New today“Cloud up 82%, cloud operating profit tripled — and it's beaten down to RSI 31 for spending an extra $15B. The market is punishing everyone who spends, but it just threw out the one making the most money doing it. When others panic, at least look at the price tag”
Technicals — The -7.1% close at $317.69 caps a -14.8% week, slipping 1.8% below the 200-day ($323.5); RSI 31 is oversold and the 2.6x volume reads as panic turnover. The March shelf at $305-310 is support below; reclaiming $323.5 is the right-side signal, with the 50-day at $366 the first target
Fundamentals — Q2 revenue $119.8B (+24%), Cloud $24.8B (+82%) with cloud operating income near-tripling to $8.8B, Search +17%, YouTube +13% — every engine firing. The only reason it's being sold: FY capex raised to $195-205B with Q2 alone at $44.9B (doubled YoY), on top of June's $80B equity raise — the market hasn't finished doing the FCF math
News — The 7/22 after-hours print beat across the board, but another $15B capex raise re-priced the 'bottomless AI spend' fear, and the same-night double act with Tesla sank the Nasdaq. The early-July final CJEU ruling upholding the $4.67B EU Android fine still lingers, alongside DMA compliance pressure. JPMorgan called the sell-off a buying opportunity outright; Barclays raised its target to $425 against the tape
Short-term · 2-4 weeks
LongLean bullish
A panic flush with fundamentals intact: RSI 31, the $305 March shelf, and +36% consensus upside make the odds acceptable. The risk is capex fear bleeding into the 7/29 MSFT/META earnings night — hence scaling in rather than full size, stop below the shelf
Entry Scale in $305-320Stop $294 (below the March shelf)Target $360-366 (50-day)
Long-term · months+
Accumulate
Cloud growing 82% with fast-rising margins is proof the AI capex is already monetizing, and Search's moat is uneroded; sub-200-day prices handed out during capex panics have historically been long-term entries
Further capex hikes pressuring FCF
Ongoing EU fines and DMA compliance drag
Gemini monetization falling short
Community Voices
WSB · Rank 7 but 2,253 upvotes leads the board — 'the quarter actually beat' is the top-voted refrain: retail is buying the dip and cheering each other onOriginal ↗
Institutional Views
华尔街共识 · 72 analysts average $433.34, +36% vs. spot, rating 1.17 Strong Buy
JPMorgan · Cut $460 to $420 (Overweight): sell-off overdone, a buying opportunitySource ↗
Barclays · Raised against the tape, $405 to $425 (Overweight)Source ↗
Morgan Stanley · Trimmed $415 to $400, cautious but still OverweightSource ↗
-11.63%07-07“HBM is sold out into 2028 and UBS just tripled its target to $1,625 — for once the WSB heat (rank 2) points the same way as the fundamentals. The most expensive mistake in a trend is getting off too early: hold, and let others guess the top”
Score8/10UBS target $535 to $1,625 in one moveHBM capacity 2026 sold out; $22B non-cancelable contractsDRAM pricing Samsung eyeing up to +20% in Q3Week +14.2%
Technicals — The +3.2% close at $990.21 caps a +14.2% week, reclaiming the 50-day ($952) and holding 4% above it; RSI 52.5 has just crossed midline with room to run. Resistance at $1,090 (early-July high) opens $1,200 beyond; support stacks at $952 and $920. An $80 ATR demands position sizing discipline
Fundamentals — The June-quarter narrative keeps compounding: memory scarcity pushed gross margin to 84.6%, 2026 HBM capacity is fully sold out, and $22B of non-cancelable contracts floor the revenue line. UBS models Q3 DRAM contract pricing +32% and Q4 +18%, with undersupply lasting into mid-2028 — this isn't a cyclical bounce being priced, it's structural shortage
News — UBS (Timothy Arcuri) set the re-rating anchor in late May, lifting its target from $535 to $1,625; this week's double catalyst: Morgan Stanley and BofA both called to buy the dip on 7/22, and Samsung is pushing up to +20% more on Q3 DRAM contracts (roughly +250% cumulative this year). The latest monthly global NAND sales hit a record, +40.7% MoM — the whole memory complex is being re-priced
Short-term · 2-4 weeks
LongBullish
A +14.2% week reclaiming the 50-day with RSI only at 52 — the move is driven by pricing and target hikes, not pure sentiment. The $880 stop gives a full ATR of cushion; losing it would mean the pricing narrative itself is breaking
Entry Hold; add on a $920-955 retestStop $880 (7.5% below the 50-day, one ATR of cushion)Target $1,090 then $1,250
Long-term · months+
Accumulate
Undersupply into mid-2028, 84.6% gross margin and non-cancelable contracts turn memory from a cyclical into an AI-infrastructure tollbooth; with +58% consensus upside, pullbacks are entries
Samsung/Hynix capacity adds breaking the shortage
An AI capex downturn hitting HBM orders
An $80 ATR that shakes out oversized positions
Signal BacktestCumulative -11.63%price itself -4.08%0/1 closed trades wonprofitable since 07-10
07-07Long$984.75 → closed $853.207-17-13.36%
07-22Long$970.82 → open → $990.21+2.00%
Community Voices
WSB · 395 mentions holding rank 2 with 1,371 upvotes — memory is that rare trade where retail and institutions agree; the heat is high but for once not a contrarian signalOriginal ↗
Institutional Views
华尔街共识 · 54 analysts average $1,568.74, +58% vs. spot, rating 1.16 Strong Buy
UBS · Target $535 to $1,625: AI data-center HBM demand exploding, 2026 capacity sold outSource ↗
Morgan Stanley / BofA · Both pushing 'buy the dip,' citing $22B non-cancelable contracts and structural demandSource ↗
New today“A $289B backlog — the toll booth of the war economy, with Patriot, Standard Missile and AMRAAM all ramping. One step from a new high, a beat-and-raise, and upgrades landing in threes: buying at a breakout isn't chasing, it's following the order flow”
Score7/10Q2 EPS $1.89, a 14% beatBacklog $289B recordFY guide EPS raised to $7.10-7.25Off 52w high -2.5%
Technicals — The +7.3% close at $209.16 on 2.5x volume sits 2.5% from the $214.5 high, confirmed by both the hot and breakout channels; RSI 71.5 is warm, but trend-style warm. The 20-day at $197 is the retest level with a $5.5 ATR; past the high there's no overhead supply, measured move $228
Fundamentals — Q2 adjusted EPS of $1.89 (14% beat) on revenue of $24.7B, $1.9B above consensus; FY EPS guidance raised from $6.70-6.90 to $7.10-7.25 with revenue to $95-96B. The Raytheon segment grew 18% — air defense and missiles are the scarcest items in global arms trade — while the $170B commercial backlog rides the engine aftermarket cycle
News — The 7/23 pre-market beat-and-raise, alongside LMT (+10.5% on $65B of new orders including a $35B THAAD award), ignited the defense complex. With tankers hit in the Red Sea and the US-Iran standoff grinding on, defense is the one macro-tailwind sector in this tape; Jefferies, which upgraded to Buy in early June ($210 to $220), reiterated after the print
Short-term · 2-4 weeks
LongLean bullish
Beat-and-raise, record backlog, 2.5% from the high, dual-channel confirmation — a textbook breakout structure. RSI 71 is the only blemish, solved by adding on the retest rather than chasing
Entry Enter $205-211, add on a $197-200 retest (20-day)Stop $196 (below the 20-day, 2x ATR)Target $228 (measured move)
Long-term · months+
Accumulate
A $289B backlog means 3+ years of revenue visibility, and air-defense restocking is a decade-long theme; the commercial-plus-defense twin engine is scarce certainty in a high-rate tape
A fast-moving peace process inverts the thesis
Legacy engine-quality liabilities
Consensus only +5% — near-term odds depend on target resets
Community Voices
StockTwits · Defense names never trend on retail boards — which is exactly the point: nobody front-runs it, so the trend moves slow and holdsOriginal ↗
Institutional Views
华尔街共识 · 27 analysts average $220.52 (+5.4%), rating 1.35 Buy — targets should reset higher after the beat-and-raise
Jefferies · Upgraded to Buy, target $210 to $220Source ↗
RBC Capital · Reiterated Outperform, $230 targetSource ↗
+3.49%07-02“One third-party SensorTower print erased HK$300B of market cap, and Huatai calls it 'sentiment mispricing' — perhaps, but the real soft spot is the AI race: Yuanbao's 57M MAU versus Doubao's 345M. Hold what you own and reconcile against the 8/12 earnings; if the books don't balance then, revisit”
Score7/10Wednesday (7/22) -7.05% to HK$440.6, a near 6-month low, on heavy HK$29.8B turnoverTrigger Bernstein/SensorTower: Q2 iOS game GMV -2.6%Southbound 5 straight days net selling, ~HK$6.18B totalQ2 earnings 8/12
Technicals — Thursday's +1.0% close at HK$445.2 stabilized, but the -6.9% week and Wednesday's (7/22) high-volume marubozu (HK$29.8B) damaged the structure; the stock sits -18.5% under the 200-day (HK$546) with RSI at 46.6. Support: the HK$420-430 2025 shelf. Resistance: HK$470 (20-day) and the HK$474 gap
Fundamentals — Q1 non-IFRS net profit of RMB 67.9B (+11%; IFRS RMB 58.1B, +21%) keeps diverging from the tape. The market's two fears: the gaming base (SensorTower's Q2 iOS GMV -2.6%, which Huatai calls too narrow a lens and a sentiment mispricing), and 2026 AI R&D plus compute spend doubling past RMB 36B into the profit line. Down 31% from the October 2025 high and squeezed out of mutual funds' top-10 holdings — the positioning has cleared faster than the fundamentals have deteriorated
News — Wednesday's (7/22) -7.05% was the week's main event: a Bernstein note citing SensorTower data lit the gaming fear, with southbound money net-selling about HK$2.77B of Tencent that same day. Q2 earnings land 8/12, with revenue expected at RMB 201.6-211.1B (+9.3-14.4%); the focus is whether the game-revenue scare survives contact with actual accounts — the third-party-data-versus-reality bet settles next month
Short-term · 2-4 weeks
LongBullish
Panics built on third-party data usually retrace half before the earnings verdict; the HK$420 shelf plus the +11% profit divergence underwrites the hold, with a clean stop below the base
Entry Hold; add at HK$425-440Stop HK$418 (below the 2025 shelf)Target HK$490-500 (above the gap)
Long-term · months+
Accumulate
Non-IFRS profit +11%, buybacks merely paused for the pre-earnings blackout (full-year program intact), all three engines (games, ads, fintech) intact, and the valuation back at early-2025 levels; the AI spend is a short-term tax on profit for a long-term claim on the ecosystem — RMB 36B can't buy Yuanbao a win, but it can afford the attempts
The 8/12 print confirming the gaming slowdown
Yuanbao's MAU gap versus Doubao widening
Southbound and fund de-positioning not yet done
Signal BacktestCumulative +3.49%price itself +3.25%profitable since 07-04
07-02LongHK$430.2 → open → HK$445.2+3.49%
Community Voices
港股通 · Southbound sold HK$6.18B over five days — but with the funds' crowded position already flushed, the stock is lighter, not heavier. Are you the herd chasing the flow, or the hunter waiting for the statement dateOriginal ↗
Institutional Views
华尔街共识 · 53 analysts average HK$689.55, +55% vs. spot, rating 1.14 Strong Buy
华泰证券 · SensorTower's lens is too narrow (no Android/PC/overseas) — the sell-off is sentiment mispricingSource ↗
Bernstein · Citing SensorTower: domestic iOS mobile-game GMV around -2.6% YoY in Q2, pressuring the gaming baseSource ↗
+13.22%07-02“Nobody cares that H1 total volume fell 16% — overseas +95% is the re-rating key: eight ro-ro ships, plants in Hungary, Brazil and Thailand. The price war bleeds at home while the profit grows abroad. Up 20% already in July; if you're on, don't get shaken off”
Score7/10June sales 403K units, 2026 monthly highOverseas 175K in June (+95%); >40% of H1 mixUBS est. RMB 8,728/unit Q2 profit, +50% QoQJuly to date +20~25%
Technicals — The +0.9% close at HK$88.65 caps a +2.4% week and a 20-25% run off the 6/30 low, leading Hang Seng Tech; above the 50-day (HK$87) but still -8.7% under the 200-day (HK$97). RSI 55.6 is healthy; resistance at HK$92 (March gap) then HK$97, support at HK$84-85
Fundamentals — June's 403K units (+5.5%) set a 2026 high, though H1's 1.81M ran -15.7% YoY — the price war's toll at home. The real delta is profit mix: UBS models per-unit profit jumping from RMB 5,831 to 8,728 in Q2 (overseas pricing plus flash-charging/premium mix), with H1 overseas volume near 790K (+68%), now over 40% of the total
News — No fresh single-day news — July's run keeps being driven by the record-volume, overseas-ramp, margin-recovery triad; Hungary/Brazil/Thailand localization is ramping and the eighth ro-ro ship is in service. With no interim pre-announcement yet, UBS's per-unit profit estimate is the bulls' key exhibit — the late-August interim report is the verdict
Short-term · 2-4 weeks
LongBullish
The +50% QoQ per-unit profit surprise isn't fully priced, and the trend structure above the 50-day is intact; with the interim report not due until late August, pullbacks before then are position-management questions, not directional ones
Entry Hold; add on a HK$84-86 retestStop HK$82 (below the 50-day and July trendline)Target HK$97-100 (200-day)
Long-term · months+
Accumulate
Going global is the structural pivot from volume-at-a-price to price-for-quality: overseas unit economics far exceed domestic, and a 40% mix is only halftime; the worst of the domestic price war is likely behind
EU tariffs and localization execution
A second leg of the domestic price war
The interim report falling short of UBS's estimate
Signal BacktestCumulative +13.22%price itself +13.22%profitable since 07-04
07-02LongHK$78.3 → open → HK$88.65+13.22%
Community Voices
港股通 · Retail is still counting price-war wounds at home while institutions model overseas per-unit profit — same financials, two readings, and the tape sides with the latterOriginal ↗
Institutional Views
华尔街共识 · 26 analysts average HK$123.99, +40% vs. spot, rating 1.25 Strong Buy
瑞银 · Q2 per-unit profit seen +50% QoQ to RMB 8,728 on overseas pricing and premium mixSource ↗
大华继显 · Target raised to HK$135, sector top pickSource ↗
Position: +1.6% into fresh highs — Goldman lifts to HK$181, modeling a +25% Q2 profit beat and a $1.5B buyback at the 8/4 print; RSI 75 is hot but the position stays
Position & watchlist: +1.1% — the cloud-up-45% preview already paid once on 7/8, flash-buy losses narrowing faster than expected and Jiang Fan pledging UE breakeven this fiscal year; hold into late-August earnings
Position: down 10.5% on the week in GOOGL's capex contagion, with the 7/29 print as the exam — the capex language sets the direction; hold without adding, trim below $580
Position: earnings land 7/29, same day as the FOMC, with GOOGL's capex panic as the cautionary tale — hold the $381 position without adding; +46% consensus upside is the backstop
Position: preliminary Q4 gross-margin guide of 15-17% (nearly doubled), $60B+ of new orders, and the SpaceXAI gigawatt data-center build — up 17.6% on the week, hold; the convert dilution is the buried risk
Position: NAND pricing seen +234% in 2026, Susquehanna at $3,250, July global NAND sales at a record — in the year's best stock, pullbacks are entries, hold
Position: +4.4% turning the week positive, +20.1% on the month — rivals pivoting to loss-cutting in the flash-buy war is an industry-wide breather, hold
Position: NVIDIA's 9.3% stake and $2B investment, the $1B+ Reflection AI contract, and a $775M GPU-collateralized loan defusing dilution fears — three catalysts behind a +16% week, hold
