{
  "@context": "https://frontierpicks.com/schemas/dossier.v1.json",
  "ticker": "POWI",
  "name": "Power Integrations, Inc.",
  "url": "https://frontierpicks.com/dossiers/POWI/",
  "json_url": "https://frontierpicks.com/dossiers/POWI.json",
  "status": "DORMANT",
  "current_conviction": "LOW",
  "graded_conviction": "MEDIUM",
  "archetype": {
    "code": "a3",
    "n": 3
  },
  "current_thesis": "The 800VDC AI-power re-rate has cracked. POWI lost the $74–78 May breakout shelf in a -20% week to $70.45 (2026-07-17) and now sits below the $76.20 average target for the first time since spring. The GaN cohort broke with it — NVTS is -41% in a month on a Wolfspeed patent suit and ON Semi's GaNEXUS entry. The 2026-08-05 Q2 print is where narrative has to convert into disclosed data-center dollars.",
  "invalidation_trigger": "A weekly close below $60 — the July swing low and the pre-breakout base — completes the round-trip of the entire 800VDC re-rate and ends the 2026 uptrend. Secondary: the 2026-08-05 Q2 print passing without a named AI data-center design win or quantified bookings, or the GaN cohort (NVTS, ON, MPWR) printing fresh lows together.",
  "catalyst_date": "2026-08-26",
  "outcome": "PLAYED_OUT",
  "outcome_date": "2026-06-15",
  "invalidation_fired": false,
  "themes": [
    "ai-datacenter-infrastructure",
    "industrial-power-grid"
  ],
  "tags": [],
  "sources": [],
  "notes": [
    "2026-06-05: broke -9.71% to ~$77 (AH ~$75.98) in a SECTOR-WIDE semi selloff (Broadcom soft AI/hyperscaler outlook, Samsung strike risk, TSMC stake-sale chatter, rare-earth supply, hot jobs print) — Microchip fell with it. NOT company-specific news.",
    "Non-GAAP and GAAP diverge widely here (Q2 2026: $0.37 non-GAAP vs $0.17 GAAP) — check which figure a headline is using.",
    "Data-center exposure was still reference-design and engagement stage, with no named customer or booking figure disclosed, as of the 2026-08-05 call.",
    "Levels quoted in coverage before August 2026 came from an unadjusted price series; levels here reference split/dividend-adjusted closes and differ by a few percent.",
    "Moving-average and RSI feeds on this name have returned inconsistent values; swing levels and the 52-week high are the more reliable structure.",
    "Insider sales clustered in late May 2026, into the parabolic leg — before the July break, not after it.",
    "Target aggregators disagree: $77.50 across 5 analysts (stockanalysis.com) vs a $72.67 consensus and a 4 Buy/3 Hold/1 Sell mix (MarketBeat), August 2026."
  ],
  "body_markdown": "\n_All price references are split/dividend-adjusted closes; the last completed daily close is $55.18 (2026-08-21)._\n\n## POWI — Power Integrations, Inc.\n\n## Current Thesis\n\nThe level this coverage published on 2026-08-15 as the thesis break — a weekly close below $60, the July swing low and the last shelf under the 2026 advance — was taken out. The weekly close on 2026-08-21 was $55.18, 36.8% below the $87.35 52-week high, with a three-month price change of -21.9% and RSI(14) at 41.3.\n\nThat closes the loop on the leg an investor was buying between April and June: NVIDIA's Kyber architecture moving AI rack power to 800VDC, with Power Integrations' 1250V/1700V PowiGaN as one of few volume-production parts sized for the socket. On the 2026-08-05 Q2 call management placed auxiliary data-center power as a potential **2028** revenue item, the GPU main-power path further out, and called the newly demonstrated 2,200V PowiGaN a technology milestone rather than a launch, with meaningful revenue several years away. No named customer, no bookings figure, no quantified data-center contribution. Stifel (Tore Svanberg, $95→$80) and Needham (David Williams, $90→$80) both cut on 2026-08-06 while keeping Buy ratings.\n\n**The narrative is dead** for the 800VDC leg, and the dates carry it: the May breakout shelf went in a -20.21% week into 2026-07-17; the payoff moved to 2028 on 2026-08-05; the sell-side marked down on 2026-08-06; the July low went on the weekly close of 2026-08-21. What remains underneath is a cyclical power-semiconductor recovery — Q3 guided to $122–130M against $125.66M consensus, gross margin 54–55%, operating margin 17–19%, channel inventory back to 7.3 weeks — which is a different security than the one bid in May and does not carry a narrative premium.\n\n## Bull Case\n\n- **Q2 2026 (2026-08-05) beat both lines:** revenue $118.939M vs $117.440M consensus, +10% sequentially and +3% YoY; non-GAAP EPS $0.37 vs $0.32.\n- **Margin inflected, not just revenue:** non-GAAP gross margin 55.1%, up ~160bp sequentially; operating margin 17.1% against 12.7% in Q1 2026.\n- **Q3 guide steps up again (2026-08-05):** $122–130M, a midpoint that brackets the $125.66M consensus, with operating margin guided 17–19%.\n- **Channel is clean:** 7.3 weeks of inventory, inside the company's target range and down from above 9 weeks entering 2026, so sell-in is not being flattered by restocking.\n- **Industrial carries the cycle:** 43% of Q2 sales and +16% YoY; communications +16% sequentially, consumer +5%, computer +5%.\n- **Roadmap extended (2026-08-05):** PowiGaN demonstrated at 2,200V, above the 1250V/1700V parts already shipping, for data centers, HVDC and EV.\n- **Sell-side sits well above the tape:** stockanalysis.com shows a $77.50 consensus across 5 analysts (2 Strong Buy / 1 Buy / 2 Hold), listed as 40.45% above market as of 2026-08-23; MarketBeat's August 2026 page shows a $72.67 consensus and a 4 Buy / 3 Hold / 1 Sell mix.\n- **Capital return affirmed:** $0.215 per share, record 2026-08-31, payable 2026-09-30.\n\n## Bear Case\n\n- **The payoff moved out three years on 2026-08-05.** A story bought for near-term delivery became a 2028 option in one call, and nothing since has dated it more precisely.\n- **The round-trip is now on the tape.** $74–78 shelf lost 2026-07-17; $60 July low lost on the weekly close of 2026-08-21 at $55.18, five sessions after the 2026-08-14 close of $62.75.\n- **Targets have not been marked to the tape.** Both $80 marks were set on 2026-08-06 with the stock materially higher than the 2026-08-21 close; the gap closes by price rising or by further cuts, and one cut round has already happened.\n- **The GAAP base is small.** Q2 2026 GAAP EPS was $0.17 against $0.37 non-GAAP, so headline multiples depend heavily on which figure is used.\n- **No company-scheduled event can re-price it inside the window.** The only dated company items in the next 30 days are dividend mechanics; the next real test is the Q3 print, estimated ~2026-11-04.\n- **Coverage is thin and split.** Two aggregators disagree on both the consensus target and the rating mix, and one of them prints a Sell.\n- **Insider selling clustered in late May 2026,** into the parabolic leg — ahead of the July break, not after it.\n\n## Setup & Price Structure\n\nThe decline since 2026-07-17 has the shape of drift rather than capitulation: RSI(14) at 41.3 on 2026-08-21 is neither oversold nor showing a positive divergence, and the move from the 2026-08-14 close of $62.75 to the 2026-08-21 close of $55.18 happened without a dated company headline in that window. A base has not formed.\n\nOverhead, $60 has flipped from support to the first thing a recovery has to prove: a weekly close back above it would be the earliest structural evidence that the round-trip has ended. Underneath, there is no shelf from the 2026 advance visible between the July low and the spring launch zone — an inference from the structure, not a level anyone defended. For gradeability, $52 is the next marker; a weekly close under it would say the decline extended past the July break with no base attempted.\n\nPositioning observables, stated as observables: sell-side targets 40%-plus above spot per stockanalysis.com on 2026-08-23; a one-time Russell 2000 Dynamic index add in June 2026 whose flow is behind; retail-facing coverage since 2026-08-05 framed around margin and P/E rather than AI content; no imminent earnings date to force repositioning. Moving-average feeds on this name have returned inconsistent values, so swing levels and the 52-week high are the more reliable reference points here.\n\n## Catalyst Calendar (next 30 days)\n\n- **2026-08-26** — NVIDIA Q2 fiscal 2027 results (quarter ended 2026-07-26), call 2 p.m. PT, per NVIDIA's scheduling release. The 800VDC/Kyber rack timeline and hyperscaler capex commentary that underwrite the whole power-delivery story are set here, not in POWI's own disclosure.\n- **2026-08-31** — Dividend record date for the $0.215 per share declared 2026-08-05 (some data providers list this date as the ex-date). Mechanical; resolves nothing about data-center revenue.\n- **2026-09-30** — Dividend payment date.\n- **~2026-11-04 (est.)** — Q3 2026 results; outside the 30-day window. The company reported Q3 on 2025-11-05 and 2023-11-07 in prior years. This is the next binary: whether revenue lands inside $122–130M and whether any data-center content is finally quantified.\n\n## What Would Change Our Mind\n\nThe structural question has already been answered once: $60 was the last shelf under the 2026 advance and it did not hold on 2026-08-21. The forward question is whether the cyclical business floor is priced. A weekly close below $52 would say the decline has run past the July break without an attempt at a base, and would put the price below every reference from the re-rate. A secondary condition sits on 2026-08-26: if NVIDIA's Q2 call passes without an 800VDC rack schedule that implicates the high-voltage GaN supply chain, the residual option value in the narrative has no dated support before November.\n\nWhat would rebuild the case, and it is specific: two consecutive weekly closes back above $60, or a named data-center customer with a dollar figure — design-win value or bookings — at the Q3 print (~2026-11-04, est.). What would deepen the bear read: Q3 revenue printing below the $122M guide floor, gross margin under the 54% low end, or a target cut below the 2026-08-06 $80 marks.\n\n## Correlation Notes\n\n- Moves with the GaN/power cohort — NVTS, ON Semiconductor, MPWR. NVTS fell 41% in a month around the Wolfspeed patent action and ON's GaNEXUS entry in July 2026; a cohort-wide fresh low is the signal that this is sector de-rating rather than company news.\n- High macro/sector beta on days with no company news: -9.71% on 2026-06-05 in a broad chip selloff (Broadcom outlook, jobs print), a 5.3% Friday drop to a $67.11 close in a separate sector-wide session, and a tariff-driven chip selloff in July 2026.\n- The 800VDC read-through runs through NVIDIA's disclosure on 2026-08-26 before it runs through anything POWI publishes.\n- Small-cap and rate beta via the Russell 2000; the June 2026 Dynamic Index add was one-time and mechanical.",
  "first_seen": "2026-04-23",
  "last_analyzed": "2026-08-23T16:14:16+00:00",
  "last_synthesized": "2026-08-23",
  "last_update_source": "watchlist_research",
  "license": "Content © FrontierPicks. Cite the canonical URL."
}