Monolithic Power Systems, Inc. (NASDAQ: MPWR) — A Best-in-Class Power Compounder Priced for an AI Story Its Own Numbers Just Stalled

An independent fundamental research note Report date: 2026-06-11 · Sector: Technology — Analog / Power-Management Semiconductors · CIK 0001280452 Price ~$1,473 · Market cap ~$72.4B · Enterprise value ~$71.0B · ~49.1M shares · Net cash ~$1.35B

⚡ Claude’s Take

This block is the author’s own independent, subjective opinion. It is general information, not investment advice, and not a recommendation to buy or sell any security. The analysis that follows takes no position, renders no price target, and stands on its own evidence.

Verdict: HOLD / not-a-short — a genuinely elite business at a demanding price. “Great house, frothy street.” For owners: hold and let it compound, but trim into euphoria. For new money: AVOID here; accumulate-on-weakness in roughly the $850–1,100 zone (~30–38x forward EPS / mid-50s EV/EBITDA), and back up the truck only on a real cyclical or AI-air-pocket break toward the high-$600s–$700s (the 52-week low was $670). Medium conviction.

Monolithic Power is one of the best fabless semiconductor businesses in the world — ~25% organic revenue CAGR for seven years, fourteen straight years of growth, stable ~55% gross margins, ~50%+ returns on the operating capital that actually matters, a fortress net-cash balance sheet, near-zero goodwill, and a fast-growing, well-covered dividend. None of that is in dispute. The problem is entirely the price and the narrative gap behind it. At ~24x EV/sales, ~83x EV/EBITDA, ~49x forward earnings and a ~0.9% free-cash-flow yield, the stock sits in the 94th percentile of its own ten-year valuation and a ~60% EV/sales premium to direct analog peers ADI and TXN — despite lower gross and operating margins than both. That premium is an AI/data-center-power bet. Yet the very segment the bet rests on — “Enterprise Data” — declined ~2% in 2025 and shrank from 32.5% to 25.2% of revenue, while the non-AI book (auto +43%, storage/compute +46%, comms +37%) carried the entire +26% year. Management, tellingly, refuses to claim AI-power share gains while Infineon, TI and ADI tout data-center-power businesses “doubling.” You are paying a top-decile price for a re-acceleration that, as of the latest data, is contested and unproven.

The framing is quality-compounder-at-the-wrong-price, not a short (the business is too good and the balance sheet too strong to bet against, and a re-accelerating AI leg is a live possibility). The reverse-DCF says the market is underwriting ~15–22% revenue growth for the better part of a decade with margin expansion and a permanently premium multiple — essentially extrapolating the past. From here the math is unkind: my scenarios put the bear at ~−50%, the base at roughly flat, and even the bull at only ~+6–10%/yr, because you start at a sub-1% cash yield with no margin of safety. Conviction: medium. The single fact that would flip me decisively bullish: two consecutive quarters of re-accelerating Enterprise Data with management quantifying AI-power socket share gains. The single fact that would flip me bearish: another two quarters of flat/down Enterprise Data and a gross-margin slip below ~53%, signaling the data-center-power contest is being lost on price. One more flag that keeps conviction from being higher: an unremediated, auditor-confirmed material weakness in tax-accounting controls that forced a ~$195M restatement of 2024 earnings, paired with zero insider open-market buying and aggressive CEO selling into all-time highs.

1. Executive Summary

Monolithic Power Systems is a fabless designer of high-performance analog and mixed-signal power-management semiconductors — DC-to-DC converters, power-management ICs, and increasingly integrated chip-plus-passive modules — sold into storage & computing, enterprise/AI data centers, automotive, communications, consumer, and industrial markets, primarily through distributors. It is, by the financial evidence, a top-tier business: revenue compounded from $582M (2018) to $2,790M (2025) at roughly a 25% organic CAGR, fourteen consecutive years of growth, gross margins parked in a tight ~55% band, ~26–30% operating margins after fully expensing heavy stock comp, free cash flow of ~$666M, a net-cash (~$1.35B) and effectively debt-free balance sheet, and returns on tangible operating capital comfortably north of 50%. It has built this almost entirely organically — goodwill is a trivial $25.9M — which removes the integration and impairment risk that dogs acquisitive peers like Analog Devices.

The investment debate is not about quality; it is about price and the durability of the AI-power narrative driving it. MPWR trades at ~24x EV/sales, ~83x EV/EBITDA, ~49x forward earnings, ~106–116x clean trailing earnings, and a ~0.9% FCF yield — the 94th percentile of its own decade and a wide premium to ADI/TXN (~14–15x EV/sales) despite lower margins than either. The premium is a wager on data-center/AI power. The uncomfortable fact: the Enterprise Data segment (cloud/AI systems) fell ~2% in 2025, dropping from 32.5% to 25.2% of revenue, while the company’s non-AI end markets grew ~40% and carried the +26% consolidated result. Management declines to claim AI-power share gains; Infineon, TI and ADI are all attacking the same sockets and tout data-center-power “doubling.” MPWR’s moat — a proprietary BCD process run on merchant foundries, plus module integration and design-in switching costs — is genuine and durable in automotive and industrial, but thinnest exactly where the multiple lives.

Two quality-of-earnings and governance items deserve front-page billing. First, 2024 GAAP net income of ~$1,592M (restated; $1,787M as first reported) is economically meaningless: pretax income was only $573M; the gap is a ~$1.1B non-cash deferred-tax benefit from a future foreign tax holiday — which was then restated down ~$195M under a newly disclosed, still-unremediated material weakness in income-tax accounting controls (auditor-confirmed not-effective at both 2024 and 2025 year-ends). FY2025’s $621.5M is the only clean earnings base. Second, insiders are one-directional sellers — zero open-market purchases, with CEO/founder Michael Hsing selling >$200M of stock (some discretionary, not 10b5-1) into the all-time-high rally.

The embedded-expectations read: the price underwrites a continuation of the historic ~25% trajectory — ~15–22% revenue CAGR for 5–10 years with margin expansion and a permanently premium multiple — even though the marquee growth driver just stalled. Scenario asymmetry from today’s price is unfavorable (bear ~−50%, base ~flat, bull ~+6–10%/yr). This memo takes no position and sets no target; that work is reserved for Claude’s Take above.

2. Business Overview

What MPWR sells. Monolithic Power Systems (incorporated 1997; ~4,500 employees; fabless) designs high-performance analog and mixed-signal power-management integrated circuits and integrated power solutions. The core franchise is DC-to-DC conversion — the chips and modules that step, regulate and distribute voltages inside electronic systems — complemented by AC-to-DC products, driver MOSFETs, power-management ICs (PMICs), current-limit switches, and lighting-control devices. The catalog spans roughly 4,000+ products sold to ~30,000 end customers. (FACT — FY2025 10-K, Item 1, filed 2026-02-27.)

The strategic arc that matters for the thesis is MPWR’s deliberate climb up the value chain: from selling discrete chips, to selling modules (the IC co-packaged with inductors and capacitors into a complete power block), to selling board- and rack-scale power systems. Management frames this as moving “from a chip company to a full-service silicon-based solution provider.” (FACT — Q1-2026 earnings call, 2026-04-30; Analyst Day, 2025-03-20.) This matters because each rung up the ladder raises the dollar content per design and deepens the customer’s dependence on MPWR’s specific solution.

The technology core — proprietary BCD process on merchant foundries. Unlike most fabless companies, which use a foundry’s standard process and design kit, MPWR developed its own proprietary process technology (BCD — Bipolar-CMOS-DMOS) and packaging, and installs that technology on partner-foundry equipment for its exclusive use. This is the crux of the model: MPWR captures IDM-like process differentiation at fabless capital intensity — it gets a differentiated power process without owning multi-billion-dollar fabs the way Texas Instruments, Infineon, NXP or STMicroelectronics do. The financial signature of that advantage is ~55% gross margins on near-trivial owned-fab capex. (FACT/INTERPRETATION — 10-K Manufacturing & R&D; Analyst Day 2025-03-20.)

Revenue by end market (FACT — FY2025 10-K MD&A):

The single most important and most overlooked fact in this report sits in that table: Enterprise Data — the cloud/AI-systems segment that the entire AI-power narrative is built on — fell ~2% in 2025 in absolute dollars, dropping from 32.5% to 25.2% of mix. The +26% consolidated growth was carried entirely by the non-AI book: storage & computing (+46%), automotive (+43%), communications (+37%), industrial (+35%). Management acknowledged this directly on the Q4-2025 call (“our non-Enterprise Data end markets grew by over 40% year-over-year”). Storage & Computing does carry AI-adjacent content (graphics-card and AI-server memory power), so the company’s total AI exposure is broader than the Enterprise Data line alone — but the pure AI-systems segment did not grow in 2025. (FACT — 10-K; Q4-2025 call, 2026-02-06.)

Business model and customer concentration. MPWR sells predominantly through third-party distributors, not direct. This concentrates revenue: in 2025, three distributors accounted for 26%, 18% and 10% of total revenue (a combined ~54%); in 2024 two distributors were 31% and 20%. No single direct end customer exceeded 10% — but that is partly an artifact of the distributor model masking the true hyperscaler/OEM concentration sitting behind the channel. Roughly 92% of revenue ships to Asia (the electronics-assembly hub), billed in USD. Lead times run 16–26 weeks, and MPWR builds inventory to forecast — creating the double-order/destock risk that produced the flat 2023. (FACT — 10-K.)

Verdict (Business Overview): A focused, high-margin, fabless power-semiconductor franchise with a genuine technology differentiator and a coherent move up-stack into modules and systems — but with meaningful channel concentration and an AI-power growth line that, contrary to the popular framing, stalled in the most recent year.

3. Industry Dynamics

Analog/power management is structurally the best neighborhood in semiconductors. The 10-K names the attractions precisely, and they line up with the framework’s “structurally good industry” tests: long product lifecycles (MPWR parts often ship for 10+ years), highly fragmented end markets (thousands of niche sockets, no single point of commoditization), technology that is difficult to replicate (analog design is an apprenticeship craft, not a node-shrink race), the primacy of experienced design engineers, low capital intensity, and end-market diversity. Crucially, analog sidesteps the two worst dynamics in the industry: the boom-bust commoditization of memory and the winner-take-all, ruinous-capex treadmill of leading-edge logic. There is no Moore’s-Law cadence forcing perpetual node migration; a well-designed power IC sells for a decade with only gentle ASP erosion. (FACT/INTERPRETATION — 10-K; corroborated by peer ADI and TXN disclosures.)

The durable financial proof of the industry’s quality is visible across the group: ADI runs ~64% gross / ~38% operating margins, TXN ~57%/~38% even off a cyclical trough, NXP ~55%/~25–28%. MPWR’s ~55%/~26–30% sits at the high-volume-power end of that band — genuinely good economics, though below the high-performance signal-chain leaders. (FACT — public comparables and peer filings, 2026-06.)

Market size and the AI-power TAM. The broad analog market is ~$93–94B (2025), with the top five (TI, ADI, ST, Infineon, NXP) holding roughly half and a long fragmented tail beneath. MPWR is a small-share player in a huge fragmented pie — management’s recurring framing is that even in served markets it holds “such a small percent” that it is “all just green space.” That cuts both ways: a long runway, but also confirmation that MPWR holds no scale-leadership position in the broad market. (FACT — Analyst Day 2025-03-20.)

The differentiated vector is data-center power, which management dramatizes through “Jinghai’s Law”: every 2–3 years AI/GPU power roughly doubles while the board area available for power delivery halves. That forces architectural change — 12V → 48V distribution, then vertical power delivery (power stacked directly beneath the processor), and now toward 800V rack-scale power as racks scale from ~100kW toward ~600kW and GPUs blow past 1,000W. Each step is a content-expansion opportunity (more silicon, higher-value modules, silicon carbide for the high-voltage conversion stage). This is a real secular driver, and MPWR was an early architect of it. (FACT/INTERPRETATION — Analyst Day; recent calls.)

The Marathon capital-cycle warning. Here the industry lens turns cautionary. High returns attract capital, and that is happening in AI power now: Infineon, TI, ADI, Renesas, ON Semi, Vicor, plus hyperscaler and NVIDIA in-house power teams and Chinese entrants are all pouring R&D and capacity into data-center power. On the Q4-2025 call, an analyst noted that “some of your competitors in the AI power space are talking about their business…doubling in 2026” and asked whether MPWR is gaining share; CEO Hsing declined to engage (“I refuse to get into a pissing contest”). MPWR itself is adding capacity into the boom, raising its manufacturing-capacity target from $4B toward “$6B in the near future” — classic late-cycle supply addition. The 2025 Enterprise-Data flatline is the early financial fingerprint of this crowding. (FACT — Q4-2025 and Q1-2026 calls; INTERPRETATION via Capital Returns framework.)

Verdict (Industry): The broad analog/power-management industry is structurally excellent — consolidated-ish, high-barrier, secularly growing on electrification, automation and AI. But the specific sub-segment that justifies MPWR’s valuation — AI/data-center power — is the least captive, most contested, most capital-attracting corner, where the normal capital cycle is in full swing. Good industry; hottest, most crowded room in it.

4. Competitive Position

Naming the moat (Greenwald taxonomy). MPWR’s advantage is a fusion of proprietary intangibles (the BCD process, packaging know-how, and accumulated analog/system design expertise) and demand-side customer captivity (design-in switching costs). It does not possess a true economies-of-scale moat — it is the smallest of its named competitors and holds low share in most markets, so scale is not the source of its edge. That places the moat on Greenwald’s two weaker-to-middle pillars: supply-side proprietary technology (which the framework warns is the most transient advantage — process leads get matched, “in the long run everything is a toaster”) and demand-side switching costs (more robust, but graded by end market).

The genuine, financially-evidenced edges:

• BCD-on-foundry process advantage (FACT/INTERPRETATION). Running a proprietary power process on merchant foundry capacity captures IDM-like differentiation at fabless capital intensity. The proof is in the numbers: ~55% gross and ~30% operating margins with negligible owned-fab capex — economics IDM peers achieve only by carrying multi-billion-dollar fabs. This is real and visible. It is also the erodable kind: SG Micro, 3Peak and the IDMs are pushing their own BCD nodes, and CEO Hsing’s “we’re far ahead” (Analyst Day) is an unverifiable hypothesis, not proof of durability.
• System-level module integration (FACT — 10-K; Analyst Day). Co-packaging chip + inductor + capacitors into a module solves the physical problem of no board space under a 1,000W+ GPU. Record module revenue in 2025. This is the strongest part of the moat: the customer designs the whole power subsystem around MPWR’s module, raising both switching cost and dollar content.
• Design-in switching costs (graded — INTERPRETATION). Captivity is strongest in automotive (multi-year qualification, functional safety, 10–20-year platform life — auto grew +43% in 2025 on sticky ADAS/infotainment content) and industrial, and weakest exactly where the valuation lives: merchant data-center/AI power, which is multi-sourced and roadmap-sensitive. The 2025 Enterprise Data flatline is the moat’s thinness made visible.

Pressure-testing the AI-power position. The 10-K names the competitive set: Analog Devices, Infineon, NXP, ON Semi, Power Integrations, Renesas, ROHM, Semtech, STMicroelectronics, Texas Instruments — plus Vicor specifically in data-center power and hyperscaler/NVIDIA in-house teams. MPWR was the first mover in multi-stage 48V/vertical power delivery, a real, validated technical lead. But the bear read has the data on its side: (i) Enterprise Data −2% in 2025 while rivals tout AI-power “doubling”; (ii) management declines to claim share gains; (iii) the segment is structurally merchant/multi-sourced; (iv) MPWR’s actual position at NVIDIA/hyperscaler platforms is unconfirmed — management never names customers (Hsing, Q1-2026, asked to quantify AI-accelerator share: “That’s a good try…No, I’m not going to give it to you”). Industry observers consistently characterize merchant data-center power as multi-sourced turf where design wins can shift and a single hyperscaler roadmap change can move the line. OPEN QUESTION: What is MPWR’s real, current content/share position in next-gen AI-power sockets, and is it rising or falling versus Infineon? Treat any “MPWR is the NVIDIA power winner” claim as unproven until corroborated by teardowns/channel checks.

Greenwald tests. ROIC/economics test: passes — sustained ~55% gross margins, ~30% operating margins, 50%+ ROIC, and 14 consecutive growth years confirm a barrier exists. Share-stability test: mixed/fails at the segment level — the violent swing in Enterprise Data’s share of mix (17.7% → 32.5% → 25.2% in three years) is the opposite of stable share, and is itself evidence that the AI-power position is contestable. The durable, stable-share core is automotive/industrial/broad power, not AI.

Verdict (Competitive Position): A real but graded and partly-erodable moat — durable in the diversified power core, contested and unproven in the AI-power segment that drives the multiple. MPWR has a genuine advantage that produces best-in-class fabless economics and a 14-year growth record, but it is not the impregnable scale-plus-captivity fortress of a TI or the high-ASP precision moat of an ADI. It rests on Greenwald’s two weaker pillars, and the part of the business the market pays the richest multiple for sits precisely where the moat is thinnest and the capital cycle hottest.

5. Growth History and Forward Opportunities

Historical growth — exceptional and almost entirely organic (FACT — EDGAR/10-K). Revenue: $582M (2018) → $628M (2019) → $844M (2020) → $1,208M (2021) → $1,794M (2022) → $1,821M (2023, flat — the destocking downcycle) → $2,207M (2024) → $2,790M (2025). That is a ~25% CAGR over 2018–2025 (roughly 5x) and ~27% over 2020–2025. Q1-2026 grew +26% YoY. Critically, this is organic — MPWR has made essentially no material acquisitions; the only deal of note is the immaterial January-2024 Axign tuck-in (Dutch Class-D audio, source of the trivial $25.9M goodwill). Growth has come from new products on its own process, content expansion, and end-market diversification. The Marathon “asset-growth anomaly” red flag does not apply: MPWR grew without serial debt-funded M&A, which means clean returns on the capital actually deployed and no integration overhang.

Forward drivers (management hypotheses — Analyst Day & recent calls):

1. Data-center power content expansion — 48V → vertical power → 800V rack conversion, silicon carbide for high-voltage stages, more silicon per ever-hotter GPU/rack. The structural driver — but, as 2025 proved, lumpy and contested.
2. Automotive — 48V/zonal architectures, ADAS, infotainment; +43% in 2025, a multi-billion-dollar SAM, captive and durable — the highest-quality growth line.
3. Enterprise CPU servers — “plain server” cited as a 2026 tailwind (Q1-2026).
4. Communications — optical modules/switches (AI-networking adjacent), +33% sequentially in Q1-2026.
5. New adjacencies — robotics/automation, high-speed interface (first DDR5 interface products sampled Q1-2026), continuing the chip → module → system march.
6. Capacity — the $4B → $6B manufacturing-pipeline goal.

Quality of growth (INTERPRETATION). High-quality on three counts: organic (no acquired-growth illusion), diversified (2025 proved the non-AI book can carry 26% growth when AI stalls — genuine resilience), and content-driven (rising $/unit, strong incremental economics). Two honest qualifiers: (i) the AI-power line is cyclically and competitively unreliable — the 2024→2025 stall shows the marquee vector is not a smooth secular ramp; (ii) MPWR is adding capacity into a hot capital cycle, and a hyperscaler-capex digestion or share loss to Infineon/TI would expose that build.

Verdict (Growth): High-quality, organic growth with a long runway off a still-small base — but the market is extrapolating the AI-power vector more confidently than MPWR’s own 2025 numbers, or its own management’s commentary, support.

6. Financial Quality

Income statement (FACT — 10-K; XBRL-reconciled), $000s:

Gross margin is the cleanest quality signal here: MPWR peaked at 58.4% in 2022 (the supercycle), then settled into a tight ~55.2–55.3% band through a destocking cycle and a 26% recovery — remarkably stable for a fabless analog firm and squarely in management’s ~55–56% target. The honest caution: 55% now looks like the ceiling, not a floor; with no gross-margin lever left, all operating leverage must come from opex. Operating margin compressed from the 29.4% 2022 peak to 24.4% (2024, over-investment through the trough) then re-levered to 26.1% (2025) and a strong 30.0% in Q1-2026 — proof the operating-leverage thesis is real when revenue accelerates, though one quarter is not a trend. R&D intensity (~13–14% and rising in absolute dollars) is appropriate for an IP-differentiated franchise but caps margins below ADI/TXN. (FACT/INTERPRETATION.)

The 2024 tax anomaly + restatement — the QoE crux. This is the most important number in the file, and it has two layers.

• The benefit. In 2024, a foreign subsidiary was granted a ten-year tax incentive beginning tax-year 2025, and MPWR recorded a ~$1.1B non-cash deferred-tax benefit to reflect estimated future cash-tax reductions (the 2024 foreign-deferred line was −$1,106M). This is not an intra-entity IP transfer or a valuation-allowance release; it is the upfront recognition of a deferred-tax asset for a future tax holiday. (FACT — 10-K Note 17.)
• The restatement. Preparing the FY2025 10-K, MPWR found it had mis-accounted for those deferred taxes and restated 2024 down $194.6M:

MPWR disclosed a material weakness in internal control over financial reporting over income taxes, restated all three 2025 interim 10-Qs, and stated prior filings “should no longer be relied upon.” (FACT — 10-K Note 2, Item 9A.)

The bottom line: 2024 pretax income was only $572.9M; the entire gap to the $1,592M reported net income is a non-cash, balance-sheet deferred-tax entry monetizing a future holiday — not earning power. Normalize 2024 at a ~15% rate and clean net income is ~$487M, overstating true earning power by ~3.3x at the headline. Any trailing P/E running through 2024 is garbage; FY2025’s $621.5M is the only clean GAAP base — putting the stock at ~116x trailing. The “cheap-looking” sub-50x prints some screens show are an artifact of the 2024 windfall. (FACT/INTERPRETATION.)

Stock-based compensation. SBC is heavy — ~8–10% of revenue, ~$227M in 2025 (2022 $161M, 2023 $150M, 2024 $206M) — but is fully expensed in GAAP operating income, so the 55% gross and 26% operating margins are after SBC (a point in MPWR’s favor versus peers who quote ex-SBC). The caution is on any company- or sell-side “adjusted” EBITDA/EPS, which adds ~$227M back and inflates adjusted operating income ~30% over the $729M GAAP figure. Net dilution is modest (the buyback mops it up — see §7), which converts the equity cost into a real cash outflow rather than runaway share growth. (FACT — XBRL.)

Cash flow and FCF (FACT — XBRL), $000s:

FCF is robust (~$666M in 2025) and converts at ~1.07x of clean net income — high quality, no cash/accrual divergence (the cash-flow statement correctly strips the non-cash 2024 tax benefit). Two watch-items: (i) the 2022 working-capital build (inventory $259M→$447M) depressed FCF that year, and inventory is now at an all-time-high $565M (~165 days) on the AI/data-center pull — a demand air-pocket would hit OCF the way it did in 2022; (ii) capex has stepped up to ~6% of revenue (from ~3% historically) and ~9% in Q1-26, building owned test capacity and the tax-incentive-jurisdiction footprint — the first crack in the asset-light story, though still far below a foundry. (FACT/INTERPRETATION.)

Balance sheet. Total assets $4,194M; equity $3,531M; liabilities $663M. Cash + short-term investments ~$1,257M (yfinance “total cash” ~$1,367M incl. long-term investments); debt negligible (~$20M) → net cash ~$1.35B. Goodwill just $25.9M, intangibles immaterial → tangible book ≈ stated book ≈ $3.5B, with no goodwill cushion to write down. The one judgment-laden asset to watch is the ~$1.1B net deferred-tax asset from the incentive — the very item that was restated and sits under the material weakness; strip it and tangible equity is ~$2.4B. (FACT — 10-K.)

Returns. ROE on the clean 2025 base is ~19.2% — solid but understated by the equity-heavy, net-cash, DTA-inflated balance sheet. The honest read for a fabless model is ROIC: ~$1.05B of true operating capital (equity less cash less DTA) against ~$580M NOPAT → ROIC comfortably 50%+. This is the core quality fact — a fabless analog franchise earning 55% gross margins on a tiny operating-capital base generates extraordinary returns on the capital that matters; the modest ~19% ROE is an artifact of hoarding cash and the DTA, not weak economics. (FACT/INTERPRETATION.)

Verdict (Financial Quality): Economics improve with scale and are genuinely excellent — stable ~55% gross margins, re-leveraging operating margins, ~$666M FCF at ~1.1x conversion, 50%+ ROIC, fortress balance sheet. The defining blemish is the tax line: a one-time $1.1B benefit, a $195M restatement, and an unremediated material weakness in the single most judgment-heavy area. Operating earnings are clean; reported earnings (2024) are not — use FY2025.

7. Capital Allocation

Dividends — a 14-year raise streak, fast-growing and well-covered (FACT — 10-K XBRL). Declared DPS: ~$2.00 (2020) → $4.00 (2023) → $5.00 (2024) → $6.24 (2025), with the Board raising the quarterly dividend $1.56 → $2.00 in Feb-2026 ($8.00 annualized, +28%). On clean FY2025 net income of $621.5M the ~$285M paid is a ~46% payout; on FCF (~$666M) ~43%. The 2026 step-up (~$390M run-rate) lifts the payout toward ~55–60% but remains comfortably covered by FCF and the ~$1.35B net-cash cushion. This is a safe, fast-compounding dividend — not a stretched one.

Buybacks — lumpy, opportunistic, currently dormant (FACT — XBRL; 10-K Issuer Purchases): 2021 $0; 2022 $0; 2023 $3.7M; 2024 $636.2M (the big year); 2025 $6.6M (~8,000 shares). A new $500M authorization (Feb-2025 through Feb-2028) is ~99% unused (~$493M remaining at 12/31/2025). The 2024 slug — deployed into a year when the stock had de-rated — was a genuinely countercyclical, well-timed move. But the overall pattern is reactive: zero in richly-valued 2021–22, a huge 2024 window, then near-total withdrawal as the stock ran from ~$650 to >$1,500. Management is not repurchasing at $1,000–1,700, which is consistent with price discipline (good) but leaves the authorization as optical cover rather than an active program. (FACT/INTERPRETATION.)

Share count. Diluted weighted shares: 48.77M (2023) → 48.84M (2024) → 48.31M (2025); outstanding ~47.9M (Oct-2025) → ~49.1M (Feb-2026) on RSU vesting. The count is essentially flat over three years — SBC dilution is roughly neutralized by the lumpy buybacks, but the float is not meaningfully shrinking. This is “tread water on dilution,” not aggressive per-share-value creation.

M&A — “build, don’t buy” (confirmed positive). Goodwill is just $25.9M, the residue of the single small Axign deal (Jan-2024). There is no acquisition of consequence in the five-year corpus. This is a clear positive: effectively zero impairment/integration risk, and a sharp contrast with Analog Devices’ ~$27B of Maxim/Linear goodwill. MPWR compounded ~14 years of growth without buying it — validating the internal R&D engine and removing a major capital-allocation landmine.

R&D/capex as capital allocation. The build-don’t-buy strategy routes cash into the P&L as R&D (~$382M, ~14% of sales, growing ~18%/yr) rather than onto the balance sheet as goodwill — and at 50%+ ROIC, that reinvestment is the highest-return use of MPWR’s cash. Capex has roughly tripled off the 2023 base to ~$172M (~6% of sales), a deliberate shift toward more owned infrastructure; still light for the sector, but worth monitoring against the asset-light thesis.

Compensation & governance (FACT — 2026 DEF 14A). CEO Hsing’s 2025 total comp was ~$19.9M (~95% variable; ~$1M base). The annual cash bonus paid at 400% of target for 2025 (a 26%-growth year). PRSUs — 100% of the long-term equity — are tied to revenue and a new (2025–27) 3-year relative-TSR condition added “after considering stockholder feedback,” with a 500% maximum and clawback. The relative-TSR overlay is a genuine improvement, forcing a chunk of pay to clear a per-share-outperformance bar rather than rewarding raw growth alone — though the magnitude (400–500% leverage) is aggressive. Governance positives: single-class stock (one share, one vote) and a classified board being phased out by 2030. Negatives: Hsing holds combined Chairman/President/CEO roles (no independent chair), and a small $10M related-party VC-LP investment (connected to the EVP/GC) was disclosed in July-2025.

The material weakness — the real governance negative. The FY2025 10-K disclosed a material weakness in ICFR over deferred income taxes, concluded not effective at BOTH 12/31/2024 and 12/31/2025 (auditor E&Y concurring), forcing the ~$195M restatement of 2024. Remediation is “ongoing,” not complete — the weakness persisted through a second fiscal year-end. The Audit Committee responded correctly (concluded materiality, restated, disclosed, has a plan), but a live, unremediated, auditor-confirmed control failure in the most complex line is a genuine, current black mark. (FACT — 10-K Item 9A.)

Insider activity (Form 4 sweep — FACT). Across the Feb–Jun 2026 window (stock $1,100–1,700): zero open-market purchases (code P), and dominant open-market selling (code S) across CEO Hsing, EVP/GC Tseng, EVP Sciammas, EVP Xiao, the interim CFO, and directors. Hsing sold >150,000 shares for >$200M at $1,100–1,520; roughly half the sales (including most of his February tranche) carry 10b5-1 footnotes, but a material subset — including his large May-20-2026 sale — are discretionary (not 10b5-1). Insiders own ~3.7% (Hsing ~2.1%). The 10b5-1 framing softens the signal — these are scheduled diversification sales by long-tenured, concentrated holders — but the complete absence of any open-market buy at any price, plus discretionary CEO selling into all-time highs, is a mildly bearish conviction tell.

Verdict (Capital Allocation): Largely intelligent, with two real caveats. The philosophy is excellent and durable — build-don’t-buy (near-zero goodwill), heavy and productive R&D, a fast-growing well-covered dividend, a fortress balance sheet, and one genuinely countercyclical 2024 buyback. The caveats: the buyback is opportunistic-bordering-on-absent (authorization ~99% unused; capital under-distributes vs. FCF, letting cash idle), and the governance/insider signals lean cautious — combined CEO roles, a related-party investment, one-directional insider selling, and most seriously a live, unremediated material weakness that forced a 2024 restatement.

8. Changes and Headwinds — Last Two Years

• The AI inflection and then stall. Enterprise Data exploded from 17.7% of revenue ($323M, 2023) to 32.5% ($716M, 2024) on the AI/data-center build-out, then went flat-to-down (−2%) in 2025, falling to 25.2% of mix. The defining recent change — and the one the bull narrative underweights. (FACT — 10-K.)
• Diversified strength offsetting it. Automotive (+43%), storage & computing (+46%), communications (+37%) and industrial (+35%) carried 2025 — evidence of genuine end-market resilience, but also a shift in the growth source away from the marquee AI line. (FACT.)
• The 2024 restatement + material weakness (disclosed Feb-2026) — see §6/§7. A new, live governance/QoE development. (FACT.)
• Dividend acceleration to $8.00 for 2026 (+28%) and a new $500M buyback authorization (Feb-2025, largely unused). (FACT.)
• Capacity expansion — the $4B → $6B manufacturing-pipeline goal, and capex stepping to ~6% of sales: a strategic shift toward more owned infrastructure, into a hot capital cycle. (FACT.)
• New product vectors — first DDR5 high-speed interface products sampled (Q1-2026), SiC for rack-scale power, robotics/automation adjacencies — the chip → module → system march continuing. (FACT.)
• Competitive intensification — Infineon/TI/ADI and others escalating data-center-power investment, rivals touting “doubling,” management declining to claim share. (FACT/INTERPRETATION.)
• Insider distribution into the all-time-high rally; zero open-market buys. (FACT.)

Verdict (Changes): On balance these modestly weaken the bull thesis at the current price. The business is strengthening structurally (diversification, dividend, product breadth), but the specific driver justifying the multiple (AI power) stalled, competition intensified, a control failure surfaced, and insiders sold — none individually fatal, collectively a yellow flag against a top-decile valuation.

9. Risk Analysis (Risk Matrix)

The dominant risks are valuation de-rating (#1) and its proximate triggers — AI-power share loss (#2) and growth deceleration (#3) — which usually arrive together and require no business failure to inflict 40–55% downside. Catastrophic/total-loss risk is very low: net cash, no debt, profitable, diversified, no goodwill to impair. This is a quality-de-rating risk, not a solvency risk.

10. Valuation Discussion (Embedded Expectations)

No price target and no recommendation in this section — embedded-expectations and scenario framing only.

Multiples and peer comps (FACT — recomputed 2026-06-11):

The premium is enormous and only partly explained by fundamentals. MPWR trades at ~24x EV/sales versus ADI/TXN ~14–15x (a ~60% premium) and ~3.8x NXPI — despite lower gross and operating margins than ADI and TXN. On a PEG lens (~1.94 vs. ADI ~0.86, TXN ~1.35, NXPI ~0.73), MPWR is the most expensive name even after adjusting for growth, by roughly 2x. The honest read: MPWR’s multiple is a growth-duration-and-AI-optionality premium, not a current-economics premium. The market pays ~2x the peer EV/sales for lower current margins, betting that MPWR’s smaller base, capital-light model and content-expansion runway let it compound at peer-beating rates for far longer. A coherent bet — but a bet on the future, priced as largely de-risked, on a ~0.9% FCF yield that leaves no margin of safety. (INTERPRETATION.)

Reverse-DCF / embedded expectations. The starting point is the ~0.9% FCF yield (the lowest in the peer set by far) — rational only if FCF compounds very fast for a very long time. At a ~9.5% discount rate, justifying the ~$71B EV requires terminal FCF roughly 4.3–4.8x today’s $666M, i.e. ~$3.0B by ~2035 — implying ~14–15% revenue CAGR for a decade with operating-margin expansion to the mid-30s. On a 5-year cross-check, the price embeds roughly 20–22% revenue CAGR with margins holding/expanding and a still-premium ~40–45x forward exit multiple. In plain terms, the price says MPWR keeps doing roughly what it did the last decade (~25% CAGR) for most of another decade, with no margin give-back and no multiple normalization. Partly plausible (the historical record is real and the base is still small) — but the AI-power leg that the premium is “priced for” declined in 2025, and management won’t claim share. The market is paying a 94th-percentile price for a re-acceleration that is, as of the latest data, unproven. (ASSUMPTION-driven; INTERPRETATION.)

Scenario analysis (5-year; illustrative, no target implied):

The bear requires no business failure — only that MPWR become an ordinary high-quality semi priced like one (growth normalizing to ~10–12% as the capital cycle digests and AI-power share is competed away), which compresses EV ~50%+. The base has the business growing into much of the current price for roughly flat total return — the “great business, you make little money from here” outcome. Even the bull pays only high-single-digit IRR, because you start at a ~0.9% FCF yield: you need both the growth and the multiple to hold just to earn a modest return. The asymmetry from today’s price is unfavorable. (ASSUMPTION/INTERPRETATION.)

Verdict (Valuation): The market is pricing MPWR as a perpetual ~20%+ compounder with margin expansion and a permanent premium multiple — extrapolating the past at the top of the historical valuation range — even as the segment that anchors the thesis stalled. The risk/reward from here is the textbook “priced-for-perfection compounder.”

11. Variant Perception

Consensus belief. MPWR is a best-in-class fabless analog/power compounder and a primary AI-power beneficiary — the nimble share-gainer riding GPU/ASIC power-delivery content, 50%+ ROIC, capital-light, secularly growing for years. The sell-side embodies this: ~4.29/5 rating and a ~$1,797 average target (third-party color only — explicitly not our target and not used as one). Short interest ~5.4% of float is a mild skeptic signal, not a crowded short — consensus is firmly bullish. (FACT.)

Strongest bull case. A genuinely differentiated fabless model (proprietary BCD process on merchant foundries, fast cycle-time, module integration) that has taken share for a decade and has a long content-expansion runway off a still-small ~$2.8B base: every new AI rack, electrified vehicle and multi-rail server raises MPWR silicon content. Capital-light economics protect margins (>100% incremental returns, ~6%-of-sales capex), so growth converts to FCF without fab-depreciation drag. If AI power re-accelerates, MPWR compounds revenue ~20%+ for years and the premium is earned.

Strongest bear case. You pay the 94th percentile of MPWR’s own decade and a ~60% EV/sales premium to ADI/TXN — for a business with lower current margins than two of three peers, on a ~0.9% FCF yield — where the AI-power segment the premium is built on actually shrank −2% in 2025 and management won’t claim share. The 2025 growth came from the more-cyclical non-AI book. Infineon, TI and ADI are all attacking data-center power and tout doubling; the moat is thinnest exactly where the valuation lives. Growth deceleration or multiple normalization (usually arriving together) takes the stock down 40–55% with the business fully intact.

The 3–5 assumptions that matter most:

1. AI/data-center power re-accelerates (the −2% 2025 stall reverses) — the load-bearing assumption.
2. MPWR holds/gains share against Infineon/TI/ADI rather than losing the contest.
3. Revenue CAGR stays ~15–20%+ for 5+ years off the larger base.
4. Operating margin holds ~30% and ideally expands — no mix-driven give-back as competition bites.
5. The premium multiple persists — the market keeps paying a top-decile own-history valuation.

What falsifies each side. Falsifies the bull: two+ more quarters of flat/declining Enterprise Data while rivals report data-center-power gains; or gross margin slipping below ~53%; or revenue CAGR slipping to low-teens. Falsifies the bear: Enterprise Data re-inflecting to double-digit growth with management quantifying AI-power socket share gains, and total revenue sustaining ~20%+ with margins holding ~30%+ — re-validating the secular step-up and earning the premium.

12. Fact vs. Interpretation Table

13. Open Questions

1. What is MPWR’s actual, current content/socket share at NVIDIA and the major hyperscaler AI platforms — and is it rising or falling versus Infineon? Management refuses to disclose; must be triangulated from teardowns/channel checks. The single most thesis-relevant unknown.
2. Why did Enterprise Data decline in 2025 — share loss, customer roadmap timing, or hyperscaler digestion? The distinction determines whether the AI leg re-accelerates.
3. When will the tax-accounting material weakness be certified remediated, and is there any SEC inquiry? The 10-K flags “increased possibility of legal proceedings and regulatory inquiries” but discloses none.
4. Which foreign jurisdiction granted the 10-year tax incentive, and how durable is the ~$1.1B DTA under OECD Pillar Two? Unnamed in the 10-K; Singapore is the most probable.
5. Is the 2025–26 buyback pause deliberate price discipline or simply a dividend preference? Filings don’t say; matters for reading capital-allocation intent.
6. Will gross margin hold ~55% as data-center-power competition intensifies, or is 55% a peak that pricing pressure erodes?
7. Does the capex step-up to ~6% of sales mark a structural shift away from the asset-light model?

14. What Must Be True (Bull and Bear, with Falsification Tests)

For the BULL (today’s ~$1,473 / ~$71B EV is justified or low): MPWR must sustain a ~18–25% revenue CAGR for 5–10 years with the AI/data-center-power leg re-accelerating and share held/gained, operating margin holding/expanding to ~32–34%, FCF compounding from $666M toward ~$2.5–3.0B — and the market must keep paying a premium exit multiple.

Falsification test: If, over the next four quarters, Enterprise Data fails to return to double-digit YoY growth, OR consolidated revenue growth decelerates below ~15%, OR gross margin slips below ~53%, the ~15–22% decade-long CAGR embedded in the price is broken and the ~0.9% FCF yield is unsupportable.

For the BEAR (today’s price embeds unrealistic expectations; ~40–55% downside on normalization): Growth must decelerate toward ~10–12% as the base scales and the capital cycle digests, AND/OR data-center-power share must be competed away by Infineon/TI/ADI, AND the multiple must compress from ~83x EV/EBITDA / ~49x forward toward peer-normal ~16–32x — without the business failing.

Falsification test: If MPWR delivers two consecutive quarters of re-accelerating Enterprise Data with management explicitly quantifying AI-power share gains, while total revenue holds ~20%+ and operating margin holds/expands ≥30%, the “decelerating compounder priced as perpetual” thesis is broken and the premium is being earned.

15. Source Appendix

See Appendix B for the full annotated source list. Primary sources: MPWR FY2025 Form 10-K (filed 2026-02-27, CIK 0001280452); Q1-2026 Form 10-Q (filed 2026-05-04); DEF 14A proxies (2022–2026); Form 4 insider filings (Feb–Jun 2026); earnings-call and Analyst Day transcripts (2022–2026); SEC EDGAR XBRL financial concepts; public market data and peer multiples for ADI, TXN, NXPI (2026-06-11). Management commentary is treated as hypothesis and validated against filings and financials throughout.

This analysis carries no investment recommendation and no price target. The only position expressed in this document is the clearly-labeled Claude's Take block at the top, which is the author’s own independent, subjective opinion and general information only — not investment advice.

APPENDIX A — Standard Diligence Questionnaire

Report date 2026-06-11. Supplemental to the research memo. Fact/Interpretation/Assumption labels applied where material.

General

What thoughtful questions have other investors asked about this company? The serious debate centers on five questions: (1) Is MPWR’s AI/data-center-power position durable, or being competed away by Infineon/TI/ADI? (the −2% Enterprise Data print in 2025 made this acute). (2) Can it sustain ~20%+ growth off a now-$2.8B base, or does the law of large numbers bite? (3) Is ~55% gross margin a floor or a peak? (4) How should one value a sub-1% FCF-yield, 94th-percentile-own-history stock? (5) Post-restatement, how much should the material weakness in tax controls and one-directional insider selling discount the multiple? The bull-bear fault line is almost entirely about price and AI-leg durability, not business quality.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? (Interpretation) Closer to a cyclical high than a low. Operating margin (30% in Q1-26) is at/near record, gross margin is at its stable-band ceiling, and revenue is +26% — but the AI-power segment specifically stalled, suggesting a mixed cyclical picture rather than a clean peak. The 2023 flat year shows the cycle is real.

Driven by external environment or internal actions? Both: the secular AI/data-center build-out and automotive electrification are external tailwinds; the proprietary process, module strategy and 14-year share-gain record are internal. The 2025 result was notably internally-driven (non-AI diversification carried it).

How stable are revenues? (Fact/Interpretation) Long product lifecycles (10+ years), ~30,000 customers and fragmented sockets give underlying stickiness, but the business is cyclical (flat 2023) and now AI-exposed (lumpy). Top-3 distributors ~54% of revenue adds channel concentration.

Outlook for products/services; how big will this market be? (Fact — management) Broad analog TAM ~$93–94B and growing; data-center power, 48V/vertical/800V rack power, automotive 48V/zonal, and silicon carbide are the expansion vectors. Market is growing, global (92% Asia-shipped), secular.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? (Interpretation) The broad analog industry is stable/consolidated; the AI-power sub-segment is getting more competitive (capital flooding in — Marathon capital-cycle signal).

How profitable (ROIC, ROE)? (Fact) ROE ~19% (understated by net cash + ~$1.1B DTA); ROIC 50%+ on true operating capital — extraordinary, the core quality fact.

How profitable is the industry; barriers to entry? (Fact) High-quality industry (ADI 64%/38%, TXN 57%/38%, NXP 55%/25–28% margins). Barriers: analog design craft, process IP, design-in qualification, long lifecycles. MPWR’s ~55%/~30% sits at the high-volume-power end.

Can the business be easily understood? (Interpretation) Yes at the business-model level (sells power chips/modules, fabless, organic). The tax accounting is complex (hence the restatement).

Undermined by foreign low-cost labor? (Interpretation) No — value is in analog design IP and process, not labor; though it depends on Asian foundry/assembly (TSMC-class) — a supply-chain/geopolitical exposure, not a labor-cost one.

Do brands matter? Nature of competition? Switching costs? (Interpretation) Brand matters modestly (engineering reputation); competition is on performance/integration/design-win; switching costs are real and graded — high in auto/industrial (multi-year qual), low in merchant data-center power.

Financial Condition & Balance Sheet

Assets not fully recognized on balance sheet? (Interpretation) The internally-developed BCD process IP and design library are expensed, not capitalized — a hidden asset. Conversely the ~$1.1B deferred-tax asset is recognized but is judgment-laden (restated).

Off-balance-sheet liabilities? (Fact) None material; minimal debt (~$20M), standard operating leases. No pension/large contingencies disclosed.

How conservative is the accounting? (Fact/Interpretation) Operating accounting is clean (revenue, margins, cash flow reconcile); the tax line is the exception — a material weakness, not-effective at both 2024 and 2025 year-ends, forced a ~$195M 2024 restatement. SBC is fully expensed (conservative). Net: conservative on operations, deficient on tax.

How CapEx-hungry? (Fact) Historically light (~3% of sales), stepping to ~6% (2024–25) on owned-capacity build — still far below an IDM/foundry, but rising.

Capital Allocation & Management

How much FCF; how is it used; philosophy? (Fact) ~$666M FCF (2025). Uses: ~46% to a fast-growing dividend, opportunistic buyback (big 2024, dormant since), heavy R&D reinvestment; balance accretes to net cash. Philosophy: build-don’t-buy, reinvest at 50%+ ROIC, return a growing dividend, hoard the rest.

Significant acquisitions? (Fact) Essentially none — only the immaterial Jan-2024 Axign tuck-in ($25.9M goodwill). A positive (no integration/impairment risk).

Buying back shares? (Fact) Lumpily — $636M in 2024, ~$0 otherwise; $500M authorization ~99% unused. Share count flat (SBC offset, not shrinking).

Issuing large amounts of new shares to insiders? (Fact) SBC ~$227M/yr (~8% of revenue) — heavy, but bought back to keep the count flat.

Compensation policy / motivations? (Fact) CEO ~$19.9M, ~95% variable; PRSUs on revenue + new 3-yr relative-TSR (improvement); cash bonus paid 400% of target in 2025. Aggressive magnitude; alignment improved by the TSR overlay. Single-class stock (positive); classified board being unwound by 2030.

Valuation & Market Data

ADR, MLP, or K-1 issuer? (Fact) No — ordinary US-domestic C-corp common stock (NASDAQ), 1099 dividends.

Dividend policy? (Fact) 14-year raise streak; $6.24 (2025) → $8.00 annualized (2026, +28%); ~0.5% yield; ~46% payout, well covered.

How profitable? (Fact) ~55% gross, ~30% operating, ~22% net margin (clean 2025); 50%+ ROIC.

Net income diverging from cash from operations? (Fact) Only in 2024, and entirely due to the non-cash deferred-tax benefit (OCF correctly strips it). On the clean 2025 base, FCF/NI ~1.07x — no divergence.

Risks & Downside

What factors would cause the stock to decline? (Interpretation) Multiple de-rating from the 94th percentile; AI-power share loss/Enterprise Data staying flat; growth deceleration; cyclical/inventory air-pocket; gross-margin pressure; escalation of the accounting/regulatory issue. A de-rating needs no business failure.

Risk of catastrophic loss? (Interpretation) Low — net cash, no debt, profitable, diversified, no goodwill to impair. The risk is valuation, not solvency.

Chance of total loss? (Interpretation) Negligible absent fraud or geopolitical catastrophe (Taiwan/China foundry shock).

Recent News & Events

Has the business environment changed recently? (Fact) Yes: the AI inflection (2024) and subsequent Enterprise Data stall (2025); intensifying data-center-power competition; the Feb-2026 restatement + material-weakness disclosure; dividend +28% and a new (unused) buyback authorization; capacity expansion ($4B→$6B goal). (Timeline built from 8-Ks, the 10-K, and earnings-call transcripts.)

Significant acquisitions? Only the immaterial Axign (2024).

Change in accounting policies? The income-tax restatement and ongoing remediation of the material weakness.

Recent changes — markets, facilities, management? New product vectors (DDR5 interface, SiC rack power, robotics adjacencies); owned-capacity build in a foreign tax-incentive jurisdiction; combined Chairman/President/CEO under founder Hsing; classified board being phased out by 2030.

APPENDIX B — Source Appendix

Report date 2026-06-11. Primary sources prioritized. Management commentary treated as hypothesis and validated against filings/financials.

Primary — SEC filings (EDGAR, CIK 0001280452)

1. Form 10-K, FY2025 (filed 2026-02-27, SEC EDGAR CIK 0001280452). Business (Item 1), Risk Factors, MD&A (revenue by end market, margins, capex), Item 9A (material weakness in income-tax ICFR), Note 2 (2024 restatement), Note 17 (income taxes / 10-year foreign incentive / ~$1.1B deferred-tax benefit), balance sheet, cash flow, Issuer Purchases, dividend policy.
2. Form 10-Q, Q1-2026 (filed 2026-05-04). Q1-26 revenue +26%, op margin 30%, segment detail, capex.
3. DEF 14A proxies (2022, 2023, 2024, 2025, 2026). 2026 (filed 2026-04-30): Summary Compensation Table (CEO ~$19.9M), PRSU metrics (revenue + 3-yr relative TSR, 500% max), beneficial ownership (~3.7% insiders; Hsing ~2.1%), classified-board phase-out, $10M related-party VC-LP investment.
4. Form 4 insider filings (Feb–Jun 2026; full corpus 436 filings, SEC EDGAR). Zero open-market purchases; CEO Hsing sold >150k shares / >$200M at $1,100–1,520 (mix of 10b5-1 and discretionary); broad officer/director selling.
5. 8-K filings (2021–2026). Earnings releases, dividend declarations, buyback authorization (Feb-2025 $500M), restatement disclosure (Feb-2026).
6. SEC EDGAR XBRL financial data — revenue, net income, OCF, capex, SBC, buybacks, dividends, shares outstanding, equity, goodwill: multi-year series 2018–2025 reconciled to the 10-Ks.

Primary — Transcripts (company IR / public transcript services)

7. Analyst/Investor Day, 2025-03-20 (~103k chars) — “Jinghai’s Law,” data-center power roadmap (48V → vertical → 800V rack), TAM framing, capacity goals.
8. Q1-2026 earnings call, 2026-04-30 — chip→module→system strategy, DDR5 interface, comms +33% QoQ, AI-share refusal.
9. Q4-2025 earnings call, 2026-02-06 — “non-Enterprise Data grew 40%+,” competitor “doubling” exchange, dividend raise.
10. Q2/Q3-2025 and FY2022–2024 earnings calls — multi-year growth, margin, end-market commentary.

Secondary — Market data & comparables

11. Public market data (2026-06-11) — price ~$1,473, market cap ~$72.4B, EV ~$71.0B, net cash ~$1.35B, ~49.1M shares; multiples (trailing/forward P/E, EV/S, EV/EBITDA, P/S, P/B, FCF yield); peer multiples for ADI, TXN, NXPI. Reconciled to filings.
12. Valuation context — sector/GICS classification, IPO 2004-11-19, ~4,500 employees, ownership, short interest ~5.4% of float, third-party analyst rating ~4.29/5 and average target ~$1,797 (color only, not used as a target); own-history valuation percentiles (P/E 90.6th, P/B 93.5th, P/S 98.5th, composite 94.2nd). Third-party aggregated — reconciled to EDGAR.

Frameworks

14. Greenwald & Kahn, “Competition Demystified” and Chancellor / Marathon, “Capital Returns” — applied to moat-type classification, share-stability/ROIC tests, and the AI-power capital-cycle read.