rather than price; and retimers for PCIe and Ethernet. Two further products are due to ship in fiscal 2028: Active LED Cables, microLED-based optical cables reaching 30 metres with copper-like reliability, from the Hyperlume acquisition; and OmniConnect, a SerDes fabric and “Weaver” gearbox that lets an AI processor address memory off the substrate. A SerDes licensing business and the PILOT diagnostics software complete the portfolio. Everything is built on Credo’s own SerDes and DSP architecture, manufactured at TSMC on mature, one-generation-behind process nodes — the source of both its cost advantage and its supply availability while leading-edge capacity is rationed.
Customers are hyperscale cloud operators — all of the major ones are engaged, five deeply, with Amazon, Microsoft, xAI and Oracle identified by analysts or by the company’s own product announcements — the “neo-cloud” GPU-rental operators, and the OEMs, ODMs and optical-module makers that build for them. Sales are direct, through a two-pronged model that wins the specification at the end customer and then ships to whichever contract manufacturer the end customer nominates.
Unit sales on purchase orders, with no long-term take-or-pay commitments. Fiscal 2026 revenue was $1,335m, up 206%; first-quarter fiscal 2027 revenue was $479m, up 115% year on year and 10% on the quarter. The company does not disclose revenue by product line. From management commentary, cables were “more than 90% of the increase” in the first quarter and remain the largest business; the optical lines together are guided to exceed $600m in fiscal 2027, roughly a quarter of the year, with optical DSPs, photonic chips and ZeroFlap transceivers each above $100m; retimers delivered record revenue; licensing is a small, lumpy remainder. Non-GAAP gross margin has held at 67–68% for six quarters, and non-GAAP operating margin reached 48–50% because operating expenses run at just 17% of revenue. GAAP margins are far lower — 25% operating in the first quarter — because share-based pay was $88m, 18% of revenue, and the DustPhotonics deal added $11.6m a quarter of amortisation.
Credo sells into the connectivity layer of AI infrastructure spending, which the five largest US hyperscalers have guided to roughly $700–750bn of capital expenditure in calendar 2026. The company frames its market as pluggable transceiver units — copper cables, laser-based optics and, from fiscal 2028, microLED cables — which third-party forecasters expect to grow from about 60m units in 2026 to 175m in 2030, a 31% compound rate, before adding scale-up GPU-to-GPU networks that are ten times denser than scale-out. A blended $200 average selling price on 60m units is a $12bn 2026 market; Credo’s fiscal 2027 revenue of about $2.5bn would be roughly a fifth of it, concentrated in the cable segment where it holds the majority share it created. Management describes the optical, microLED, OmniConnect and scale-up opportunities as each “multibillion-dollar” and the total as “tens of billions”. Those are the company’s numbers and are read here as ceilings, not forecasts; none of them is used in the model.
On the cycle: hyperscaler capex is still accelerating — Amazon alone spent $54bn in the June 2026 quarter, up 68% — the largest customers describe supply as spoken for through 2028, and the two speed transitions that raise Credo’s content per port, 200G per lane in fiscal 2028 and the shift of scale-up networks toward near-package optics, are ahead rather than behind. The risk is not that 2026 or 2027 disappoint; it is that the market already assumes 2028 and 2029 will not, and the fiscal 2028 consensus of $3.65bn, up 47%, leaves nothing for a digestion year.
The 10-K names Broadcom, Marvell and Astera Labs, “as well as various cable and optical transceiver suppliers”. Marvell and Broadcom are the incumbent optical-DSP and SerDes leaders and both sell DSPs into cables made by cable houses; Astera Labs leads in PCIe retimers and is moving into scale-up fabric switches. Nvidia is developing its own connectivity, including co-packaged optics for scale-up, and the hyperscalers can and do design silicon in-house. Credo’s answer to all of them is the same: it is the only vendor that owns the whole link — SerDes, DSP, photonic chip, cable or module, firmware and telemetry — and sells reliability at the system level rather than a component on a specification sheet. That is a real advantage today and an execution-dependent one; Section 06 scores it.
| $m, FY ends ~1 May | FY2024 | FY2025 | FY2026 | Q1 FY27 | Comment |
|---|---|---|---|---|---|
| Revenue | 193 | 437 | 1,335 | 479 | +126% · +206% · +115% y/y |
| Gross margin (GAAP) | 62% | 64.8% | 68.0% | 64.5% | The first-quarter dip is $11m of acquired-intangible amortisation; non-GAAP 68.0% |
| Operating income (GAAP) | (29) | 37 | 445 | 121 | 33% margin in FY26; 25% in Q1 on $88m of share-based pay |
| Operating income (non-GAAP) | n/d | 115 | 639 | 231 | 26% → 48% → 48% margin |
| Net income (GAAP) | (28) | 52 | 472 | 129 | Effective tax about 0–1% (Cayman) |
| Net income (non-GAAP) | n/d | 130 | 662 | 236 | EPS $0.70 → $3.46 → $1.20 (Q1) |
| Share-based compensation | ≈43 | 77 | 183 | 88 | 13.7% of FY26 revenue; 18.4% in Q1; Q2 guide about $104m |
| Operating cash flow | n/d | ≈95 | ≈464 | 90 | Q1 held back by a $151m working-capital build |
| Free cash flow | n/d | ≈70 | ≈406 | 83 | FY26 free cash flow was 61% of non-GAAP net income |
| Cash and investments | ≈410 | 431 | 1,443 | 764 | −$679m in Q1: $736m net cash paid for DustPhotonics |
| Shares outstanding (m) | ≈165 | 171.2 | 185.4 | 187.95 | +9.8% in fifteen months: equity pay, an at-the-market offering, the acquisition |
Sources: Form 10-K for fiscal 2026, Form 10-Q for the first quarter of fiscal 2027, results releases. Fiscal 2024 and 2025 cash-flow figures and fiscal 2026 annual operating and free cash flow are assembled from quarterly disclosures on the results calls and are approximate (≈). n/d: not disclosed on that basis.
| Driver | Evidence |
|---|---|
| Cable penetration and speed | Deeper attach at five hyperscalers, new neo-cloud accounts, and the 200G-per-lane / 1.6T transition that raises content per cable — “some contribution” in the second half of fiscal 2027, “a much bigger way” in fiscal 2028 |
| Optical from near zero to $600m-plus in fiscal 2027 | Record DSP revenue, first photonic-chip revenue in the first quarter, two photonic-chip design wins “with major players” for fiscal 2028 ramps, ZeroFlap transceivers in production at hyperscalers and neo-clouds, first 1.6T DSP revenue later this year |
| Scale-up networks | Retimers at 100G and 200G per lane for Ethernet, UALink and PCIe Gen6; OpenCPX membership for near-package optics, with design wins expected to ramp in fiscal 2028 |
| Two new categories in fiscal 2028 | Active LED Cables, with demonstrations at OCP in October, and OmniConnect memory fan-out, which management sizes at “thousands of dollars of Credo content per GPU”, with Positron as first customer. Neither is in the base case |
| Operating leverage | Non-GAAP operating expenses guided to grow 55% against revenue growth above 85%, taking non-GAAP net margin to about 50% |
| Risk | Evidence |
|---|---|
| Customer concentration | Four end customers are 84% of revenue; one contracting party is 43% of revenue and 57% of receivables. Purchase orders can be cancelled or rescheduled on short notice and there is no contracted backlog |
| The second-half step | Fiscal 2027 requires $1.46bn in the second half. The optical ramp is a new manufacturing model, contract-built modules, with supply-chain and yield risk the cable business did not carry |
| Vertical integration and competition | Nvidia’s own connectivity and co-packaged optics, hyperscaler in-house silicon, and Marvell and Broadcom DSPs inside competing cables can compress share or price at each speed transition |
| Dilution and pay | Share count up 9.8% in fifteen months; unrecognised share-based pay of about $1.3bn; a 1.44m-unit special award to the chief executive; insiders sold about $465m of stock in twelve months under trading plans and bought none |
| Tax step-up | The company expects the OECD 15% minimum tax to apply from fiscal 2028, against an effective rate near zero today |
| Supply concentration | TSMC is the sole foundry; BizLink is the sole cable manufacturer; assembly and test are in Taiwan |
| Cycle | A hyperscaler digestion year in fiscal 2029 or 2030 is normal for capex-driven semiconductors and is not in the consensus |
| Source | Score | Assessment |
|---|---|---|
| Switching costs | 7 | Two-to-three-year design cycles, multi-month qualification of the product and its contract manufacturers, firmware and SDK integration into the customer’s network operating system. Once designed in, “customers are unlikely to change to another design until the next generation.” The flip side: every generation is a fresh contest |
| Cost advantage | 7 | Own SerDes and DSP on mature TSMC nodes (“similar performance at lower cost and on more available legacy nodes”), vertical integration from silicon to cable or module, a 68% gross margin with 17% operating expenses. Real, and the mature-node strategy is hard to copy for a rival committed to the leading edge |
| Brand | 6 | “Purple cable” and ZeroFlap are recognised inside hyperscaler networking teams; purple is a registered trademark when associated with these cables. Brand matters in a market that buys on reliability, but it does not travel to the component business, where Credo is one of three |
| Intellectual property | 6 | The SerDes and DSP architecture is the company’s foundation and has been licensed to chip makers for a decade. The patent estate is thin for the size of the business — 86 issued US patents and 52 in China — so protection rests on trade secrets and speed |
| Distribution | 6 | Direct engagement with all major hyperscalers, who then mandate Credo parts to their OEM, ODM and module suppliers. Efficient, and the same door the customers can close |
| Customer loyalty | 6 | Microsoft since the first cable, Amazon through the ramp, Oracle as ZeroFlap co-developer; five hyperscalers “deep”. Loyalty is to the system-level relationship and to reliability outcomes, and it has survived Marvell’s and Broadcom’s efforts to sell DSPs into competing cables |
| Scale | 5 | $2.5bn of revenue against Broadcom’s ~$106bn and Marvell’s ~$12bn; 807 employees. Sufficient for the cable niche and improving fast in optics after DustPhotonics; not a barrier to either incumbent |
| Data | 4 | Link-level telemetry across a growing installed base is a genuine, if early, differentiator that no cable or component vendor has assembled |
| Network effects | 3 | None in the classical sense. PILOT telemetry improves with the installed base, and participation in standards (OpenCPX, UALink, ESUN) shapes conventions — weak, indirect effects |
| Regulation | 2 | No regulatory protection; export controls and tariffs are a cost, not a moat |
Composite: 5.2 out of 10. A narrow moat, strongest in switching costs and cost position, weakest in structural lock-in.
| Competitor | Where they overlap | Their advantage | Credo’s advantage |
|---|---|---|---|
| Broadcom | Optical DSPs, SerDes, retimers, switches, co-packaged optics | Scale, the switch socket itself, leading-edge process, hyperscaler custom-silicon relationships | System-level cable and module ownership; speed; willingness to serve cost points Broadcom will not |
| Marvell | Optical DSPs (market leader), cable DSPs sold to cable houses, PCIe retimers, custom silicon | Installed base in optics, breadth, $12bn scale | A complete-cable franchise Marvell only supplies into; telemetry; 200G-per-lane timing parity |
| Astera Labs | PCIe and CXL retimers, scale-up fabric switches, signal conditioning | PCIe leadership, Nvidia rack-scale attach, the fabric-switch ramp | Ethernet and UALink breadth, copper and optical media, cost |
| Nvidia (in-house) | Own DSPs, co-packaged optics for scale-up, reference-design cables | Owns the platform and can bundle | Hyperscalers’ own silicon programmes want a merchant, multi-protocol connectivity vendor; Credo is deliberately “agnostic” |
| Cable and module houses (Asia) | Cables with merchant DSPs; standard optical transceivers | Manufacturing scale and price | Qualification rigour, firmware, PILOT; ZeroFlap sold above commodity prices |
Strongest advantage: being the only vendor that designs, builds, qualifies and monitors the entire link, on cheaper silicon than the incumbents use. Biggest threat: the next speed node — 200G per lane and scale-up optics — is the moment every socket is re-competed, and both Nvidia and the hyperscalers have reasons to bring it in-house. Direction: expanding in breadth — five product lines ramping where there was one, and optics now a stack — but not yet deeper: customer concentration and purchase-order economics are unchanged, and the company’s own risk factors still list “vertical integration by significant customers”. We score the moat narrow and expanding.
| Concern | Severity | Evidence and why it matters |
|---|---|---|
| Customer concentration | 8 · HIGH | End customers 33 / 28 / 13 / 10% of first-quarter revenue, the top four 84%; contracting-party Customer A 43% of revenue and 57% of receivables, Customer B 28% and 28%. The top ten were about 90% of fiscal 2026 revenue. Purchase orders only — “customers may cancel purchase orders on relatively short notice and without penalty.” One account pausing for two quarters is a 15–30% revenue hole with no contractual recourse. Mitigant: the largest account was 86% in January 2025 |
| No contracted backlog behind an 85% growth guide | 7 · HIGH | Remaining performance obligations were $4.2m at 1 August 2026, all within twelve months. Management refers to “unfulfilled backlog” when explaining inventory but discloses no number. The $1.46bn second-half requirement rests on customer forecasts, design wins and a beat-every-quarter record, not contracts |
| Share-based pay and dilution | 6 · MED-HIGH | $88m in the first quarter (18.4% of revenue), guided to about $104m in the second; $183m in fiscal 2026. Unrecognised expense about $1.27bn (8.3m RSUs at a $104 average, 3.1m PSUs at $131). Shares outstanding 171.2m in May 2025, 187.95m in August 2026, +9.8%, of which about 2m came from an at-the-market offering and 0.76m from the acquisition; the rest is equity pay. No buyback. Non-GAAP figures exclude all of it |
| Vertical integration and platform risk | 6 · MED-HIGH | The 10-K lists “our ability to capitalize on, and prevent losses due to, vertical integration by significant customers” as a competitive factor. Nvidia’s own connectivity and co-packaged optics, hyperscaler in-house DSPs and merchant DSPs inside third-party cables are all live substitutes at the next speed node. With two-to-three-year design cycles, a lost socket is lost for a generation |
| Insider selling | 5 · MEDIUM | About $465–470m sold over twelve months with no purchases, all reported sales under Rule 10b5-1 plans. The chief operating officer sold about 106,000 shares ($23.9m) on 15 July 2026 at $225, retaining 3.1m; the chief executive filed a Form 144 on 15 September 2026 for 33,338 shares (about $5m); the chief financial and technology officers also sell regularly. Plan-driven and small relative to holdings, but one-way through both the March low and the June peak |
| The GAAP to non-GAAP gap | 5 · MEDIUM | First-quarter GAAP operating margin 25.2% against non-GAAP 48.2%; GAAP net income fell 23% on the quarter while non-GAAP rose 4%. The reconciling items are real costs to shareholders (dilution) and real cash already spent (the acquisition). The company frames all guidance on the non-GAAP basis |
| Working capital and inventory | 4 · MEDIUM | Inventory $313m (+25% on the quarter, about 168 days), receivables $289m (+24%, about 55 days), refundable capacity deposits $88m with a further $102m committed for fiscal 2027. Working capital absorbed 26% of incremental revenue in fiscal 2026 and $151m in the first quarter. Normal for a doubling business — and the first place a demand shortfall would show |
| Goodwill from a $1.25bn acquisition | 4 · MEDIUM | DustPhotonics: $896m of goodwill and $362m of intangibles for a business with $4m of receivables and other assets; pro forma results “not material”. The $310m contingent consideration ($300m in shares) is recorded in equity at a fixed dollar value, so the share count it produces rises as the price falls. Impairment risk is tied to the optical ramp |
| Tax step-up | 4 · MEDIUM | Effective tax rate −0.5% in the first quarter and 0.7% in fiscal 2026 through a Cayman holding structure. The 10-K: “we expect we may be subject to legislation based on the OECD’s 15% global minimum tax regime in our fiscal year 2028.” Consensus non-GAAP EPS uses about 1%. PFIC status is assessed annually |
| Supply concentration | 4 · MEDIUM | TSMC the sole wafer source; BizLink the sole cable manufacturer, under the only long-term supply contract; Amkor, ASE, KYEC and Sigurd in Taiwan. Moving to a new foundry or assembler would take “approximately 9 to 12 months” plus requalification |
| Accounting, legal, governance | 2 · LOW | No material litigation; controls effective; no restatements; auditor attestation clean; revenue is unit sales on shipment. A Cayman exempted company — shareholder remedies are weaker than Delaware’s, and the company is ineligible for some US indices |
Overall red-flag score: 5 out of 10, moderate. Concentration, uncontracted demand, heavy equity pay and a large strategic acquisition — none of it hidden, all of it disclosed in the filings and quantified here. The two that matter most for the valuation are the second-half requirement and the tax step-up, and both are priced in Section 03.
| Strength | Weight | Evidence and why it matters |
|---|---|---|
| Cash generation at scale | 8 · STRONG | Fiscal 2026 free cash flow about $406m, 30% of revenue, while funding a $230m working-capital build and $58m of capex; fourth-quarter fiscal 2026 free cash flow $177.5m, 41% of revenue. With capex at 1–4% of sales and a 68% gross margin, the business converts at 60%-plus of non-GAAP net income once growth slows to normal. Net cash $754m, no debt |
| Concentration falling, engagement widening | 7 · STRONG | From one customer at 86% in January 2025 to four at 33 / 28 / 13 / 10% and five hyperscalers “with deep relationships”, plus neo-clouds and Positron. Every incremental hyperscaler ramp is both diversification and growth |
| The broadest product cycle in the company’s history | 7 · STRONG | 1.6T cables, 200G-per-lane DSPs, photonic chips with two named-tier design wins, near-package optics via OpenCPX, ZeroFlap modules, Active LED Cables and OmniConnect all have first or step-up revenue in fiscal 2028. The base case values Active LED Cables and OmniConnect at zero and near-package optics only through the general growth rate |
| Cost structure | 7 · STRONG | Non-GAAP operating expenses at 17% of revenue against 22% at Marvell and 35% at Astera in their latest quarters; mature-node silicon gives both lower cost and available capacity while advanced nodes are rationed. 807 employees produce $2.5bn of revenue |
| The board’s own anchor | 5 · MODERATE | The chief executive’s award set its first price hurdle at $244.70, described in May 2026 as “a substantial premium” to the April 2026 average of $150.02 — the level the shares revisited on 15 September before rebounding to $194. Six revenue hurdles from $2.5bn to $7.5bn map the range of outcomes the board considered plausible |
| Telemetry as an emerging data moat | 4 · MODEST | PILOT captures eye height, signal-to-noise and post-error-correction histograms on every link and integrates with switch software; management says the dataset will improve next-generation designs and failure detection. Early, but it is the one asset here that compounds with the installed base rather than with the next design win |
| Guidance record | 4 · MODEST | Revenue above the top of the guided range in every quarter of the ramp; fiscal 2026 finished about 60% above where consensus started the year. The pattern is a reason to weight the base above consensus rather than below it; we have not done so |
Overall green-flag score: 7 out of 10. The underappreciated items are real but mostly forward-looking, so they are carried as scenarios — bull 20%, super-bull 5% — rather than as base-case cash flows, the same treatment given to forward-looking risk in the bear.
| Dimension | Score | Assessment |
|---|---|---|
| CEO track record | 5 | Bill Brennan has led Credo since 2014; the founders, Chi Fung “Lip-Bu” Cheng and Lam Yat Tung, remain as chief technology and chief operating officers. Revenue $193m in fiscal 2024, $437m, $1,335m, about $2.5bn guided, with the cable category created from nothing and the optical stack assembled in nine months. Few semiconductor chief executives have delivered this trajectory |
| CFO credibility | 4 | Dan Fleming’s guidance has been exceeded at the top of the range every quarter of the ramp; he quantifies what he can (customer percentages, cash, inventory, share count) and declines what he cannot (product split, backlog). First-quarter operating expenses above guide and the drop in cash conversion were explained, not spun |
| Guidance accuracy | 4 | Consistently conservative: fiscal 2026 finished about 60% above where the year’s consensus began, and every quarter above the guided range. The deduction is for the pattern itself — a guide that is always beaten is one the market learns to inflate, which is part of why a $7m beat produced a 20% fall |
| Transparency | 3 | Good: the end-customer concentration table, capacity-deposit and purchase-obligation detail, acquisition allocations, an explicit Pillar Two warning. Poor: no product-line revenue, no backlog or bookings metric, “unfulfilled backlog” cited without a number, first-half optical revenue not quantified despite a $600m annual target |
| Capital allocation | 3 | Three acquisitions in nine months — Hyperlume ($92m, microLED), Comira ($35m, link-layer IP), DustPhotonics ($1.25bn including $310m contingent) — all strategic and all pre-scale; DustPhotonics’ revenue was “not material”. A roughly $348m at-the-market raise in October–December 2025 at about $150–190 was well timed. No return of capital. The judgement is on strategy, which is coherent, not yet on returns, which are unproven |
| Acquisitions — integration | 4 | DustPhotonics recognised revenue in its first quarter and produced two design wins within months; Hyperlume’s microLED cable is on schedule for fiscal 2028 revenue. Early, but on time |
| Buybacks | 1 | None. With share-based pay at 18% of revenue, the absence of any offset means shareholders absorb the full dilution |
| Dilution | 2 | Shares +9.8% in fifteen months; 8.3m RSUs and 3.1m PSUs unvested; 2.1m RSUs and 1.9m PSUs granted in a single quarter at $227 and $155 average fair value; a $300m earn-out to be settled in shares. Non-GAAP diluted share count guided to about 200m from 197.5m |
| Insider ownership | 3 | Founders and management retain meaningful stakes — the chief operating officer alone about 3.1m shares, roughly 1.7% — but selling has been continuous and one-directional, about $465m in twelve months with no purchases, through both the $88 low and the $308 peak |
| Compensation | 2 | The May 2026 special award — 1.437m PSUs in six tranches, revenue hurdles of $2.5–7.5bn and price hurdles of $244.70–489.40, described as the chief executive’s only equity grant for five years — is well structured: both hurdles required, a one-year post-vest holding period, no interpolation. It was nonetheless granted after a five-fold share-price run, to a chief executive already heavily incentivised, and its first revenue hurdle is roughly this year’s guidance. Tranche 1 is worth about $59m at its own hurdle price |
| Board quality | 3 | Conventional for a recent listing; independent committees; Cayman incorporation limits shareholder remedies. No evidence of governance failure; limited evidence of restraint on equity issuance |
| Communication style | 4 | Consistent vocabulary (“reliability is our north star”, “playing the long game”), candid on competition and on cable growth slowing relative to optics, disciplined about not naming customers. Prone to market-size language (“tens of billions”) where a number was asked for |
Average: 3.2 out of 5. Operators first, owners second: they run the business like owners and issue stock like a venture-stage company. In product strategy and execution the answer to “do they act like long-term owners?” is yes — the mature-node decision, the system-level cable model and the optical build-out are patient, multi-year bets that have paid off. In the treatment of the share count, not yet: dilution is running at a pace that costs shareholders roughly a fifth of revenue a year in fair-value terms, there is no buyback, and insiders sell into every price. The special award aligns the chief executive with a $5bn-plus revenue outcome and a share price above $245, which is exactly what shareholders want; it also confirms that the board regarded $150 as a base, not a ceiling.
| Window | Event | Upside | Downside | Impact · confidence |
|---|---|---|---|---|
| 21–23 Sep 2026 | ECOC 2026, Málaga: a 1.6T 2×DR4 module on the Cardinal 802 DSP and Kfir 200G photonic chip; 800G/1.6T ZeroFlap optics with PILOT. Follows the AI Infra Summit (15–17 Sep) and FTSE All-World inclusion (21 Sep) | First full-stack 1.6T module demonstrations firm up the fiscal 2028 narrative | None material | Low · high (company release) |
| Sep–Oct 2026 | Insider 10b5-1 sales; chief executive’s Form 144 filed 15 Sep for 33,338 shares | A pause or a purchase would be the first in a year | Continued selling into the rebound reads badly | Low · high (Forms 144 and 4) |
| 6 Oct 2026 | Marvell investor day — long-term connectivity strategy | Marvell frames cable and optical DSPs as shared growth, validating the market size | An aggressive cable-DSP or module push at hyperscaler accounts | Medium · high |
| Oct 2026 | OCP Global Summit — Active LED Cable demonstration; annual meeting likely (proxy filed 25 Aug) | Named microLED cable customers; hyperscaler endorsement of the reach | Slip to late fiscal 2028; say-on-pay dissent over the special award | Medium · high on the event |
| Late Oct – early Nov 2026 | Hyperscaler September-quarter results and 2027 capex commentary (Amazon, Microsoft, Alphabet, Meta, Oracle) | 2027 capex guided up again; “supply spoken for through 2028” reiterated | Any “digestion” or “optimisation” language — the bear-case trigger | High · high on dates |
| ≈1 Dec 2026 | Q2 FY27 results and the Q3 guide | Revenue above $535m, a Q3 guide near $690m or above, first-half optical revenue quantified, gross margin 68%-plus | A Q3 guide below about $650m puts the full-year outlook in doubt; cost overruns repeat | Critical · high |
| Dec 2026 – Jan 2027 | First 1.6T DSP revenue; ZeroFlap ramps at more hyperscalers and neo-clouds; remeasurement of the DustPhotonics contingent consideration | Named 1.6T module and DSP wins | An earn-out written down means optical targets were missed | Medium · medium |
| ≈2 Mar 2027 | Q3 FY27 results — the inflection quarter | The step-up delivered; a fifth customer above 10% | A miss confirms the guide was aspirational; inventory write-downs | Critical · high |
| Mar 2027 | OFC 2027; Nvidia GTC | Near-package optics engines shown with design partners | Nvidia’s co-packaged optics roadmap pulls scale-up connectivity in-house | Medium · high on the events |
| May 2027 | Fiscal 2028 begins: the 15% minimum tax expected to apply; first revenue from Active LED Cables, OmniConnect and near-package optics; the 1.6T cable ramp | New categories add a fourth and fifth growth leg | Non-GAAP EPS re-based for a 15% tax rate, about −14% mechanically | High · medium |
| ≈1 Jun 2027 | Q4 FY27 results and the fiscal 2028 outlook | A fiscal 2028 guide at or above the $3.65bn consensus (“outsized growth”) | A guide of +25–30% resets the multiple even if fiscal 2027 was delivered | Critical · high |
| Through FY27–28 | The tariff regime (guidance assumes the current one); the Taiwan supply chain; the annual PFIC determination | Tariff relief for AI infrastructure components | Section 232 semiconductor tariffs on Taiwan-assembled product | Medium · low |
| Any time | M&A — Credo as acquirer (Hyperlume, Comira, DustPhotonics in nine months) or as target | Change-of-control provisions in the chief executive’s award interpolate to the offer price | Another $1bn-plus deal from a $764m cash base means equity issuance | Medium · low |
Buybacks: none, and none expected. Dividends: none. Results dates are estimated from the company’s pattern of reporting on the first business day of June, September, December and March. Confidence reflects the timing of the event, not the direction of its outcome.
BULLGrowth. Seven straight quarters above 100%, and the growth is broadening as it compounds: four customers at or above 10%, records in three product lines, optical going from nil to more than $600m in a year. The pluggable market goes from about 60 million units to 175 million by 2030 and Credo has more content per port at every speed step. Fiscal 2028 adds five new revenue lines. Consensus at +47% is, on this company’s history, a floor.
BEARGrowth is 84% four accounts and 100% one industry’s capital spending, and the second half of this year is 59% of the plan with $4.2m of contracted backlog. Cables — still the great majority of revenue — are, in the chief executive’s own words, “clearly a slower growth overall.” The strategic plan the board wrote in May gets to $5bn by fiscal 2031; consensus gets to $3.65bn by fiscal 2028. One of those implies the other decelerates to about 11% a year immediately after next year.
BULLValuation. Even after the rebound the stock is on 24.6× next year’s earnings and 9.8× sales with a 47% year ahead; Marvell, guiding to 50% itself, is on 38× and 13×; Astera is on 63× and 23×. This report’s own model gives $187 when stock pay and tax are treated the way every peer’s multiple and the consensus treat them — within 4% of the price — and its bull case is $207, above it. Apply the group’s multiples and you get $174–221, and the price is inside that. The June high of $308 was a 2030 story; $194 is a 2028 story, and 2028 is the year the company says will be “outsized.”
BEARCheap relative to a group priced for perfection is not cheap, and a 29% move in seven sessions on no company news is a rally, not a revaluation. Every cell of the sensitivity grid is below the price — the most generous by 13%. The $187 is exactly the point: it is what the business is worth if $450m a year of equity pay costs nothing and the 15% tax the 10-K expects never arrives. Charge both and the multiple is 31× a fiscal 2028 number that itself needs a 47% year with no backlog. The sell-side’s $245–310 targets are 31–39× the non-GAAP number — the same valuation that failed on 2 September.
BULLBusiness quality. Sixty-eight percent gross margins on mature silicon, 17% operating expenses, a 50% net margin, about $400m of free cash flow last year while doubling working capital. The only vendor that owns the whole link and can prove reliability with telemetry — and reliability is what a hyperscaler actually pays for when one link flap idles a cluster worth more than Credo. Category creator once with cables, plausibly twice with ZeroFlap optics, with a third in Active LED Cables in the wings.
BEARPurchase-order economics, no long-term contracts, a socket re-competed every generation, and customers who publish their own silicon roadmaps. Nvidia is building co-packaged optics precisely to remove the pluggable transceiver from scale-up networks, and Credo’s own 10-K lists “vertical integration by significant customers” as a competitive factor. Eighty-six US patents is a thin estate for a $36bn company. The moat is speed, and speed is not a moat when Broadcom decides to run.
BULLFinancials and cash. Net cash of $754m after paying $736m for DustPhotonics in cash, no debt, and a 68% gross margin held flat for six quarters through a doubling of volume. The first-quarter dip in cash conversion is inventory and receivables in a quarter where revenue rose 10% and the company prepaid for capacity — exactly what you want to see ahead of an inflection. Working capital reverses when growth normalises.
BEARGoodwill is a third of assets. Inventory is 168 days and up 25% in a quarter; if the second half is late, that is the first write-down. GAAP operating margin fell ten points in a quarter. The company added $538m to paid-in capital in one quarter — $169m of shares for an acquisition, $300m of promised shares for an earn-out, $88m of stock pay — and the share count has grown 10% in five quarters. Free cash flow was 38% of non-GAAP net income last quarter.
BULLManagement. A chief executive who has compounded revenue thirteen-fold in three years, guides conservatively, beats every quarter, and has just accepted an award that pays nothing below $2.5bn of revenue and $245 a share. The founders still run the technology and operations. Three acquisitions in nine months, all integrated on time. That is what long-term ownership looks like in a hyper-growth semiconductor.
BEARInsiders sold about $465m of stock in a year and bought none; the chief executive filed on 15 September to sell more, with the shares at $150. Stock pay is 18% of revenue with no buyback. The special award was granted after a five-fold run, and its first revenue hurdle is roughly this year’s guidance. Product-line revenue — the single most important disclosure for judging the optical claim — is withheld. They are excellent operators; they are not yet stewards of the share count.
BULLCatalysts. December’s third-quarter guide, March’s delivery, June’s fiscal 2028 outlook and October’s OCP demonstration of Active LED Cables — four dated events, each of the kind this management has historically cleared. Hyperscaler capex commentary has been raised at every turn this year. The tape has turned: the 14-day RSI went from 19 to 70 in seven sessions, the price is back above its 200-day average, and the 2 September sellers have been absorbed.
BEARThe same four events are the four ways the thesis breaks, and the bar is the highest it has been: a $690m quarter, then $770m, then a fiscal 2028 guide that consensus already has at +47%. A single hyperscaler using the word “digestion” in October re-prices the whole group. And a 29% rebound on a sector rally is a description, not evidence: the shares went from $144 to $88 between January and March, and from $88 to $308 between March and June, on nothing company-specific in either direction.
On the business, the bull has the better of it. The quality of the franchise — margins, cost structure, cash generation, product cadence and the direction of diversification — is not in serious dispute, and the bear’s strongest structural point, vertical integration, has been present through the entire ramp without materialising. On the valuation, the bear has the better of it. The full-cost model does not reach the price in any cell of the grid. At $150 the gap was one cost convention wide; at $194 it takes both — stock pay and the coming tax treated as free — and still falls 4% short, and with both charged it needs the bull revenue path as the central expectation or an 8% discount rate. The sell-side anchors that failed on 2 September were built on exactly those conventions. What is uncertain is the shape of fiscal 2028–2030: consensus and the board’s plan cannot both be right, and the difference between them is roughly the difference between $121 and $147 a share. Verify next: first-half optical revenue, disclosed or inferred from the second-quarter print; the third-quarter guide against the roughly $690m requirement; end-customer and contracting-party concentration for any change in the top two; inventory days; fiscal 2028 gross-margin commentary as modules scale; and the pace of Form 144 filings. The committee’s ruling is that the shares are overvalued: the price reflects a good business, an unproven second half and a fully delivered fiscal 2028 product cycle, in proportions that leave nothing for the second and third going wrong. The bull’s strongest point — that on peer multiples, and with both costs waived, the shares are close to fair — is conceded and disclosed; it does not change a cash-flow verdict.
Inside an AI data centre, thousands of computer chips have to talk to each other at enormous speed. Credo makes the cables, and the tiny chips inside those cables, that carry the conversation between a server and the network switch it plugs into. Its purple cables are copper with a Credo chip in each end, which lets copper do a job that used to need more expensive, more fragile fibre-optic cables. It also sells the chips that go inside other companies’ fibre-optic cables, and since May 2026 it makes complete fibre-optic modules of its own. Almost all of its customers are the handful of giant cloud companies building AI computing capacity.
It sells hardware, one order at a time. There are no long-term contracts: customers place orders a few months ahead and can change them. Because Credo designs its own chips on older, cheaper manufacturing processes and outsources the factory work, it keeps about 68 cents of every sales dollar after making the product, and roughly 50 cents after running the company on its own preferred measure. That is unusually profitable.
Sales have more than doubled year on year in every quarter for almost two years — from $437m in the year to May 2025 to about $2.5bn expected in the year to May 2027. Very few companies of any kind grow like that while making money. The shares rose from $88 in March 2026 to $308 in June, fell back to $150 by mid-September and have since rebounded to $194.
Employees and executives are paid in a lot of stock — about 18 cents of every sales dollar — and the company pays almost no tax today but expects a 15% rate from 2027. The profit figure most people quote leaves both of those out. Count them and the business is worth well below today’s price; leave them out and it is worth about what it costs. Whether those two costs are real is most of the disagreement about this stock. We think they are.
Our estimate of what the business is worth is $113–169 a share, most likely about $135. The price of $194 is above that range, so we call the shares overvalued. On 15 September, at $150, the price was inside it; nothing about the business changed in between.
Right: the cloud companies keep spending; Credo’s new optical products sell as fast as its cables did; two more product lines arrive in 2027–28 as promised; no single customer matters as much as it used to; and the market keeps valuing Credo the way it values its peers, which would be roughly $175–220 a share.
Wrong: the second half of this fiscal year arrives late and the shares fall again; one of the four big customers pauses or builds its own version, and that customer could be a third of sales; the AI building boom slows for a year, as building booms do; and existing shareholders keep being diluted by stock pay while insiders keep selling.
The results around 1 December, and in particular the guidance for the following quarter, which needs to be near $690m. What the big cloud companies say about their 2027 spending in late October. And the number of shares outstanding each quarter.
| Status | What it covers |
|---|---|
| Verified from Credo filings and releases | Revenue, gross, operating and net income on both bases, share-based compensation, cash-flow items, working-capital lines, cash and investments, goodwill and intangibles, remaining performance obligations, capacity deposits and purchase obligations, customer concentration on both the end-customer and contracting-party bases, share counts, the DustPhotonics purchase-price allocation and earn-out, the Pillar Two statement, guidance, and the terms of the chief executive’s special award — from the Form 10-K for the fiscal year ended 2 May 2026 (filed 15 June 2026), the Form 10-Q for the quarter ended 1 August 2026 (filed 2 September 2026), the results releases of 1 June and 1 September 2026, the 1 September 2026 conference call, the 8-K of 1 June 2026, and Forms 4 and 144 filed through 15 September 2026. |
| Verified from market data | The $194.35 intraday price of 24 September 2026, taken at 13:08 New York time; peer prices taken between 13:08 and 13:12 the same day; the daily close series from 25 September 2025 to 23 September 2026 behind the drawdown, volatility and technical figures (14-day RSI 70, from 19 on 15 September; 20-day average $182; 50-day $210; 200-day $176; 52-week range $86.49–308.67); and peer margins, growth and guidance from each company’s latest results release, dated in Section 03. |
| Third-party, attributed | Consensus figures — about $2.48bn and $5.10 for fiscal 2027 and $3.65bn and $7.90 for fiscal 2028 — gathered from public aggregators after the 1 September print, secondary and not independently verified; sell-side price targets; third-party unit forecasts for the pluggable transceiver market. Attributions of the company’s numbered customers to named hyperscalers are drawn from published broker research and press reports and have not been confirmed by the company. Peer forward figures marked ≈ are derived from company outlooks, not consensus. |
| Our estimate or judgement | Every scenario input in Section 03; the product-line revenue split; the fundamental beta and the 10.0% rate; the 30 / 45 / 20 / 5 weights; the sensitivity grid, the reverse DCF, the $113–169 band and the $135 central estimate; fiscal 2025 and 2026 annual operating and free cash flow, assembled from quarterly disclosures; days of inventory, receivables and payables; enterprise value and every multiple; and all scores in Sections 06, 07 and 08. The model on this page is the model that produced those figures — the scenario cards load the published paths and the sensitivity grid is reproducible from the base settings. |
| Still open | Revenue by product line is not disclosed, so the optical and cable split — and with it the gross-margin path — is inferred from management’s $600m-plus optical target and its “more than 90%” cable commentary. No backlog or bookings figure is disclosed beyond $4.2m of remaining performance obligations. Lumentum’s net debt after its fiscal fourth-quarter convertible equitisation was not verified. The identity of the numbered customers is not corroborated. |
Method. Unlevered discounted cash flow, methodology version 1.3, under the cash-generative operators framework: a ten-year explicit forecast for fiscal 2027–2036 valued as at 24 September 2026, with a stub for the remaining three quarters of fiscal 2027 and a mid-year convention thereafter, and a Gordon-growth terminal value at 3.0%. Free cash flow to the firm is non-GAAP operating profit less share-based compensation, after tax at 1.5% in fiscal 2027 and 15% thereafter, plus depreciation, less capital expenditure and less working capital absorbed by growth; amortisation of acquired intangibles is excluded. The first quarter’s $83m of free cash flow, already in the cash balance, is netted out; $754m of net cash is added to enterprise value and the result divided by 191.2m shares. One discount rate and one terminal growth rate are used across all scenarios. Scenario weights are owned by the analyst, the fair-value band is read from the sensitivity grid, and the reverse DCF is run against the market price. The rating changes to fairly valued if the price falls below $169 with no change in the evidence, and to undervalued below $113; the conclusion is most sensitive to the discount rate, the share-based compensation path and the fiscal 2028–2031 revenue slope, in that order. Ratings are undervalued, fairly valued or overvalued according to where the price sits relative to the band; Sonde does not issue buy, sell or hold recommendations or price targets. The reference price used throughout is the $194.35 intraday quote of 24 September 2026, and every multiple in this report scales directly with it.