Semiconductors
A real, recovering analog franchise (TMR current-sensing + auto/data-center) trading like a hyperscaler — the business is fixed, the stock is not cheap, and a ~32% Sanken overhang still wants out.
Research
The verdict
A real, recovering analog franchise (TMR current-sensing + auto/data-center) trading like a hyperscaler — the business is fixed, the stock is not cheap, and a ~32% Sanken overhang still wants out.
Allegro designs analog and mixed-signal sensor ICs and application-specific power ICs — the chips that let an electromechanical or power-conversion system "sense, move, and manage power." Two product lines: Magnetic Sensors (MS) — $538.5M, 60% of FY26 sales — and Power Integrated Circuits (PIC) — $351.6M, 40%. End markets: Automotive ($628.6M, 71%) and Industrial & other ($261.5M, 29%). Inside auto the growth engine is "Focus Auto" — ADAS (electric power steering, braking) and xEV (electrified powertrain) content; inside industrial the new engine is AI data center (fan control, power supplies, high-voltage current sensing), plus robotics and energy infrastructure.
Business model: classic analog — thousands of long-life parts, designed-in via multi-year "design wins," then shipped for the 5–10-year life of a car platform or industrial system. ~55% of FY26 sales went through distributors; the largest non-affiliated distributor was only 9.4% of sales and no single end-customer reached 10%. That diversification is a genuine quality marker — there is no Apple-style single-customer cliff here.
Headquarters: Manchester, New Hampshire; ~186.2M shares out (as of May 18 2026); listed NASDAQ: ALGM since the Oct-2020 IPO. Key contract structure is design-win-driven and non-cancellable-ish only by switching cost, not by take-or-pay — revenue is "sticky by socket," not contractually locked.
Allegro is fabless on wafers, hybrid on back-end.
Chokepoints: (1) foundry capacity on specialty BCD nodes — "limited production capacities with little ability to quickly expand" is the company's own language; (2) single/limited-source materials; (3) Asia concentration in both manufacturing and demand (China + Japan + Other Asia ≈ 71% of ship-to revenue) — a live tariff/geopolitical exposure.
The moat is specialty analog + magnetic-sensing IP, not raw scale.
Bargaining power: moderate and asymmetric. Over customers — decent within a design (high switching cost) but weak at the negotiating table vs. Tier-1 auto buyers who squeeze ASPs every year. Over suppliers — weak; it is a sub-scale fabless buyer competing for specialty-node capacity against far larger analog peers (TI, ADI, Infineon). The moat is the IP and the socket, not the supply chain.
Allegro reports one reportable segment, so the breakouts below are product/market/geography splits, all ``:
By product (FY26 vs FY25):
| Product | FY26 $ | FY25 $ | YoY |
|---|---|---|---|
| Magnetic Sensors (MS) | $538.5M | $474.6M | +13.5% |
| Power ICs (PIC) | $351.6M | $250.4M | +40.4% |
| Total | $890.1M | $725.0M | +22.8% |
By market (FY26 vs FY25):
| Market | FY26 $ | FY25 $ | YoY |
|---|---|---|---|
| Automotive | $628.6M | $535.2M | +17.4% |
| Industrial & other | $261.5M | $189.8M | +37.8% |
Read: Power ICs and Industrial are accelerating much faster than the core — that is the data-center/AI-power story showing up in the mix. The cause is explicit in the MD&A: data-center applications, industrial automation, robotics, plus Focus-Auto (ADAS/xEV) recovery. By geography, Greater China (+36%) and Other Asia (+75%) drove the rebound; the US was flat-to-down (−1.1%) and Japan −1.9% — the recovery is Asia-led, which is both the opportunity and the geopolitical risk.
Trend caveat: these are off a depressed FY25 base. The right denominator is FY24 peak ($1,049M) — FY26 revenue is still ~15% below the prior cycle peak. This is a recovery, not yet a new high-water mark.
Three-year P&L (FYE late-March; $000):
| FY2026 | FY2025 | FY2024 | |
|---|---|---|---|
| Total net sales | 890,096 | 725,006 | 1,049,367 |
| Gross profit | 411,970 | 321,527 | 574,529 |
| Gross margin | 46.3% | 44.3% | 54.7% |
| Operating income (loss) | 18,487 | (19,802) | 196,244 |
| Net income (loss) attrib. ALGM | (14,897) | (73,010) | 152,697 |
| Diluted EPS (GAAP) | $(0.08) | $(0.39) | $0.78 |
The shape is a textbook semis cycle: FY24 peak → FY25 trough (operating loss, −$0.39) → FY26 recovery (back to operating profit, near-breakeven net). FY26 still printed a small GAAP net loss of −$0.08 — but that is below the operating line: $22.1M interest expense on the term loan plus a $9.4M equity-method loss on Polar/PSL dragged a +$18.5M operating profit into a −$14.9M net loss. Operationally the business turned; the reported bottom line is still cleaning up financing + JV noise.
Non-GAAP tells the cleaner story: FY26 non-GAAP EPS $0.54, up 125% from FY25's $0.24. The GAAP↔non-GAAP gap (≈$0.62/sh) is SBC (~$47.9M ), intangible amortization, and the JV/financing items — large but mostly legitimate for a recently-IPO'd, acquisitive analog name.
Latest quarters (the acceleration):
Guidance (Q1 FY27, qtr ending Jun 26 2026): net sales $245–255M (~+23% YoY midpoint), non-GAAP GM 50–51% (a step up toward the historical ~55% target), non-GAAP EPS $0.19–0.23. Management said data center should grow >20% in FY27.
Balance-sheet flags (FY26): cash $168.8M, total debt $285.0M (2026 Refinanced Term Loan), net debt ≈ $117M — modest leverage (~0.7x on recovering EBITDA). OCF $163.1M (up from $61.9M), capex only $38.2M → FCF ≈ $125M. Note the capex collapse: FY24 capex was $124.8M (building/expanding capacity) and is now ~$38M — under-investment that flatters near-term FCF but bears watching if the upcycle demands capacity. Working capital $357.7M; no receivable/inventory blow-out flagged.
Market reaction — the key tell: despite the Q4 beat, the stock fell ~16% to ~$43 on the May-7 print — a "sell the news / valuation digestion" move after a huge run. It then rallied to ~$59 by mid-June on the subsequent quarter's data-center momentum and a Jefferies PT hike to $62. Translation: this name now trades on data-center narrative and multiple, and is volatile to it.
No transcripts on disk (transcripts/ empty) — sentiment is reconstructed from the prints and press ``.
Tone arc over the last ~4 quarters: trough-defensiveness (mid-FY25) → "inflection" language → confident "broad-based recovery" (FY26 Q3/Q4). The recurring phrases management now leans on: "Focus Auto" (ADAS + xEV content growth), "data center" (the new hero — explicitly called the standout, >20% FY27 growth), "TMR" and current-sensing leadership, "design wins," and gross-margin recovery toward the 55% model. What they stopped saying: the FY25 inventory-correction / channel-destocking refrain has largely dropped out. New structural message: a sales-force reorganization by end market rather than geography (FY26) — a deliberate pivot to chase data-center/industrial growth pockets. Net sentiment: improving and increasingly growth-forward, with the caveat that the market punished the very-high expectations on the May print.
Allegro vs. analog/sensing peers.
| Company | Ticker | Mkt cap | EV/Sales | EV/EBIT(DA) | P/E | Div yield | Note |
|---|---|---|---|---|---|---|---|
| Allegro | ALGM | ~$11.0B | ~12x | ~110x P/E trailing non-GAAP | ~110x (non-GAAP TTM) | 0% (none) | Recovery + data-center premium |
| Texas Instruments | TXN | ~$278B | n/a | n/a | n/a | ~2–3% | Scale analog benchmark |
| Analog Devices | ADI | n/a | n/a | EV/EBITDA ~33.6x; P/E ~63.6x | as left | ~1.7% | Closest quality comp |
| onsemi | ON | ~$27B | n/a | n/a | n/a | 0% | Auto/industrial + SiC overlap |
| NXP | NXPI | n/a | n/a | n/a | n/a | ~1.7% | Auto-analog peer |
| Microchip | MCHP | n/a | n/a | n/a | varies | ~2% | MCU/analog, deeper trough |
| Industry (semis) | — | — | — | ~17–18x EV/EBITDA TTM | ~25–26x | — | Broad benchmark |
Read: on trailing numbers ALGM is wildly expensive — ~110x non-GAAP P/E vs an industry ~25x and even vs ADI's already-premium ~64x. That multiple is a bet on the forward trough-to-recovery EPS ramp, not trailing reality. On a forward run-rate (annualizing the $0.19–0.23 guide ≈ $0.85–0.95 FY27 non-GAAP EPS ), ALGM is ~62–70x forward — still a clear premium to the group, justified only if data-center + auto compound 20%+ with margins recovering to 55%. The comp verdict: priced as a growth-semiconductor, not as the cyclical analog supplier it has historically been.
Mostly ``:
What the market actually reacts to: (1) the analog/auto cycle turning (revenue direction), (2) the data-center/AI-power narrative (the new multiple driver), (3) Sanken supply (overhang/technical), and (4) guidance vs. the embedded premium — at ~110x, even beats can sell off. Macro/rate moves matter via the multiple, not the fundamentals.
Forensic lens. Every figure `` unless noted.
Regulatory findings (required sub-section):
regulatory/regulatory-findings.md (generated 2026-06-22 via SEC EDGAR EFTS) returned 0 LR and 0 AAER naming Allegro MicroSystems for 2021-06-22 → 2026-06-22.Bottom-up from FY26 actuals + the Q1-FY27 guide. All outputs ``; inputs labeled. (No forecast.ts logged — watchlist breadth mode.)
Anchors: FY26 revenue $890M, non-GAAP EPS $0.54. Q1-FY27 guide: $245–255M rev (~+23% YoY), non-GAAP EPS $0.19–0.23, GM recovering to 50–51%. Run-rating the guide midpoint ($250M/qtr, $0.21/qtr) ≈ ~$1.0B revenue / ~$0.84 non-GAAP EPS FY27 baseline before further ramp.
| Scenario | FY27E rev | FY27E non-GAAP EPS | FY28E EPS | FY29E EPS | Logic |
|---|---|---|---|---|---|
| Bear | ~$960M (+8%) | ~$0.70 | ~$0.80 | ~$0.90 | Auto cycle stalls, data-center decelerates, GM stuck ~50%; ASP pressure from Tier-1s |
| Base | ~$1.02B (+15%) | ~$0.90 | ~$1.20 | ~$1.55 | Data center >20%, auto +mid-teens, GM → 53–55%, operating leverage on flat-ish opex |
| Bull | ~$1.10B (+24%) | ~$1.10 | ~$1.55 | ~$2.10 | Sustained DC ramp + xEV content + GM back to 55% peak; multiple holds |
Mechanics behind base: revenue +15% (DC growing >20% off a 14% base + auto mid-teens) → ~$1.02B; GM 53.5% → gross profit ~$546M; opex held ~$400M as the FY26 sales-force reorg drives efficiency → operating income ~$146M; less ~$18M interest, less Polar noise, ~12% tax → net ~$110M; ÷ ~187M diluted shares ≈ ~$0.59 GAAP / ~$0.90 non-GAAP. By FY29 the peak-cycle analog model (op margin high-20s%, GM 55%) gets you toward $1.50+ non-GAAP — which is the only path that retro-justifies today's multiple.
Brier seed (not logged): "ALGM FY27 (FYE Mar-2027) non-GAAP EPS ≥ $0.85, p ≈ 0.60" — the guide run-rate alone roughly clears it; the risk is a second-half auto air-pocket.
Bull case. A genuine analog-IP franchise (#1 in magnetic sensors, near-unique production TMR current-sensing) is simultaneously levered to two secular content-growth waves — vehicle electrification/ADAS and AI-data-center power — having just broken free of parent control (Sanken). Power ICs +40% and Industrial +38% in FY26 show the data-center mix-shift is real and high-margin; management guides data center >20% growth and GM recovering toward the 55% peak model. Operating leverage off a fixed analog cost base means EPS can compound faster than revenue. If TMR becomes the default current-sensor for SiC/GaN power stages, Allegro owns a structurally growing, ASP-rich socket in every AI server power supply and every xEV inverter. The Sanken separation removed a control discount; a clean float re-rates.
Bear case (permanent-impairment lens). (1) Valuation is the risk. ~110x trailing / ~65x forward non-GAAP for a business still 15% below its prior peak revenue and structurally exposed to the auto cycle — the stock has priced the entire recovery and the data-center dream. A single guide-down de-rates it 30–40% (the May −16% on a beat is the warning shot). (2) Sub-scale vs. giants. TI ($278B), ADI, Infineon can out-invest, out-foundry, and out-price Allegro in any socket they prioritize; current-sensing is attractive enough that they will. (3) Sanken overhang. A ~32% holder with its own financial pressure is a perpetual seller into strength — a structural cap on the stock and a governance complication. Pre-mortem (18 months out, thesis broke): auto demand rolled over in late CY2026, data-center current-sensing got commoditized by TI/ADI second-sources, GM stalled at 50%, Sanken dumped another tranche — and a ~110x stock that needed everything to go right round-tripped to the low-$30s. Contrarian view the market is refusing to see: the bull narrative quietly re-rated ALGM from "cyclical auto-analog supplier" (deserves ~15–20x) to "AI-data-center compounder" (gets ~50x+) on a segment that is still only ~14% of sales — the category re-label, not the fundamentals, is doing most of the work in the price.
Dismantling the bull case.
SKYT is no longer a foundry — it is an IonQ deal-stub trading $38.40 vs a $35.00 closed-vote merger value, i.e. a +10% premium that is pure leveraged FTC-cleared bet on IonQ rallying past $60.13; the standalone foundry (43% Infineon, negative op cash flow, $22M cash) is the break-case floor, not the thesis.
The MEMS-timing monopolist is being bid as if it has already won the AI clock — at ~35x sales and ~85x forward earnings, a flawless +88% quarter and the transformational Renesas deal are not just priced in, they are required.
This is no longer a chip stock — it's a near-closed merger-arb. TI buys SLAB for $231 cash; HSR cleared and shareholders approved, only China SAMR left, ~6% gross spread to close in 1H2027. The trade is deal certainty, not IoT growth.