Swing Trade Ideas — AI Infrastructure & Semis (US + HK)

Report date: May 30, 2026 (data as of close May 29, 2026)

Horizon: Catalyst-anchored — exit each idea at the cited catalyst regardless of timing

Bias: 3 longs / 2 shorts (net long, mostly into US AI capex prints)

Author note: This is a swing book sized for a 4–8 week window. Read the declared blind spots before sizing anything.

---

1. Executive Summary

• AI capex prints are the calendar. AVGO confirmed for June 3 (Q2 FY26, AMC); MU expected last week of June (estimated). The book is built around these.
• The Friday tape was bifurcated. Networking + memory ripped (AVGO +4.7%, MU +5.1%, CRDO +6.2%, ARM +5.4%, DELL +32.8%, Lenovo HK +22.0%) while NVDA, INTC, MPWR and the HK foundries (SMIC −7.5%, Hua Hong −5.1%) sold off on big volume. That's classic late-cycle leadership rotation, not a top — but it argues for picking spots in the laggards as shorts, not selling AI infra wholesale.
• Posture: Net long the US AI capex complex into the AVGO/MU print window; paired with technical-break shorts in INTC (broken parabolic) and SMIC (HK foundry geopolitics + roll-over) to hedge correlation and macro risk.

---

2. Methodology & Declared Blind Spots

This report uses live US/HK quotes via the local yfinance MCP, supplemented by WebFetch / Sonar-Pro web search for catalyst dates and the HKEX shortable list. Day-percent moves are computed as `(today_close − prior_close) / prior_close` from daily history, not from the (often-stale) previous_close field on the live quote.

The following cannot be checked from the data layer in this repo. Do not assume they were:

1. No options data. Implied vol, skew, and the implied move into AVGO/MU are unknown. The earnings-anchored ideas are directional, not vol trades.
2. No sell-side consensus / revisions. Forward EPS / NTM multiples are not verifiable. Where consensus is referenced, it is [UNSOURCED].
3. No positioning data. Short interest, days-to-cover, gamma, and CTA flow are unknown.
4. No Stock Connect Northbound/Southbound flow. A major HK driver — especially for SMIC and Lenovo — is invisible here.
5. yfinance is best-effort. Particularly around earnings prints, dual-listed ADRs, and HK names with halts. The quote.previous_close field shipped a wrong DELL Day% on May 29 — see the Watch List for what that means for DELL today.
6. HKEX shortable list verified for 0981.HK via HKEX's May 29, 2026 designated-securities revision (per Sonar-Pro / HKEX circular).

---

3. Regime Read

3.1 US: SOXX leadership intact, internals rotating

• SPY 756.48 — through 750, ATH territory, 50-DMA rising.
• SOXX 569.08 — extraordinary run from 323 (early March) to 569 (May 29) = +76% in three months. Friday's −0.6% with internal divergences is the kind of pullback that often resolves higher post-AVGO, but it is also the first sign of distribution we've seen since April.
• SMH 598.93 — same shape.
• Subsector relative strength (Friday close vs. 1-month):
• AI accelerators / silicon: AVGO strong (gap), MU strong (parabolic), ARM strong (breakout)
• AI servers: DELL +32.8% on what looks like an earnings/guidance event (volume 41.7M, ~3x avg); SMCI +11.6% sympathy
• Semis losing leadership: NVDA −1.5% (post-print drift), INTC −5.2% (broken parabolic), MPWR −4.1% (weakest of the day on volume)

3.2 HK: HSI rolling over, China-tech heavy

• HSI 25,182 — closed under the 25,500 weekly pivot, lower-high since the late-March 26,500 print. Below the weekly 50-DMA cluster around 25,800.
• HSCEI 8,425 — same shape.
• Internals divergence: Lenovo (0992.HK) +22.0% on its FY results print (likely best HK single-day in tech YTD), but the foundries SMIC (0981.HK) −7.5% and Hua Hong (1347.HK) −5.1% broke down on big volume after a May rally — geopolitics + competitive pressure narrative.

3.3 Catalyst calendar (4–8 weeks)

---

4. Trade Ideas — Summary Table

R/R is computed from the entry-zone midpoint to the named target vs. stop. All Day% values are recomputed from daily history per Section 2.

---

5. Idea-by-Idea Writeups

Idea #1 — LONG AVGO (Broadcom)

• Setup: AVGO gapped through the $430 multi-week consolidation top on Friday on 2.5x volume, closing at the high of day ($448.90 intraday high). Three trading days before a confirmed print, with the AI capex complex as the leading thread on tape. Stock has not traded above $448 prior to May 29 in the 3-month sample.
• Thesis:
• AI custom-silicon + AI networking read; AVGO is the most direct beneficiary of hyperscaler AI capex outside NVDA, and unlike NVDA the print is a fresh catalyst rather than a backwards look.
• Friday's volume profile (41.7M vs. ~17M avg) is institutional accumulation, not retail froth. 2:1 buy/sell tape at the open with a high-of-day close.
• The May 26–28 base ($422–427) becomes the floor; any pullback in the next two sessions to that zone is the optimal entry.
• Entry zone: $442–452 (mid $447). Either fill on a dip to the upper end of the gap or buy strength above $450 if it consolidates.
• Stop: $420 (−6.0%) — beneath the May 27 close ($421.86) and the prior consolidation top ($430), forfeiting the breakout if it round-trips.
• Target: $510 (+14.1%) — measured-move projection from the $385–430 base height onto the $430 pivot, plus continuation room post-print. Conservative target $480 if you scale.
• R/R: ($510 − $447) / ($447 − $420) = 63 / 27 = 2.33
• Sizing: Risk 0.5% of book at the stop. For a $1M book, ~185 shares (~$83k notional) costs $5,000 at the stop.
• Time stop: Exit at June 4 close, regardless of result. This is a directional event trade, not a vol trade — see blind-spot #1.
• Risks:

1. Implied move not measurable here. If AVGO has implied a ±8–10% move and prints in line, the directional long can lose without the thesis being wrong. Treat 1R as the realistic loss case, not 0.
2. AI custom-silicon (XPU) revenue beat is already partly priced after the gap. A "good but not great" print can sell off.
3. China revenue exposure to export-control commentary on the call is binary.

---

Idea #2 — LONG MU (Micron)

• Setup: MU has gone parabolic from $700 (May 1) to $971 (May 29) on the HBM/memory-pricing thesis — a +39% advance in four weeks. The three sessions May 27–29 are a tight $904–981 consolidation at the highs on rising volume. Friday's print is a fresh ATH close.
• Thesis:
• Memory super-cycle thesis intact: HBM3e supply is sold out into 2027 and pricing has stepped up ~3 quarters in a row [UNSOURCED — verify against MU IR before sizing].
• The Q3 (May Q) print captures pricing realization at peak DDR5 + HBM3e mix.
• Volume-weighted accumulation since May 1 (cumulative 800M+ shares) implies institutional positioning is not yet complete.
• Entry zone: $940–960 (mid $950). Use any pullback toward the May 27–28 lows ($904–909) as a deeper add. If no pullback, buy a tight breakout above $981.
• Stop: $890 (−6.3%) — beneath the May 26–28 closing zone ($895–928) and the May 22 last-look swing high (~$780). Wide enough that normal post-parabolic noise doesn't take you out. If your sizing demands a tighter stop, shrink the position — don't tighten this stop into a rip.
• Target: $1,100 (+15.8%) — measured-move from the ~$700 base onto $920 = ~$1,140; round to $1,100.
• R/R: ($1,100 − $950) / ($950 − $890) = 150 / 60 = 2.50
• Sizing: ~83 shares per $5,000 risk on a $1M book.
• Time stop: Exit at the next earnings print (estimated June 26) or by July 3, whichever first.
• Risks:

1. Catalyst date is not confirmed by IR yet (per Sonar-Pro web search); if MU pushes the print into July, the trade carries a longer time-stop than expected.
2. Parabolic trends end vertically — a 10% one-day reversal would not be unusual after this kind of run; the −7.4% stop accepts that risk.
3. China memory subsidization headlines (CXMT) are an overhang on long-term pricing — not a 4-week catalyst, but a sentiment risk.

---

Idea #3 — LONG ANET (Arista Networks)

• Setup: ANET pulled in from ~$170 in late April to a $148–162 base. Friday's +1.4% with the rest of the AI infra complex puts it back in the upper half of the base.
• Thesis:
• Arista is the cleanest pure-play AI-networking name; an AVGO beat with strong networking commentary tends to drag ANET 4–6% in sympathy.
• The base since May 12 ($148–168) is a textbook bull flag after the March–April advance from $115. Breakout buy zone is $168.
• As an FX-insensitive US name, ANET is the natural complement to MU (cyclical) and AVGO (event-driven) in this book.
• Entry zone: $156–162 (mid $159). Buy on the next pullback into the base, OR buy a breakout above $168 with a tighter stop.
• Stop: $148 (−6.9% from mid) — just below the May lows.
• Target: $185 (+16.4%) — prior March highs ($175) plus continuation, calibrated to the AVGO read-through.
• R/R: ($185 − $159) / ($159 − $148) = 26 / 11 = 2.36
• Sizing: ~454 shares per $5,000 risk.
• Time stop: July 17, 2026 (6 weeks) or stop, whichever first.
• Risks:

1. ANET re-rate is contingent on AVGO networking commentary; an AVGO beat driven entirely by custom silicon (XPU) without networking commentary would not move ANET.
2. The $148 swing low is heavily defended; a break would imply a topping pattern from $175, much wider downside than the stop captures.
3. NVDA Spectrum-X / InfiniBand share commentary at GTC / earnings is a recurring overhang for ANET multiples.

---

Idea #4 — SHORT INTC (Intel)

• Setup: INTC went vertical from $43 (Jan) to $130 high on May 11 — a near-3x in four months on AI-foundry thesis revival. Since May 11 it has been distributing in the $108–130 range. Friday's −5.2% on 2x volume broke the 5/22 swing low ($118.09) intraday and closed at the dead low.
• Thesis:
• Parabolic moves into resistance with 2x-avg volume distribution + a swing-low break are the textbook short setup.
• The 50-DMA is around $95 and the 5/13–5/15 gap fill region is $108–116 — overhead supply.
• Friday's intraday range ($113.54–$126.64) puts the close in the bottom 10% of the daily range, the highest-conviction technical bear bar.
• Entry zone: $116–122 (mid $119). Sell into a back-test of the broken support / 5/28 close zone.
• Stop: $128 (+7.6%) — above the 5/29 high ($126.64) with buffer. A close above $128 invalidates the breakdown.
• Target: $95 (−20.2%) — the 50-DMA cluster + the late-April breakout pivot zone ($82–95).
• R/R: ($119 − $95) / ($128 − $119) = 24 / 9 = 2.67
• Sizing: ~556 shares per $5,000 risk.
• Time stop: Cover on a daily close above $128 (50-DMA reclaim) or by July 24 (one week before estimated Q2 earnings), whichever first.
• Risks:

1. No options-positioning visibility — INTC is a heavily-optioned name and a gamma-driven squeeze can run 10–15% against the short on any AI-foundry headline.
2. Trump-administration foundry-policy or CHIPS-Act headlines are recurring catalyst risk in this name and unhedgeable here.
3. The May 11 high was on huge volume; that level may be defended a second time before the trade works. Position sizing must accept >1R adverse move.

---

Idea #5 — SHORT 0981.HK (SMIC)

• Catalyst: Technical breakdown post-vertical move; PBoC LPR fix June 20 + the ongoing rotation out of HK foundries. Next interim results August (outside window).
• Setup: SMIC ran from HK$54 (early April) to a HK$93 high on May 26 — a +72% advance in 7 weeks. May 26 printed an outside-down day (open HK$91.80, high HK$93.00, low HK$84.10, close HK$84.40). Three days of distribution since (HK$84.40 → HK$85.20 → HK$88.25 → HK$81.60). Friday's close HK$81.60 is below the May 26 low (HK$84.10) on 215M shares.
• Thesis:
• Parabolic exhaustion with a confirmed break of the prior swing low + 3x-avg volume = textbook reversal setup.
• HK foundries collectively rolled — Hua Hong (1347.HK) printed −5.1% Friday on similar volume profile; this is a sector move, not a single-name event.
• Northbound flow is a blind spot here (see #4 in declared blind spots), so the thesis is technically driven, not flow-driven.
• Entry zone: HK$83–87 (mid HK$85). Sell into any small bounce / back-test of the broken support.
• Stop: HK$91 (+7.1%) — above the 5/27 intraday high (HK$89.45) with buffer; close below the May 26 high (HK$93.00).
• Target: HK$68 (−16.7%) — back to the early-May / mid-May pre-parabolic level (HK$68–72) and the 50-DMA cluster around HK$73–75 with buffer below.
• R/R: (HK$85 − HK$68) / (HK$91 − HK$85) = 17 / 6 = 2.83
• Sizing: ~6,500 shares (~HK$420k = ~US$54k notional) per US$5,000 risk; verify HKD/USD before placing.
• Time stop: Cover on a daily close above HK$91 or by July 17, whichever first.
• Risks:

1. Stock Connect Northbound flow is the dominant marginal buyer for SMIC and is not visible here — a sudden Northbound surge can squeeze 5–10% intraday.
2. China-policy headlines (semiconductor self-sufficiency, fab subsidies) are recurring positive catalyst risk for SMIC specifically.
3. HK borrow availability and rates are not checked — confirm with the prime broker that 0981.HK is borrowable at a workable rate before sizing the short.

---

6. Portfolio Construction & Risk Notes

Net / gross exposure (assumed $1M book, 0.5% NAV risk per idea = $5k 1R)

Modest gross. The book leaves room to add on conviction or to expand position sizing if the early prints (AVGO June 3) confirm the thesis.

Correlation budget — read this before sizing

• AVGO + MU + ANET are essentially one bet: US AI capex. A bad AVGO print on June 3 will likely hit MU and ANET in sympathy. Treat the book as carrying ~2.0R of correlated AI-capex exposure on the long side, not 3R.
• INTC short partially hedges this — Intel sells off if AVGO sells off (semis-wide weakness), so the natural hedge ratio is ~0.5x. But INTC also has unique foundry-policy risk that doesn't correlate.
• SMIC short hedges the China-tech leg of any global semi rollover, but is dominated by its own Northbound flow regime.

Beta estimate (rough, from 90-day weekly returns)

• AVGO ~1.6 to SPY, MU ~2.0, ANET ~1.5, INTC ~1.5, SMIC ~1.0 to HSI.
• Book beta to SPY ≈ +0.6 (longs minus shorts, weighted by notional). Net long but not aggressively so. SPY drawdowns are not the dominant risk; idiosyncratic event prints are.

Macro events that can break the book regardless of single-name thesis

1. Jun 3 AVGO print — sets the tone for the entire AI capex complex. A miss takes AVGO + MU + ANET down together.
2. Jun 9–10 FOMC — a hawkish surprise hits AI capex (high duration) and helps the short book; a dovish surprise does the opposite.
3. Jun 20 China LPR / any China stimulus surprise — risk to the SMIC short specifically.
4. Any export-control headline out of the US Commerce Department — bilateral risk for both INTC long-term thesis and SMIC short thesis.

Sizing discipline

• 1R = 0.5% of book per idea = 5 × 0.5% = 2.5% NAV at risk if all stops hit simultaneously.
• Time-stop discipline: if any idea is not at +0.5R by the 3-week mark, cut size in half. The catalyst-anchored framing means the catalyst either works or it doesn't — there is no "give it another month" in this book.

---

7. Watch List — Screened, Not Taken

---

8. Caveats — Read Before Sizing

• Numbers and prices are end-of-day May 29, 2026. They will be stale on Monday open.
• No options data anywhere in this report. Earnings-anchored ideas are directional only.
• DELL printed a `quote.previous_close` value that disagreed with daily history by 30+ percentage points. That's a yfinance / Yahoo data layer issue and is exactly the failure mode the skill warns about. The Day% values in this report are recomputed from daily history — do not rely on the live `change_pct` field for any of these names.
• Earnings dates: only AVGO is confirmed. MU late-June and CRDO Q4 FY26 are estimated and unconfirmed. Treat the time stops accordingly.
• HK shortable list verified for 0981.HK via the May 29, 2026 HKEX revision. Re-verify on the day of execution — the list is updated at each quarterly review.

---

End of report. Generated May 30, 2026.