Monolithic Power Systems MPWR

Reference table

1. Eight quarters of growth, and what produced it

Quarterly results, from inc_q and earn_cal, $M except per share:

Three observations.

Revenue has increased in every one of the last eight quarters and is 93.3% higher than it was two years ago. That is an unbroken sequential run, which is rare in semiconductors and does not happen by accident.

The gross margin did not move. 55.1% to 55.2% across a quarter in which revenue grew 47.6%. A semiconductor company that grows revenue by half without giving away a point of gross margin is being pulled by demand rather than pushing on price, and the entire 84.4% increase in operating income is therefore scale over a fixed research and selling base.

The beat pattern is accelerating. +2.2%, +1.9%, +1.1%, +4.1%, +10.7%. Consensus is now visibly behind the ramp, which cuts both ways — it means the FY2026 estimate is likely conservative, and it means the market is repricing this name on estimate revisions rather than on a stable multiple.

One caution on the table: the December 2024 quarter shows net income of $1,449.4 million on revenue of $621.7 million. That is a one-off tax event — the FY2024 annual row carries an income tax BENEFIT of $1,213.8 million, producing FY2024 EPS of $36.76 against FY2025's $12.82 on 26.4% higher revenue. Every trailing earnings comparison spanning FY2024 is distorted by it and none is used in this dive.

The annual series, for context ($M):

Research and development spending has risen every year, from $137.6 million to $382.3 million — 178% over five years — while gross margin has drifted down from 58.4% to 55.2%. The company is investing hard and accepting three points of gross margin to do it, which is the correct trade in a design-cycle business and is stated here because a reader watching only the margin line would misread it.

2. The concentration — the single most important table in this file

Revenue by geography, from seg_geo. The block reconciles to consolidated revenue EXACTLY and is used ($M):

The sum matches consolidated revenue to the dollar in every year shown, and the series is consistent back to fiscal 2016. This is a clean block and the finding it delivers is the largest single risk in this dive.

Three quarters of Monolithic Power's revenue ships to greater China. Three and a half per cent ships to the United States.

Three qualifications, all of which matter and none of which changes the conclusion.

These are ship-to destinations, not necessarily end markets. A power-management chip shipped to a contract manufacturer in Shenzhen may end up in a server installed in Virginia. The revenue is nonetheless booked against a customer in that jurisdiction, and it is that customer relationship — and that jurisdiction's export and import regime — which determines whether the sale happens.

The concentration has been stable, not sudden. China plus Taiwan was 68.2% of revenue in FY2023 and 79.6% in FY2024 before settling at 75.0% in FY2025. This is the structure of the business and not a recent drift.

And it has been growing, not shrinking. China revenue rose 31.1% in FY2025 to $1,544.3 million. Korea (+50.5%) and South East Asia (+88.1%) grew faster in percentage terms but from bases one-sixth the size.

What this means for a shareholder, stated plainly. A tariff, an export control, an entity-list action or a cross-strait disruption does not compress this company's margin — it removes its customers. There is no operational mitigation available on a two-year horizon for a business whose design wins sit inside Chinese and Taiwanese systems. The balance sheet's $1.08 billion of net cash and the 55.2% gross margin are excellent and neither of them addresses this. It is the reason the risk score is 7 rather than 4, and the reason the verdict is Watch rather than Buy.

Product disclosure is thinner. seg_prod has not been updated since fiscal 2023 and reports only two lines — DC-to-DC Products $1,718.6M and Lighting Control Products $102.4M against FY2023 revenue of $1,821.1M, which does reconcile. There is no product-level split for FY2024 or FY2025 anywhere in the payload, and none in the extractable text of the filings in this archive.

3. Balance sheet and cash

From bal_a at 2025-12-31 and the 10-Q at 2026-03-31, $M:

Monolithic Power has no borrowings. Its entire $24.1 million of "total debt" is $20.0 million of operating-lease obligations and $4.1 million of other short-term items. Net cash of $1,075.2 million on a $65.59 billion market capitalisation is 1.6% — small in relative terms and absolute in kind: there is no refinancing risk, no covenant, no interest expense (interestExpense reads $0 in every year of inc_a) and no leverage of any sort.

Cash flow, from cf_a, $M:

Free cash flow of $666.2 million in FY2025 is a 1.02% yield on the current market capitalisation — low, because the market is paying for the growth rather than for the cash. The capital-expenditure field is CLEAN: freeCashFlowOperatingCashFlowRatioTTM is below 1.0 and the annual arithmetic ties exactly in every year. capexToDepreciationTTM is elevated at roughly 3.3x, which for a company more than doubling revenue and building test and packaging capacity is expected rather than suspicious.

Two items deserve naming. Stock-based compensation of $227.5 million in FY2025 is 8.2% of revenue and 34.1% of free cash flow — high, normal for a semiconductor designer, and a real economic cost that flows into the share count rather than the income statement. And the buyback is erratic: $636.2 million in FY2024 and $7.7 million in FY2025, so it is opportunistic rather than programmatic and should not be modelled as a recurring return.

The dividend is small and rising. The 10-Q records $2.00 per share declared in the March 2026 quarter, $99.3 million, against a dividendPerShareTTM of $7.12 — a 0.53% yield. At $2.00 a quarter the forward run rate is $8.00, so the trailing figure understates it.

4. Valuation — priced in or room?

At $1,335.03 (market cap $65.59B, 49,129,000 shares):

Estimate coverage is adequate on the near years and collapses beyond. FY2026 and FY2027 each rest on 9 analysts for EPS and 11 for revenue, with an FY2027 EPS range of $32.390 to $39.645 — a 22.4% spread, which is wide and appropriate for a business growing 26-47%. FY2028 rests on 5 analysts. FY2029 and FY2030 rest on ONE analyst each for EPS and are excluded from every conclusion in this dive.

A useful sanity check the file supports. First-half 2026 revenue was $1,784.8 million ($804.2M + $980.6M); the FY2026 consensus of $4,115.3 million implies a second half of $2,330.5 million, and the Q3 consensus of $1,140.6 million implies a fourth quarter near $1,190 million against the June quarter's $980.6 million. The consensus requires two more sequential increases of roughly 16% and 4%. Given eight consecutive sequential increases, that is demanding rather than heroic.

est.ebitdaAvg and est.ebitAvg are REJECTED for a fixed-ratio fabrication signature. From FY2023 through FY2030, ebitdaAvg is exactly 28.626% of revenueAvg and ebitAvg exactly 26.598%, in every year. The company's realised EBITDA margin has ranged from 22.0% (FY2020) to 31.3% (FY2022) and was 29.3% in FY2025, so the fabricated ratios happen to be near the realised level and are still fabricated. Note also that netIncomeAvg EXCEEDS ebitAvg in every forward year — FY2027 shows net income of $1,713.7 million against EBIT of $1,378.9 million, and FY2030 $2,653.6 million against $2,136.5 million — which is arithmetically impossible for a taxpaying company. Both documented signatures fire. All forward valuation uses epsAvg.

Enterprise value — CLEAN, and no omission is possible. enterpriseValueTTM is below the market capitalisation because the company holds net cash, minorityInterest is 0 and preferredStock is 0. Neither of this programme's standard EV defects can occur here.

Peer context — the set is close to useless. The vendor peers are ASE Technology, Electronic Arts, Garmin, NXP Semiconductors, Seagate, Atlassian, Take-Two Interactive, Ubiquiti, Western Digital and Block. Two are video-game publishers, one is a payments company, one is enterprise software and two are hard-disk makers. Only NXP Semiconductors and ASE Technology are semiconductor businesses, and neither is an analog power-management designer. Texas Instruments, Analog Devices and Infineon — the three companies against which this business is actually judged — are all absent. No peer multiple comparison is drawn.

4a. What today's price assumes (the inversion)

At $1,335.03 — 49.2x FY2026 consensus, 38.3x FY2027, 17.8x book — the price embeds:

• Revenue reaches $4,115 million in FY2026 and $5,184 million in FY2027. (Consensus; 11 analysts each.) The first half delivered $1,784.8 million, so FY2026 needs $2,330.5 million in the second half — two more sequential increases after eight consecutive ones.
• Gross margin holds near 55%. (Our derivation; the margin is the company's.) It has been between 55.1% and 55.4% for nine straight quarters through a doubling of revenue. Every point of gross margin is roughly $41 million of annual operating income at the FY2026 consensus revenue.
• The greater-China concentration does not become an event. (Our derivation from the filed seg_geo.) This is the most fragile assumption in the price, and it is fragile in a way the others are not: it is binary rather than continuous, and no operational response is available. 75.0% of revenue.
• The market keeps paying 35-45x forward earnings. (Our number.) At 30x FY2027E the stock is $1,044; at 45x it is $1,567. A fifteen-turn band on a business growing revenue 26-47% is the honest width, and it is why the fair-value range below spans 850 to 2,000.
• A permanent chief financial officer is appointed without incident. (Our number.) Robert Dean has signed two consecutive earnings 8-Ks as interim.

4b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: EPS growth (+28.2%, FY2026E $27.153 to FY2027E $34.816) + multiple compression (assumed, 49.2x on the forward year to about 43.9x on the then-forward year, −11%) + dividend yield (+0.53%) ≈ +15%.

The base case assumes earnings grow 28% and the multiple compresses 11%, and the two nearly cancel — leaving a return of roughly 15% that is entirely earnings and slightly negative on re-rating. That is the correct shape for a company at this stage: 49x forward earnings on 47% revenue growth is a statement about the growth rate, and growth rates decelerate — consensus itself has revenue growth going 47.5%, 26.0%, 13.5%. The practical consequence is that the thesis requires the earnings and nothing else, which is the return shape this programme prefers. What it does not do is protect the entry: if the earnings disappoint, the multiple compresses anyway and there is a very long way down, which is what the $850 bear case describes.

If the multiple compressed to 30x FY2027E the price would be $1,044 (−21.8%). At 45x, $1,567 (+17.4%).

4c. Variant perception (where we differ, what would surprise)

• We are materially BELOW the street. Our base of $1,530 is 15.0% below the consensus target of $1,800.71 and the median of $1,800. The street's LOWEST published target, $1,500, is 12.4% above spot, and 22 of 25 analysts rate the shares Buy with none at sell. We do not disagree with the direction; we think 38.3x forward earnings on a company with 75% of its revenue in one geopolitical jurisdiction should carry a discount that the street is not applying.
• We think the geographic concentration is the whole variant perception and it is under-priced. seg_geo is one of the cleaner blocks in this batch, it reconciles exactly, and it says three quarters of the revenue ships to China and Taiwan. Watchable number: greater-China revenue share in the fiscal 2026 10-K, against 75.0%. A fall toward 65% would be a genuine de-risking; a rise above 80% would make the multiple harder to defend.
• We think the gross-margin stability is under-appreciated in the other direction. 55.1% to 55.2% across a 47.6% revenue increase. Watchable number: gross margin in the 2026-10-29 print. Holding 55% at $1.14 billion of quarterly revenue would be strong evidence that pricing is structural.
• We note the absent knowledge base and think it is a genuine gap. Zero claims out of 52,021, on the company that supplies power-delivery silicon for the AI build — while three independent claims elsewhere in this same batch identify optical interconnect as a bottleneck. The store has a view on optics and none on power.
• Positive surprise that would force a re-rate: a Q3 print above $1.2 billion of revenue with gross margin above 55.5%; a permanent chief financial officer appointed; disclosure reducing greater-China share below 70%; or the FY2027 consensus revised above $38.
• Negative surprise that would break the thesis: any export-control, tariff or entity-list action touching China or Taiwan revenue; gross margin below 54%; Q3 revenue below $1.05 billion; a further quarter with an interim chief financial officer; or a break below the 200-day average of $1,214.35.

Synthos fair values

All three anchors are multiples of the FY2027 consensus EPS distribution (mean $34.816, low $32.390, high $39.645, 9 analysts). The 22.4% analyst spread and the geopolitical binary together justify an unusually wide range, and no greater precision is claimed than those two facts permit.

• Bear ~$850 — 26.2x the FY2027 consensus LOW of $32.390, and 31.3x FY2026E. Cross-check: 6.6% above the 52-week low of $797.51; 30.0% below the 200-day average. The scenario: a trade or cross-strait action removes a material share of greater-China revenue, growth decelerates hard, gross margin gives way, and a 1.71-beta stock at 49x forward earnings de-rates to the low twenties on reduced estimates. −36.3%.
• Base ~$1,530 — 43.9x the FY2027 consensus MEAN of $34.816, and 38.8x FY2028E. Cross-check: 56.3x FY2026E; 9.4% below the 52-week high of $1,689.89; 15.0% below the street consensus of $1,800.71; 20.4x book. Sensitivity, stated openly: 38.3x FY2027E gives $1,335 — spot exactly — and 45x gives $1,567. Section 4b notes that the base assumes earnings grow 28% while the multiple compresses 11%, and that the two nearly cancel. The scenario: the consensus revenue path is met, gross margin holds near 55%, the concentration does not become an event, and the multiple compresses in line with the decelerating growth rate. +14.6%.
• Bull ~$2,000 — 50.4x the FY2027 consensus HIGH of $39.645, and 50.8x FY2028E. Cross-check: 18.4% above the 52-week high; 4.8% below the highest published street target of $2,100. The scenario: AI power-density growth outruns the consensus, revenue exceeds $5.5 billion in FY2027, gross margin expands back toward 57%, geographic diversification improves, and the market pays an analog-scarcity multiple. +49.8%.

Base is 14.6% above spot; asymmetry roughly 1.37:1 (36.3% down, 49.8% up), plus a 0.53% dividend. A 14.6% base with a payoff ratio below 1.5:1 on a 1.71-beta security whose largest risk is binary and political does not clear a Buy bar, and the upgrade conditions below name the price and the disclosure that would change it.

5. Knowledge base — thirteen raw hits, ZERO claims, and a conspicuous gap

Raw hits: 13. Entity matches after a case-sensitive re-run: ZERO. Text matches: ZERO. Name-level claims on Monolithic Power: ZERO. Discarded: 13.

The primary sweep ran the entity terms MPWR, Monolithic Power and Monolithic Power Systems, plus free text on power management IC, VRM and voltage regulat, across all 52,021 distilled claims. It returned 13 hits across 9 channels — real_vision (3), all_in (2), cognitive_revolution (2) and one each from jensen_huang, ml_street_talk, arthur_hayes, nassim_taleb and anthony_pompliano_show. Not one names this company, and the case-sensitive entity re-run returns nothing at all.

Breadth 0, claim count 0, net conviction empty. This is the sixteenth void this programme has found and reported honestly.

And it deserves one further paragraph rather than a shrug, because of what sits alongside it in this same batch. The free-text terms were chosen deliberately: voltage-regulator modules and power-management integrated circuits are the components that sit between a data-centre power supply and a processor, and power delivery at extreme current density is one of the two or three genuine engineering constraints on accelerated computing. Elsewhere in this batch, three independent knowledge-base claims — from jensen_huang, andreas_steno and jordi_visser, two with named speakers — identify optical interconnect as a bottleneck and name the two companies that supply it. One of those claims uses the phrase "a choke point as electricity becomes the AI bottleneck." The store therefore contains a well-sourced, current, multi-voice view that electricity is becoming the constraint — and no claim whatsoever about the company that designs the silicon which delivers it. That is a genuine and specific gap in the knowledge base's coverage of a theme it otherwise covers well, and it is reported as one.

No concentration test, attribution note or speaker analysis is possible on zero claims and none is manufactured.

6. Data integrity — what we rejected and why

Four findings, and this is one of the cleaner payloads in this batch.

1. est.ebitdaAvg and est.ebitAvg trip TWO documented signatures simultaneously — REJECTED. (a) Fixed-ratio fabrication: from FY2023 through FY2030, ebitdaAvg is exactly 28.626% of revenueAvg and ebitAvg exactly 26.598%, in every year. (b) netIncomeAvg EXCEEDS ebitAvg in every forward year — FY2027 net income of $1,713.7M against EBIT of $1,378.9M, FY2028 $1,987.7M against $1,565.1M, FY2030 $2,653.6M against $2,136.5M — arithmetically impossible for a taxpaying company, and the signature the data contract measures at 8.9% of files. All forward valuation uses epsAvg.

2. The FY2029 and FY2030 estimate rows rest on ONE analyst each for EPS — EXCLUDED. numAnalystsEps reads 1 for both, against 7 analysts on revenue in each year. A single-contributor EPS figure is not a consensus and printing a 28x FY2029 multiple from it would be a fabrication. Both rows are excluded from every conclusion.

3. seg_prod has not been updated since fiscal 2023 — REJECTED for current use. The most recent row reports DC-to-DC Products $1,718.6M and Lighting Control Products $102.4M, summing to $1,821.0M against FY2023 revenue of $1,821.1M — so the block is accurate for the year it labels and two years stale. There is no product-level revenue split for FY2024 or FY2025 anywhere in the payload, and none in the extractable text of the filings.

4. FY2024 earnings figures are distorted by a $1,213.8 million income-tax BENEFIT and are not used. inc_a reports FY2024 net income of $1,786.7 million on operating income of $539.4 million and pre-tax income of $572.9 million, with incomeTaxExpense of MINUS $1,213.8 million. The December 2024 quarter carries $1,449.4 million of that. FY2024 EPS of $36.76 against FY2025's $12.82 on 26.4% higher revenue is the visible consequence. Every trailing multiple spanning FY2024 is meaningless and none is used; all forward valuation is on epsAvg.

Verified CLEAN — recorded because clean checks are findings, and this name has an unusual number:

• Share count — EXACT. The 10-Q's statement of stockholders' equity shows 49,129 thousand shares at 2026-03-31 against 49,130,000 implied by market capitalisation ÷ price. An exact match.
• seg_geo is CLEAN and reconciles EXACTLY in every year from fiscal 2016 to fiscal 2025 — FY2025's eight regional lines sum to $2,790,479 thousand against consolidated revenue of $2,790,459 thousand, a 0.0007% difference. It correctly identifies China as by far the largest market. This is the block that delivers the largest finding in this dive, and its reliability is what makes that finding usable.
• capitalExpenditure is CLEAN. FY2025 capex of −$172.0M against operating cash flow of $838.2M gives free cash flow of $666.2M, and the arithmetic ties exactly in all four annual rows. capexToDepreciationTTM is elevated at roughly 3.3x, which is expected for a company more than doubling revenue and building capacity rather than a corruption signature.
• Enterprise value — no omission is structurally possible. The company holds net cash, minorityInterest is 0 and preferredStock is 0, so neither the noncontrolling-interest nor the preferred omission this programme finds on most names can occur.
• totalDebt is NOT inflated in any misleading way — the $24.1M figure is disclosed as comprising $20.0M of operating-lease obligations and $4.1M of other short-term items, and interestExpense reads $0 in every year of inc_a, consistent with a company that has no borrowings.
• tech.max_dd_from_peak of −20.999% coincides exactly with pct_from_hi. Per the data contract this is correct, not a defect — the six-year peak falls inside the last twelve months. Described as the current distance from the high, never as a maximum drawdown.
• quote versus tech. tech.last of $1,335.03 equals quote.price exactly. quote.yearHigh/yearLow ($1,714.09/$774.14) against tech.hi52/lo52 ($1,689.89/$797.51) — 1.4% and 3.0%. tech is used throughout.

Vendor composite rating — B- / 3 overall, with priceToEarningsScore 1 and priceToBookScore 1 against returnOnAssetsScore 5. The valuation sub-scores are correct — this is an expensive stock — and the return-on-assets score of 5 out of 5 on a debt-free balance sheet is likewise correct. The debtToEquityScore of 1 for a company with $1.08 billion of net cash and no borrowings is not, and is disregarded.

Non-equity tripwire — checked and passed. MPWR is common stock, NASDAQ-listed. Price of $1,335.03 is not par-like; beta is 1.71, the second highest in this batch; volume was 1.21M shares (~$1.62 billion of turnover, the second heaviest); the 52-week band of $797.51 to $1,689.89 is a 112% range; the dividend is a regular quarterly common dividend of $2.00. No preferred stock, no noncontrolling interests, no borrowings. This is common equity.

7. Technicals, governance and insiders

• Price $1,335.03. −21.0% from the 52-week high of $1,689.89; +67.4% above the low of $797.51. Position within the annual range: 60th percentile.
• Split moving averages. 5.6% BELOW a 50-day average of $1,437.06; +9.9% above a 200-day average of $1,214.35. The short-term trend has rolled over while the longer one holds.
• RSI 48.4 — neutral. MACD −20.43 — the most negative reading in this batch in absolute terms, reflecting the three-month decline.
• Relative performance: 3-month −15.1% against SPY +7.6% — a 23-point deficit; 6-month +13.8% against SPY +11.1%; 12-month +69.9% against SPY +24.3% and QQQ +30.8%.
• Today's session is worth its own sentence. MPWR opened at $1,402.10, traded as high as $1,425.99 and as low as $1,333.58, and closed at $1,335.03 — $1.45 above the day's low, having given up $92.41 from the high, a 6.5% intraday range, on $1.62 billion of turnover. A stock that closes at the bottom of a wide range five days after a 10.7% earnings beat is being sold by someone, and this dive does not know who or why.
• Sentiment: 22 buy, 3 hold, 0 sell across 25 analysts — the most one-sided distribution in this batch; consensus target $1,800.71 (+34.9%), median $1,800, high $2,100 (+57.3%), low $1,500 — 12.4% ABOVE spot. No covering analyst has a target below the price and none rates it sell. On a 1.71-beta security whose largest risk is geopolitical and binary, that unanimity is treated as a caution.

Governance, from the 8-Ks — two items, both minor and both worth recording.

A $65.6 billion company is running with an interim chief financial officer. Robert Dean signed the earnings 8-K of 2026-04-30 and the earnings 8-K of 2026-07-30, in both cases as "Interim Chief Financial Officer." No permanent appointment, and no disclosure of a search, appears anywhere in this archive. Two consecutive quarters is not alarming and it is not nothing.

The annual meeting of 2026-06-11. Two Class I directors were elected to three-year terms until 2029, confirming a classified board. Victor K. Lee received 39,298,042 votes for and 4,440,848 withheld — 10.2% withheld — against Jeff Zhou's 2,602,941, or 6.0%. Ernst & Young was ratified with 72,479 against out of 45.4 million. Say-on-pay passed with 42,164,328 for and 1,411,084 against — 96.8%. Nothing here is a governance problem; the 10.2% withheld vote against one director is the only figure above routine and it is recorded rather than interpreted.

Insiders: the insider block for this name contains no transactions in the extracted payload. No purchase, no sale, no award and no withholding is available to this dive, so no insider inference of any kind is drawn.

8. Verdict, kill-criteria and flip conditions

Watch.

The case for the business is close to the strongest in this batch. Revenue +47.6% to $980.6 million in the June quarter, operating income +84.4%, net income +92.4%, a 10.7% earnings beat — the fifth consecutive beat and the largest. Eight consecutive quarters of sequential revenue growth, from $507.4 million to $980.6 million, a 93.3% increase. Gross margin of 55.2% against 55.1% a year earlier — the growth was not bought. Research and development spending up 178% over five years to $382.3 million. A balance sheet with $1,256.5 million of cash and short-term investments, $24.1 million of total debt of which $20.0 million is operating leases, net cash of $1,075.2 million, no borrowings and no interest expense in any year on file. A founder chief executive and 4,501 employees producing $2.79 billion of revenue. And consensus carrying revenue to $5,884 million by FY2028.

Why this is Watch, in three parts.

First, and dominating everything else: 75.0% of revenue ships to greater China. seg_geo reconciles exactly and reports FY2025 revenue of $1,544.3 million to China and $550.1 million to Taiwan, against $96.7 million — 3.5% — to the United States. This is not a margin exposure, a supply-chain exposure or a currency exposure. It is where the customers are. A tariff, an export control, an entity-list action or a cross-strait disruption removes revenue, and no operational response is available on a two-year horizon. The $1.08 billion of net cash does not address it.

Second, the arithmetic does not pay for that risk. Base $1,530, +14.6%, asymmetry 1.37:1, on a 1.71 beta. At 38.3x FY2027 consensus the price already assumes the concentration does not become an event, and a 14.6% base case is not compensation for a binary political variable.

Third, two smaller items. A $65.6 billion company has signed two consecutive earnings releases with an interim chief financial officer. And the stock closed today at $1,333.58 — the exact low of a $92.41 range — five days after a 10.7% beat.

And the honest counterweight, which is why this is Watch rather than Avoid. This is an exceptional business at a price that is high but not absurd for 26-47% revenue growth, and the entry has already improved: the stock is 21.0% below its high and has fallen 15.1% over three months while the index rose 7.6%. At a lower price the same concentration risk becomes a reasonable trade, and the upgrade conditions below say where.

Pre-registered KILL criteria — what would take this to Avoid:

• Any export control, tariff, entity-list action or cross-strait disruption materially affecting China or Taiwan revenue, which is 75.0% of the company.
• Gross margin below 54%, against 55.1-55.4% for nine consecutive quarters.
• Q3 2026 revenue below $1.05 billion against the $1,140.6 million consensus, which would break an eight-quarter sequential streak.
• The FY2027 consensus EPS falling below $31, against $34.816 with a low of $32.390.
• A third consecutive quarter with an interim chief financial officer, or the departure of founder chief executive Michael R. Hsing.
• A break below the 200-day average of $1,214.35 on company-specific news.

Pre-registered UPGRADE conditions — what would take this to Buy — Tactical:

• A price below approximately $1,100, which is 31.6x FY2027E and would put the base case above 39% with the same fundamentals. This is the cleanest trigger and it requires nothing of the company.
• Greater-China revenue share disclosed below 70% in the fiscal 2026 annual report, against 75.0% — the only disclosure that would materially reduce the single largest risk.
• A Q3 print above $1.2 billion of revenue with gross margin above 55.5%, demonstrating that pricing holds as volume scales.
• A permanent chief financial officer appointed.
• The FY2027 consensus revised above $38, at which spot is 35.1x.
• Any independent knowledge-base claim naming this company. At present there are zero, on a theme the store otherwise covers with three well-sourced claims.

Where MPWR fits in the Synthos Framework Portfolio. No position today, with a price trigger at approximately $1,100 and a disclosure trigger on the greater-China share. The technology sleeve would take this at a target weight of 1.5%, entered in halves. Sizing note, and it is the operative discipline: a risk that is binary and political cannot be managed by position sizing alone — it can only be paid for at entry, and at 38.3x forward earnings it is not being paid for. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $1,335.03, with the fair-value anchors, kill criteria and upgrade conditions all gradeable.

Single biggest risk: 75.0% of revenue ships to China and Taiwan. The seg_geo block on this name is one of the cleaner data structures in this batch — it reconciles to consolidated revenue to within seven ten-thousandths of a per cent in every year back to fiscal 2016 — and it reports FY2025 revenue of $1,544.3 million to China, $550.1 million to Taiwan and $96.7 million to the United States. China revenue grew 31.1% last year, so the concentration is deepening rather than easing. Everything else about this company is exemplary: eight consecutive quarters of sequential revenue growth, a 47.6% year-on-year increase achieved without giving up a single point of the 55.2% gross margin, $1.08 billion of net cash and no borrowings at all, 4,501 employees producing $2.79 billion of revenue, and a founder still running it. None of that provides any defence at all against a decision made in Beijing or Washington. At 38.3 times the FY2027 consensus, on a 1.71 beta, with a base case of 14.6% and a payoff ratio of 1.37 to 1, the price is not compensating for a risk that cannot be hedged, diversified or managed — only avoided or accepted. At a price near $1,100 the same risk becomes an acceptable trade, and that is the number to wait for.

Provenance & disclosures

• Traceability: ZERO name-level knowledge-base claims name Monolithic Power Systems out of 52,021 distilled claims (raw hits 13, entity matches after a case-sensitive re-run: 0, text matches 0, discarded 13; breadth 0, claim count 0, net conviction empty). The sweep ran the entity terms MPWR, Monolithic Power and Monolithic Power Systems, plus free text on power management IC, VRM and voltage regulat, and returned 13 hits across 9 channels — real_vision (3), all_in (2), cognitive_revolution (2) and one each from jensen_huang, ml_street_talk, arthur_hayes, nassim_taleb and anthony_pompliano_show — none naming this company. The gap is specific and worth recording: elsewhere in this same batch, three independent claims with two named speakers identify optical interconnect as an AI bottleneck, one of them using the phrase "a choke point as electricity becomes the AI bottleneck." The store therefore holds a well-sourced, current, multi-voice view that electricity is becoming the constraint on accelerated computing, and no claim whatsoever about the company that designs the silicon which delivers it. This is the sixteenth void found and reported honestly. No concentration test, attribution note or speaker analysis is possible on zero claims and none is manufactured.
• Data as-of: fundamentals — share count, stockholders' equity, dividends declared and the balance-sheet detail through 2026-03-31, filing-verified from the 10-Q filed 2026-05-04; the June 2026 quarter has no 10-Q in this archive, so its revenue, gross profit, operating income and net income come from inc_q and its earnings surprise from earn_cal, and both are labelled as vendor-sourced; annual figures for fiscal 2020 through fiscal 2025 from the vendor payload · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873601 = 2026-08-04T20:00:01Z ($1,335.03, −0.69%; 50-DMA $1,437.06; 200-DMA $1,214.35; RSI 48.4; MACD −20.43) · knowledge-base claims 2026-08-04. Monolithic Power's fiscal year is the calendar year. All figures come from the Synthos vendor data file for MPWR or from the SEC filings in the MPWR archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-27 (fiscal 2025); 10-Q filed 2025-11-05 (September 2025 quarter); 10-Q filed 2026-05-04 (March 2026 quarter — source of the 49,129-thousand share count, the $3,677.4 million equity balance and the $2.00 per share dividend declaration); 8-K filed 2026-04-30 (Item 2.02, first-quarter results, signed by Robert Dean as INTERIM Chief Financial Officer; the Exhibit 99.1 press release is furnished but not in this archive); 8-K filed 2026-06-12 (Items 5.07 and 8.01 — the 2026 annual meeting of 2026-06-11, electing two Class I directors to three-year terms to 2029 and confirming a classified board, with Victor K. Lee drawing 4,440,848 withheld votes, 10.2%, against Jeff Zhou's 6.0%; Ernst & Young ratified; say-on-pay approved with 96.8% support); 8-K filed 2026-07-30 (Item 2.02, second-quarter results, again signed by Robert Dean as INTERIM Chief Financial Officer; exhibit likewise not in this archive). No 10-Q covering the June 2026 quarter exists in this archive. All carry preserved [TABLE] statement data — 80 tables in the March 10-Q.
• Where the filings contradicted or corrected the vendor (detailed in Section 6): est.ebitdaAvg/ebitAvg, which trip BOTH documented signatures — fixed at exactly 28.626% and 26.598% of revenueAvg in every year from FY2023 to FY2030, AND with netIncomeAvg EXCEEDING ebitAvg in every forward year, which is arithmetically impossible for a taxpaying company; the FY2029 and FY2030 estimate rows, each resting on ONE analyst for EPS and excluded; seg_prod, whose most recent row is fiscal 2023 and which carries no product split for FY2024 or FY2025; and the FY2024 annual and December-quarter earnings figures, distorted by an income-tax BENEFIT of $1,213.8 million producing FY2024 EPS of $36.76 against FY2025's $12.82 on 26.4% higher revenue — every trailing multiple spanning FY2024 is meaningless and none is used. The two corporate-governance facts in this dive — the interim chief financial officer signing two consecutive earnings 8-Ks, and the classified-board election of 2026-06-11 — appear only in the filings and in no vendor field. Where vendor and filing AGREED — recorded, because this is one of the cleaner payloads in the batch: the share count is an EXACT match (10-Q 49,129 thousand at 2026-03-31 against 49,130,000 implied); seg_geo reconciles to within 0.0007% of consolidated revenue in FY2025 and is consistent back to fiscal 2016, which is what makes the 75.0% greater-China finding usable; capitalExpenditure is CLEAN with the free-cash-flow arithmetic tying exactly in all four annual rows; enterprise value carries no possible omission, the company holding net cash with minorityInterest and preferredStock both zero; totalDebt of $24.1M is correctly composed of $20.0M of operating-lease obligations and $4.1M of other short-term items, with interestExpense at $0 in every year; and tech.max_dd_from_peak of −20.999% coincides with pct_from_hi correctly.
• Estimate coverage: 9 analysts on FY2026 EPS and 11 on revenue; 9 on FY2027 EPS and 11 on revenue — the anchor for all three fair values, with an EPS range of $32.390 to $39.645, a 22.4% spread; 5 on FY2028 EPS; ONE analyst each on FY2029 and FY2030 EPS, both excluded from every conclusion.
• Peer note: the vendor peer set — ASE Technology, Electronic Arts, Garmin, NXP Semiconductors, Seagate, Atlassian, Take-Two Interactive, Ubiquiti, Western Digital, Block — is substantially unusable, containing two video-game publishers, a payments company, an enterprise-software company and two hard-disk makers. Only NXP Semiconductors and ASE Technology are semiconductor businesses and neither is an analog power-management designer. Texas Instruments, Analog Devices and Infineon — the companies against which this business is actually judged — are all absent. No peer-multiple comparison is drawn.
• Fair-value caveat: the $850 / $1,530 / $2,000 anchors are multiples of the FY2027 consensus EPS distribution — 26.2x the low of $32.390, 43.9x the mean of $34.816, and 50.4x the high of $39.645 — cross-checked against FY2026 and FY2028. Stated arithmetic, not a discounted cash flow. The range is deliberately wide because two things justify it: an analyst EPS spread of 22.4%, and a largest risk that is binary and political rather than continuous. The base is sensitivity-disclosed: 38.3x FY2027E gives $1,335, which is spot exactly; 45x gives $1,567. Section 4b states that the base assumes earnings grow 28.2% while the multiple compresses 11%, and that the two nearly cancel — so the expected return is entirely earnings and slightly negative on re-rating. The base of $1,530 is 15.0% BELOW the street consensus of $1,800.71, whose lowest published constituent of $1,500 is already 12.4% above the current price.
• Timing: second-quarter 2026 results were released 2026-07-30, five days before this dive, and BEAT the consensus EPS estimate by 10.7% ($6.50 against $5.87) on revenue 8.6% above estimate — the fifth consecutive beat and, at 10.7%, by far the largest after 2.2%, 1.9%, 1.1% and 4.1%. The next print is 2026-10-29, 86 days away, on consensus EPS of $7.62 and revenue of $1,140.6M, implying a 16.3% sequential step on the June quarter's $980.6M. The insider block contains no transactions, so no insider inference of any kind is drawn. 2026-08-04 carried no company filing, and the stock closed at $1,335.03 — $1.45 above the exact low of a $92.41 intraday range, having opened at $1,402.10, on approximately $1.62 billion of turnover.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.