MPWR — Monolithic Power Systems, Inc.

The Street and the tape disagree sharply, which is why this reads Neutral rather than Bullish. Analyst sentiment is loudly positive: 22 Buy / 3 Hold / 0 Sell, a consensus price target of $1,827.25 (recent-average $1,953.25), and StoneX initiating with a Buy and $1,800 target in September 2026 following the July 2026 Q2 blowout (revenue $980.6M, +47.6% YoY, adjusted EPS $6.50 vs $5.87 expected, plus a Q3 guide ~16% above consensus). But the market itself is voting the other way — as of 2026-09-13 MPWR was -13.37% over one month and -21.74% over three months, sitting below its 20-, 50- and 200-day averages. That gap is best explained by multiple compression across high-P/E semis (10-year Treasury yields drifting toward 4.8%) rather than by anything in the P&L, but a 21.7% drawdown in a name that just beat and raised is real information, not noise.

Revenue has inflected hard, and the acceleration is getting stronger, not fading: $474.9M (Q3 2023) → $620.1M (Q3 2024) → $737.2M (Q3 2025) → $980.6M (Q2 2026). YoY growth ran 18.9% (Q3 2025), 20.8% (Q4 2025), 26.1% (Q1 2026), then 47.6% (Q2 2026), with Q2 up 21.9% sequentially — a genuine step-change, not a blip. Gross margin is strikingly stable at 55.1–55.4% across all 12 quarters (55.18% in Q2 2026), so essentially all the upside is operating leverage: operating margin troughed at 20.85% in Q1 2024, recovered to ~26.5% through 2025, then jumped to 29.99% (Q1 2026) and 30.99% (Q2 2026) as R&D grew only 23% YoY against 47.6% revenue growth. Diluted EPS went $2.78 (Q2 2025) → $5.23 (Q2 2026), +88%. Two caveats worth flagging plainly. First, Q4 2024 shows net income of $1.449B against operating income of only $163.3M — an obvious one-off non-operating/tax gain that has nothing to do with the operating business and that poisons any trailing-EPS comparison. Second, cash conversion is lagging: TTM operating cash flow of ~$822M is actually down ~8% YoY, TTM free cash flow of ~$586M is down ~17%, and Q2 2026 capex nearly doubled to $82.4M from $50.1M. That is almost certainly working-capital build ahead of the ramp rather than earnings-quality decay (the accruals ratio is only +0.63% of total assets as of 2026-06-30), but it is the one line in the model that is not confirming the story.

As of 2026-09-13 at $1,234.46, the trailing optics look expensive: 75.27x TTM P/E, 18.53x TTM P/S, 15.57x book, 59.10x EV/EBITDA. Two of those are distortions rather than signals. The PEG of -1.31 is a pure artifact — the prior-year TTM EPS base includes the $29.88 one-off quarter (Q4 2024), so 'EPS growth' mathematically prints negative. And TTM revenue understates the current run-rate: annualizing Q3 2026 guidance (~$1.15B) with a normal Q4 gets FY2026 revenue to roughly $4.15B, which puts P/S at ~14.6x — essentially dead on the 5-year average of 14.61x. On earnings, my own FY2026E GAAP EPS is ~$22.1, putting the stock at ~56x forward versus a 5-year average P/E of 57.64x; on the $27.42 FY2027 consensus EPS in the dataset, it is only 45x. Verdict: expensive on trailing multiples, roughly fair to modestly cheap versus its own history on forward numbers, and unambiguously priced for the AI power-module ramp to keep compounding. Multiple compression, not earnings, is the main downside channel from here.

This is the cleanest part of the setup: price and estimates are moving in opposite directions. Q3 2026 guidance of $1.14–1.16B sits ~16% above the ~$982M consensus that existed before the July 2026 print, and forward EPS estimates are being revised up hard — yet as of 2026-09-13 the stock is -21.74% over three months. That is price lagging earnings, not price running ahead of them, which is the more favorable of the two mismatches. The honest counter-read: a 21.7% drawdown in a company that just beat and raised usually means the market is discounting something a quarterly print cannot show — most plausibly that the Enterprise Data surge (+45% sequential in Q2 2026) is a cyclical spike rather than a durable step-up, or simply that 75x-trailing-P/E semis de-rate when the 10-year pushes toward 4.8%. Playbook: if Q3 lands at or above $1.16B, the revision tailwind should eventually win and this looks like an estimate catch-up trade; if it merely lands inside the range, expect further multiple compression regardless of the print.

As of 2026-09-13, RSI(14) was 43.08 — neutral, neither oversold nor overbought, with no capitulation signal. Price at $1,234.46 sat 3.3% below the 20-day (≈$1,277), 6.38% below the 50-day (≈$1,319) and 1.83% below the 200-day (≈$1,258), with the previous close at $1,186.08 and a +4.08% session on the last print, which looks like a bounce attempt off the lows. The term structure matters more than the levels: -21.74% over 3 months but still +17.28% over 6 months, inside a 52-week range of $831–$1,714. That is a sharp pullback/consolidation within a larger uptrend, not a broken downtrend. Implication for timing: this is a reasonable accumulation zone rather than a falling knife — RSI at 43 leaves room to run, and the 200-day at ~$1,258 is the pivot. A decisive close back above it confirms the pullback is over; a break below ~$1,150 would argue the market is pricing something real that the last quarter did not show.

• AI datacenter capex cycle: the dominant demand driver and the reason Q2 2026 Enterprise Data sales spiked 45% sequentially. Any hyperscaler capex digestion would hit MPWR's fastest-growing segment first.
• Interest rates: the 10-year Treasury drifting toward 4.8% is the most plausible explanation for a 21.7% three-month drawdown in a 59x EV/EBITDA name — rate-driven multiple compression hits the highest-multiple semis hardest regardless of earnings.
• Sector rotation: capital rotating out of crowded high-P/E AI winners; MPWR is caught in that flow even while its own estimates rise.
• Analog/power semiconductor supply chain: tight leading-edge and specialty capacity is a real constraint, which is precisely what the September 9, 2026 GlobalFoundries 300mm Singapore agreement is designed to relieve — but volume production does not start until early 2027.
• Trade and export policy: power-management silicon has meaningful China automotive and industrial exposure, leaving the sector exposed to tariff and export-control headlines that are outside the company's control.
• End-market cyclicality: automotive and industrial (the diversification that offsets AI concentration) are themselves mid-cycle cyclical and can soften without warning.

• Q3 2026 earnings, expected late October / early November 2026: the single biggest near-term event, testing whether revenue lands inside or above the $1.14–1.16B guide. A beat would validate the acceleration and likely drag estimates toward the $1,800+ Street targets; landing mid-range would leave the multiple exposed.
• GlobalFoundries Singapore 300mm fab ramp: volume production slated for early 2027, with progress updates through late 2026. A credible capacity unlock directly de-risks the supply constraint and supports the 2027 revenue ramp.
• Continued Enterprise Data design wins and power-module content growth in next-gen GPU/CPU platforms — the segment that drove the +45% sequential Q2 2026 surge.
• Estimate revisions and price-target resets: with only 4 analysts in the recent-target pool (avg $1,953) against 25 ratings, broader coverage updates after Q3 could widen the gap between the tape and the Street.

• Multiple compression risk, which is the primary one: 75.27x TTM P/E and 59.10x EV/EBITDA leave no room for disappointment. A 10% P/E de-rating on unchanged earnings costs ~$120/share.
• AI datacenter concentration: Enterprise Data drove the sequential spike and is now the swing factor in the growth story. A single hyperscaler capex pause or a share loss in a power-stage socket would compress both estimates and the multiple simultaneously.
• Cash generation not confirming the P&L: TTM operating cash flow of ~$822M is down ~8% YoY and TTM FCF of ~$586M is down ~17% while revenue accelerated. Read as working-capital build today, it becomes a genuine quality problem if it persists past the ramp.
• Cyclicality: this is a semiconductor company growing 47.6% YoY. The analog/power cycle can turn, and the current growth rate is by definition unsustainable — the question is only where it normalizes.
• Opex creep: R&D rose 23% YoY in Q2 2026 and capex nearly doubled. If the AI ramp slows before that spending is absorbed, the operating-margin expansion that drove the EPS beat reverses quickly.

This is a high-quality growth business having an exceptional run that the market is currently distrusting. Q2 2026 revenue was up 47.6% YoY with guidance ~16% above consensus, operating margin has expanded from ~25% to ~31% in three quarters, gross margin is rock-steady at ~55%, and there is zero long-term debt against $1.0B of cash. On forward numbers the valuation is far less scary than the 75x trailing P/E suggests — roughly 56x my FY2026E EPS of ~$22.1 (right at its 5-year average of 57.6x) and 45x the $27.42 FY2027 consensus EPS. The stock is 21.7% off its three-month-ago level while estimates rise, which historically favors patience over panic. It is not a bargain and multiple compression remains the dominant downside channel, but this is a good business priced at roughly its own historical average on the numbers that actually matter.

Applied a 52x forward P/E — a ~10% discount to MPWR's own 5-year average trailing P/E of 57.64x, reflecting the working-capital-driven FCF drag and the risk of rate-driven multiple compression — to the $27.42 FY2027 consensus EPS in the dataset, giving $1,426, rounded to $1,425. Cross-check: the 2026-09-13 price of $1,234.46 implies only ~45x those same FY2027 earnings and ~56x my own ~$22.1 FY2026E EPS, so the fair-value gap is roughly 15% and depends on the AI power-module ramp delivering rather than merely persisting.