Cloud Computing
A quietly excellent RF-and-photonics compounder that the market has finally noticed — fab-lite 56% gross margins, three secular legs (defense, AI data-center optics, telecom), a net-cash balance sheet after killing its 2026 converts, but now priced at ~54× forward with an AI-optical multiple it must keep compounding into; great business, demanding entry.
Research
The verdict
A quietly excellent RF-and-photonics compounder that the market has finally noticed — fab-lite 56% gross margins, three secular legs (defense, AI data-center optics, telecom), a net-cash balance sheet after killing its 2026 converts, but now priced at ~54× forward with an AI-optical multiple it must keep compounding into; great business, demanding entry.
MACOM designs and manufactures high-performance analog, RF, microwave, millimeter-wave, and photonic/optical semiconductor components — ICs, MMICs, multi-chip modules, diodes, amplifiers, switches, lasers and optical subsystems — that customers drop into larger systems: radar, basestations, optical-network line cards, satellites, test-and-measurement, medical. It is "the analog plumbing of high-frequency signal chains," not a digital-compute company. HQ Lowell, Massachusetts; >70 years of heritage (traces to Microwave Associates).
Three end markets (one reportable segment, but managed and disclosed as three):
Business model. ~6,000 end customers; thousands of catalog + custom parts. Sold via direct sales, independent reps, and distributors. No long-term volume contracts — almost entirely purchase-order and "turns" business (ship from inventory shortly after order), which makes backlog a weak predictor and quarters lumpy. Some products have 5–20-year life cycles (long-tail catalog annuities), others <5 years.
Customer concentration: no single direct customer ≥10%, but two resellers/distributors each exceeded 10% in FY25 — one at 12.4%, a second at 11.2% (these are the channel partners, e.g. the Richardson RFPD / Arrow-type distributors, not end demand). Distributors were 32.3% of revenue (up from 29.3% / 24.0%). Top-10 customers = 56.7%; top-25 direct = 45.6%. International = 56.3% of revenue; China 28.4%, rest-of-APAC 11.5% — meaningful geopolitical exposure.
Upstream → MACOM → end customer, named where disclosed:
Chokepoints: (1) single-foundry-per-process external wafer dependency; (2) Asian back-end concentration; (3) China both a 28% revenue market and a sanction/export-control flashpoint (BIS rules of Oct-2022/2023). Mitigants: multi-sourcing where feasible, U.S.-based front-end fabs, in-house Trusted Foundry.
Bargaining power. Mixed. Over customers: weak on the giants (no >10% direct customer, but PO-only relationships mean customers can cancel/defer with little penalty) yet strong where it holds a Trusted-Foundry-only or sole-qualified-source position. Over suppliers: limited (single-foundry-per-process dependence). The honest read: the moat is technology + accreditation + design-win lock-in, not scale (rivals ADI/Broadcom/Marvell are far larger and explicitly "have greater financial resources and scale than us").
End-market revenue, FY (in $000s) and latest quarter:
| Market | FY23 | FY24 | FY25 | FY25 YoY | Q2-FY26 | Q2 YoY | Q2 mix |
|---|---|---|---|---|---|---|---|
| Industrial & Defense | n/a | 351,639 | 419,785 | +19.4% | 120,652 | +22.4% | 41.8% |
| Data Center | n/a | 197,875 | 292,836 | +48.0% | 98,188 | +36.0% | 34.0% |
| Telecom | n/a | 180,064 | 254,637 | +41.4% | 70,115 | +7.6% | 24.2% |
| Total | 648,407 | 729,578 | 967,258 | +32.6% | 288,955 | +22.5% | 100% |
MACOM reports a single operating segment, so segment EBITDA/EBIT by market is not disclosed — n/a at the segment-profit level. Trend read:
The cleanest way to read MACOM right now: operations are firing; GAAP is noisy from FY25's debt refinancing.
Income statement:
Guidance (the real catalyst). For fiscal Q3 FY26: revenue $331–339M (mid $335M) and adjusted EPS $1.31–1.37 (mid $1.34) — both well above the ~$306M / ~$1.17 consensus going in. That implies ~+16% sequential revenue and a steep adjusted-EPS ramp.
Balance-sheet flags:
Market reaction & what was priced in. The Q3 guide drove the stock sharply higher; Barclays raised its target $400→$450 (2026-05-22) and Needham $250→$400 (2026-05-08). The tape says the market now prices MACOM as an AI-optical compounder, not a cyclical RF name. Unusual vs its own history: 56.9% GM is the highest in the dataset; the prior-year H1 GAAP loss is a refinancing artifact, not operational.
No transcripts on disk (transcripts=0); this lens is ``/inference from releases and is the dossier's softest. From the FY25 10-K and Q2-FY26 release language, management's consistent focus: Data Center optical (100G→1.6T→3.2T), defense program ramps, design wins, and gross-margin expansion via the RTP fab and mix. The objective signal of rising confidence is behavioral: guiding Q3 ~16% above consensus and repaying the 2026 converts in cash rather than rolling them. Tone has shifted from "cyclical-recovery/integration" (FY24, RF-business + OMMIC + ENGIN-IC integration) to "secular AI/data-center demand" (FY25→FY26). To verify against the actual Q1/Q2-FY26 call transcripts — book-to-bill, lead times, China commentary, and whether the Telecom deceleration is timing or trend.
Peer set: AI-optical/connectivity names MACOM competes with on the growth narrative, plus the RF-analog peers it competes with operationally. Multiples are ``, dated; ROE 5-yr averages and several EV multiples are not sourced rather than fabricated.
| Company | Ticker | Mkt cap (USD) | Fwd P/E | EV/Sales | EV/EBIT | Div yield | 5-yr avg ROE |
|---|---|---|---|---|---|---|---|
| MACOM | MTSI | ~$29.9B | ~54× | n/a | n/a | 0% | n/a |
| Credo Technology | CRDO | ~$50.7B | ~43× | n/a | n/a | 0% | n/a |
| Coherent | COHR | ~$70–78B | ~186× trailing | n/a | n/a | ~0% | n/a |
| Lumentum | LITE | ~$66–70B | high (P/E ~159× ttm) | n/a | n/a | 0% | n/a |
| Marvell | MRVL | ~$272B | n/a | n/a | n/a | low | n/a |
| Analog Devices | ADI | large-cap | ~32× | n/a | EV/EBITDA ~32× | ~1.2% | n/a |
| Qorvo | QRVO | mid-cap | ~15× | n/a | EV/EBITDA ~12× | n/a | n/a |
| Skyworks | SWKS | mid-cap | ~13× | n/a | EV/EBITDA ~9× | ~4.0% yield | n/a |
Read. MACOM sits in a valuation no-man's-land that is actually flattering: at ~54× forward it is far cheaper than the NVIDIA-anchored optical pure-plays (Coherent/Lumentum at 150×+ trailing, each with a ~$2B NVDA strategic stake ) and modestly richer than Credo (~43× forward), yet a 3–4× premium to the RF-analog peers (Qorvo ~15×, Skyworks ~13×) that lack the AI-optical growth. The market is paying MACOM for the data-center optical story while it still carries a defense + telecom base the optical pure-plays don't have. The bull says that diversification deserves a premium to Credo; the bear says 54× already assumes the optical ramp compounds for years.
Pattern over the last ~5 years and recently:
Income statement / balance sheet / cash flow scan:
Regulatory findings:
Built bottom-up from the latest actuals + the company's own Q3-FY26 guide, into base/bull/bear. All outputs ``; inputs labeled. MACOM's fiscal years: FY26 ends ~2026-10-02, FY27 ~2027-10-01, FY28 ~2027-09(FY28 end).
Anchors: H1-FY26 adjusted EPS ≈ $1.09 (Q2) + ~$1.0x (Q1); Q3-FY26 guide adj-EPS mid $1.34; analyst framing ~+13–14% revenue CAGR, ~+16% EPS growth through FY28. FY25 revenue $967.3M.
| Path | Key assumptions | FY26 adj EPS | FY27 adj EPS | FY28 adj EPS |
|---|---|---|---|---|
| Base | Q3 guide holds (~$335M), Q4 ~$345M → FY26 rev ~$1.24B (+28%); GM ~58% adj; ~+15% rev FY27, +13% FY28; SBC steady; ~75M dil. shares | ~$4.90 | ~$6.0 | ~$7.1 |
| Bull | Data-center optical (1.6T/3.2T) compounds 35%+, GM →60% adj, telecom re-accelerates | ~$5.20 | ~$7.0 | ~$9.0 |
| Bear | China/export-control hit + data-center digestion; rev growth fades to high-single-digits FY27–28, GM slips to 55%, SBC dilution | ~$4.40 | ~$4.6 | ~$4.8 |
Valuation cross-check. At ~$391, base FY26 ~$4.90 ⇒ ~80× FY26 / ~65× FY27 adj EPS — richer than the ~54× "forward" figure the screens quote (the screen is likely blending a higher Street FY27 or a NTM number). Either way the multiple is full: MACOM is priced for the bull path to substantially hit. Forecast NOT logged to forecast.ts — --watchlist breadth rule (log a Brier forecast only on genuine committed conviction, done in /thesis). The honest base case: a high-quality compounder whose fundamentals justify a premium but whose current price already embeds several years of the optical ramp.
Bull case. Three secular legs rarely found together in one mid-cap: (1) AI data-center optics — the only leg the market is paying for, growing 36–48%, re-mixing the company toward higher-value coherent/photonic content at 100G→3.2T; (2) U.S. defense — Trusted-Foundry-protected, budget-tailwind, sticky; (3) telecom/5G/SATCOM — cyclical but real. Layer on 56.9% record gross margins and a fab-lite + net-cash balance sheet after the convert cleanup, a proven operator (Daly/Hittite), and interest-free 2029 leverage. If data-center optics compounds and telecom re-accelerates, FY28 adj EPS pushes toward $8–9 and the premium is earned. Contrarian view the market underrates: MACOM is the diversified, profitable, U.S.-fab way to own AI optical content — less single-customer-fragile than the NVDA-anchored pure-plays, yet it trades at a fraction of their multiple.
Bear case — what could permanently impair or de-rate:
Pre-mortem (18 months out, thesis broke). Most likely: AI data-center optical demand "digests" after a hyperscaler capex pause; MACOM misses a quarter or guides flat; the optical complex de-rates from 150×→80×, dragging MTSI from ~54× to ~30× forward — a 40%+ drawdown — even though revenue still grew ~15%. The business is fine; the price was the risk.
Dismantling the bull case. What structurally breaks the money machine? Revenue concentration sits in the channel, not the logo — two distributors are >10% each and distributors are 32% of sales; a channel de-stock (which the company can't see, by its own admission of "limited visibility into customers' inventories") would crater a quarter. The moat is narrower than bulls think on the growth leg: the Trusted-Foundry/defense moat is real, but in data-center optics MACOM competes head-on with Broadcom, Marvell, Credo, Coherent and Lumentum — far larger, better-capitalized, several NVDA-anchored. MACOM is a content supplier into optical modules, not the architect of the link; it can be designed around. The most dangerous competitor bulls underestimate: Broadcom/Marvell vertical integration in optical DSP + the silicon-photonics push (NVIDIA-funded at Coherent/Lumentum) could commoditize the discrete analog/photonic content MACOM sells. Worst capital-allocation/accounting tells: none egregious, but a serial-acquirer with $336M goodwill, $79M-and-rising SBC flattering adjusted EPS, a combined Chair/CEO, and a CEO trimming 100k shares into the run is a profile that demands scrutiny. What must hold for ~$391: ~15%+ revenue CAGR and a sustained 50×+ multiple for years. If growth disappoints 20–30%: at ~30× a $4.4 bear FY26 EPS, fair value ≈ $130 — a ~65% drawdown from here. Single scenario that permanently impairs: a China export-control regime that removes 28% of revenue and hands the share to subsidized domestic competitors — plausible-but-not-base.
The largest independent AI neocloud — renting NVIDIA compute to OpenAI, Microsoft and Meta under $60B of contracts, financed by an equally vast pile of debt.
A real, cash-generating neocloud retrofitter trading at ~18x trailing sales on a single $865M Nscale contract and a still-71%-Bit-Digital-controlled cap table — the build is genuine, but the multiple already prices the NC-1 inflection that hasn't happened yet.
The default arms dealer of the AI buildout — a real moat compounding a $15B backlog into 30% organic growth, but priced at 82x for perfection while insiders sell 65:0 and EMEA orders are already cracking.