{
  "@context": "https://orbyd.app/schemas/dossier.v1.json",
  "ticker": "MPWR",
  "name": "Monolithic Power Systems, Inc.",
  "url": "https://orbyd.app/dossiers/MPWR/",
  "json_url": "https://orbyd.app/dossiers/MPWR.json",
  "status": "DORMANT",
  "current_conviction": "LOW",
  "graded_conviction": null,
  "archetype": {
    "code": "a2",
    "n": 2
  },
  "current_thesis": "AI data-center power narrative still accelerating fundamentally (Enterprise Data ~2x YoY to $262.8M Q1; Q2 guide $890–910M, +35% YoY), but the tape bounced ~12% off the ~$1,200 shelf to ~$1,468 and now stalls at the 50-day into a binary 2026-08-03 Q2 print. Restatement, an unremediated material weakness and a one-way insider wall keep a fresh entry low-conviction until the print clears.",
  "invalidation_trigger": "A weekly close below $1,200 loses the rising 200-day / 20-week EMA shelf that caught the June low, turning the correction into a trend change toward the 52-week base. Secondary: a 2026-08-03 Q2 print under the $890M guide floor, Enterprise Data's first sequential decline, or the AI-datacenter-power theme flipping to SATURATED.",
  "catalyst_date": "2026-08-03",
  "outcome": "OPEN",
  "outcome_date": null,
  "invalidation_fired": null,
  "themes": [
    "ai-datacenter-infrastructure",
    "ai-chips-memory",
    "industrial-power-grid"
  ],
  "tags": [],
  "sources": [],
  "notes": [
    "Enterprise Data (AI datacenter power, $262.8M Q1, ~2x YoY) is BOTH the bull thesis and the single point of failure — watch for any Rubin-gen NVIDIA power-architecture / socket-share-loss headline; that re-fires the 2024 Blackwell fear.",
    "Archetype: picks & shovels — power-management content supplier to NVIDIA accelerator boards; co-moves with SOXX/SMH/NVDA and analog-power peers (Power Integrations, Amkor, Infineon). Not a retail squeeze despite the leveraged-ETF noise.",
    "Q2 2026 guide: revenue $890–910M (~+35% YoY), GAAP GM 55.1–55.7% (given 2026-04-30). Beat-and-raise with Enterprise Data extending = act trigger; guide-floor miss or sequential Enterprise Data decline = invalidation.",
    "RESTATEMENT OVERHANG: 8-K 2026-02-26 non-reliance; restated 10-K filed ~2026-02-27. Non-cash deferred-tax error on a one-time foreign tax incentive cut FY2024 income-tax benefit by $194.6M; 2025 net income +$5.5M. Revenue and non-GAAP metrics UNCHANGED, but a material-weakness governance discount until remediation is confirmed.",
    "INSIDER SELLING WALL: ~127 sales / 0 buys trailing 6mo. One-directional selling into the highs — first insider buy or a selling pause is the sentiment inflection to flag.",
    "PT range now sits ABOVE spot (~$1,468, mid-July): low target lifted to $1,575, KeyBanc high $2,000, average ~$1,789–1,800, median ~$1,662, 15-analyst Strong Buy. Headroom restored after the June drawdown.",
    "RETAIL-SATURATION FLAG: 2X leveraged single-stock MPWR ETF launched 2026-04-24 (Tradr) — late-cycle product signal, watch for crowd unwind on any theme wobble.",
    "Theme read: narrative ACCELERATING fundamentally, tape MATURING after the June distribution/bounce; IBM July AI-capex commentary reheated demand backdrop. The 08-03 print referees pullback-vs-top."
  ],
  "body_markdown": "## Current Thesis\nThe fundamental narrative is still accelerating; June was the tape stress-testing whether that mattered. MPWR sells power-management content into NVIDIA accelerator boards, and Q1 2026 (reported 2026-04-30) put hard numbers on the AI leg: Enterprise Data revenue nearly doubled YoY to $262.8M, total revenue set a record $804.2M (+26.1% YoY), and the Q2 guide of $890–910M implies ~+35% YoY — a content ramp still mid-cycle. Broadcom's 2026-06-03 soft AI-chip guide repriced the whole accelerator-power complex, and MPWR took three ~9–10% down days into a 06-26 low of $1,313.32, roughly 23% off the $1,714.09 high and under the 50-day. Since then the structure changed: the tape based on the rising 200-day near $1,200 and bounced ~12% back to ~$1,468 (mid-July), aided by IBM's July AI-infrastructure-spend commentary reheating the theme. The result is a name sitting mid-range, testing the 50-day (~$1,470) from below, ~16 days ahead of a binary Q2 print, with a restated 10-K and a one-directional insider-selling wall still overhead. Narrative intact, structure unresolved — the 2026-08-03 print settles it.\n\n## Bull Case\n- **Enterprise Data ~2x YoY to $262.8M in Q1 2026** (reported 2026-04-30) — direct evidence MPWR is holding power-management content on AI accelerator boards, refuting the Nov-2024 fear it ceded Blackwell vertical-power sockets to Infineon/Renesas.\n- **Q2 2026 guide $890–910M** (given 2026-04-30), ~+35% YoY at the midpoint off a record quarter, GAAP GM guided 55.1–55.7% — sequential acceleration, not a stall.\n- **Communications $111.5M, +55.5% YoY / +33.1% QoQ** (Q1 2026) — a second end-market firing alongside data center, broadening the story past a single line.\n- **Margins held through the mix shift** (Q1 2026): GAAP GM 55.3%, operating profit $241.2M (+42.9% YoY), non-GAAP EPS $5.10 vs the $4.90 estimate.\n- **The entire sell-side range now sits above spot** (July 2026): 15-analyst Strong Buy, average target ~$1,789–1,800, high $2,000 (KeyBanc), and a low target lifted to $1,575 — even the floor implies ~+7% on the ~$1,468 quote, with headroom restored after the June drawdown.\n- **The 200-day held.** The selloff bottomed at $1,313.32 (06-26) on the rising 200-day / 20-week EMA shelf and reversed ~12%, and IBM's July AI-capex commentary handed the theme a fresh demand data point.\n\n## Bear Case\n- **The bounce is testing the 50-day from underneath.** Price reclaimed ~$1,468 into the declining 50-day (~$1,470) without a decisive volume close above it; a rejection here re-opens the $1,313 June low and then the $1,200 shelf.\n- **A binary print sits ~16 days out.** Q2 2026 reports 2026-08-03; after a +35%-YoY guide the bar is high, and Broadcom's 06-03 miss showed the complex will punish any hint of AI-power demand decelerating.\n- **Material weakness is a live governance discount.** The restated 10-K (filed ~2026-02-27) cut the FY2024 income-tax benefit by $194.6M off a deferred-tax error on a one-time foreign tax incentive (2025 impact +$5.5M net income); revenue and non-GAAP metrics are unchanged, but the internal-controls flag stays a discount until remediation is confirmed.\n- **The insider wall has not paused.** Roughly 127 insider sales and zero buys trailing six months — one-directional selling into the highs, with no first buy to mark a floor.\n- **Retail-saturation flag.** A 2X leveraged single-stock MPWR ETF launched 2026-04-24 (Tradr) — the kind of product that shows up once a move is already crowded.\n- **Valuation leaves no cushion.** The name still trades at a premium to analog-power peers and well above conservative fair-value estimates; a guide-floor miss on 08-03 removes multiple support fast.\n\n## Setup & Price Structure\nAs of mid-July 2026 the quote is ~$1,468.35, down ~14% from the $1,714.09 52-week high and up ~12% from the 06-26 low of $1,313.32. The picture is a mean-reversion bounce stalling at the 50-day (~$1,470): the rising 200-day / 20-week EMA near $1,200 caught the June flush, the $1,313 low is the near-term higher-low reference, and $1,470–1,500 is the pivot the tape must clear to confirm a pullback read over a lower-high read. Momentum reset without overshooting — the RSI sits mid-range after working off June's oversold reading, so there is no extreme to lean on in either direction. The clean read: a fresh entry into a mid-range name testing overhead resistance ~16 days before a binary print is low-quality; the higher-probability setups are a volume reclaim of $1,500 that holds, or the post-print reaction traded off the $1,200 shelf. Standing aside into 08-03 is the better path than front-running it.\n\n## Catalyst Calendar (next 30 days)\n\n- **Late July (est.) — restatement remediation cadence.** Watch any interim 8-K/10-Q language on progress remediating the internal-controls weakness tied to the deferred-tax restatement; a clean interim would lift the governance discount.\n- **Ongoing — Form 4 flow.** The trailing six months ran ~127 sells / 0 buys; a first insider buy, or a clear pause in selling, is the early sentiment inflection to flag.\n- **Ongoing — AI-capex read-through.** Hyperscaler and IBM AI-infrastructure commentary (the July IBM signal already reheated the theme) sets the demand backdrop MPWR's Enterprise Data line prints against.\n\n## Elapsed catalysts\n\n- **2026-08-03 — Q2 2026 earnings (after close).** The binary. Guide is $890–910M revenue (~+35% YoY), GAAP GM 55.1–55.7%; a beat-and-raise with Enterprise Data extending its ramp confirms the leg, while a guide-floor miss or a first sequential Enterprise Data decline breaks it. Avoid fresh entries inside the 3-trading-day window before the print. *(passed 6d ago)*\n\n## What Would Change Our Mind\nBullish confirmation: a weekly close back above $1,500 that holds, plus a 2026-08-03 print that clears the $890–910M guide with Enterprise Data extending sequentially — that converts the June drawdown into a completed pullback and re-arms the accelerating narrative for a run at the ~$1,662 median target and beyond. Bearish invalidation: a weekly close below $1,200 loses the rising 200-day / 20-week EMA shelf that caught the June low and turns the correction into a trend change toward the 52-week base; a guide-floor miss on 08-03, an Enterprise Data sequential decline, or the AI-datacenter-power theme flipping to SATURATED would each independently break the leg. Between those markers the name grades as a wait-for-the-print, low-conviction setup.\n\n## Correlation Notes\nMPWR is a picks-and-shovels supplier, not a retail squeeze despite the leveraged-ETF noise — it co-moves with SOXX / SMH / NVDA and the analog-power peers (Power Integrations, Amkor, Infineon). The concentrated risk is NVIDIA socket share: Enterprise Data ($262.8M Q1, ~2x YoY) is simultaneously the bull thesis and the tail, so a Rubin-generation NVIDIA power-architecture or socket-share-loss headline is the specific event that re-fires the 2024 Blackwell fear. Beta to the broad AI-infrastructure trade is high; on down days the name printed 9–10% air-pockets alongside the semis, so sizing has to assume index-correlated drawdown. The June episode showed the correlation cuts both ways: Broadcom's guide dragged it lower, IBM's capex commentary helped drag it back.",
  "first_seen": "2026-04-24",
  "last_analyzed": "2026-07-18T08:21:52+00:00",
  "last_synthesized": "2026-07-18",
  "last_update_source": "watchlist_research",
  "license": "Content © orbyd. Cite the canonical URL."
}