+6.02%07-02“The ad company announced it's selling compute, and overnight the market refiled it from 'cash burner buying GPUs' to 'landlord selling water.' A narrative changing drawers is worth more than earnings changing gears — holders, raise the stop and enjoy the re-rating”
Score9/10Day / week +6.0% / +10.1%New business Meta Compute: selling surplus AI compute and modelsIn-house silicon Iris chip ships September, ~14GW targeted next yearYTD Flipped positive
Technicals — Up 6.0% on 1.9x volume to $669.21, +10.1% for the week, clearing both the 50- and 200-day in two sessions and flipping YTD positive. RSI 66.3 is warm, not extreme; the $700 round number and $740 prior-high zone are staged targets, with gap support at $640.
Fundamentals — Meta Compute reprices 'excess capex' as 'cloud revenue optionality': Wolfe Research models ~20% EPS upside per gigawatt of cloud capacity. The Iris chip ships in September targeting ~14GW next year, and the Muse Spark/Image model suite lands across the ad stack — for the first time, ads and compute form a twin-engine narrative.
News — Shares rose 6% on announcement day, turning positive for the year, with WSB mentions exploding a second day (57→240→sustained). Wolfe backed the new story with an $800 target, and the market began re-marking its capex through the AVGO/ORCL compute-valuation lens.
Short-term · 1–3 weeks
LongBullish
Holding: narrative upgrade, dual-average breakout and the YTD flip give triple confirmation. Stop raised from $590 to $620 (under the gap's lower rim) to lock the re-rating gain; a gap-fill below $620 before late-July earnings would mean the market isn't paying for the new story.
Entry Hold; a retest of $640 (the gap's rim) is the add-watch zoneStop $620Target $720
Long-term · months+
Accumulate
Ad cash cow, Meta Compute optionality and Iris cost-down run in parallel — a rare megacap whose new story just opened to page one; if the 14GW target lands, cloud becomes the second revenue curve.
Unproven margins and utilization on external compute sales
Above-plan capex squeezing free cash flow
Iris production-yield risk
Signal BacktestCumulative +6.02%price itself +9.19%0/1 closed trades wonprofitable since 07-10
07-02Long$612.91 → closed $582.907-04-4.90%
07-07Long$600.29 → open → $669.21+11.48%
Community Voices
WSB · Last week they mocked 'Zuck burning cash again'; this week they're computing EPS per gigawatt — retail learns new valuation frameworks at a speed exactly proportional to the share priceOriginal ↗
Institutional Views
华尔街共识 · 68 analysts, average target $823 (+23% vs. spot), 1.13 — strong-buy territory
Wolfe Research · ~20% EPS upside per GW of cloud capacity, $800 target — capex flips from cost line to option bookSource ↗
+8.28%07-04“Two days ago it was ceding the mic to MU and SKHY; then META said '14 gigawatts' and the mic came right back — in the compute arms race it's always the arms dealer. Above the 50-day, the repair trade shifts from defense to offense”
Technicals — Up 4.0% to $210.96, reclaiming the 50-day ($209.20) on volume — a 7% week completing the two-stage repair from 200-day defense to 50-day recovery. RSI 57 is healthy; the $220 May gap is next resistance, beyond which the prior-high retrace opens.
Fundamentals — META's Meta Compute reveal is a stealth tailwind: within the 14GW target, in-house Iris is supplementary — GPUs remain the bulk buy. The bear logic of 'customer silicon = peak demand' just got contradicted by 'customers expanding both capacity and purchases.' With H200 China access on top, both demand-side fears dulled within a week.
News — A 7% week led the trillion-dollar club as WSB rank recovered. Memory stole the SKHY-week spotlight, but Friday's flows reaffirmed: in a compute expansion cycle, GPUs are still the biggest line item.
Short-term · 1–3 weeks
LongBullish
Holding: the 50-day reclaim completes repair stage two. Stop raised from $188 to $196 (half an ATR under the 20-day); target extends to $235 beyond the $220 gap-fill.
Entry Hold; a $205 retest (above the 50-day) is the reboarding watch levelStop $196Target $235
Long-term · months+
Accumulate
The customer buildout wave (META's 14GW, MARA's 4.8GW) proves demand explosion, not substitution — the arms-dealer logic strengthens, and 21x forward remains a three-year low.
In-house silicon eroding premium share long term
H200 policy reversals
Technical supply at the $220 gap zone
Signal BacktestCumulative +8.28%price itself +6.77%profitable since 07-07
07-04Long$194.83 → open → $210.96+8.28%
Community Voices
WSB · 'MU is the shovel, NVDA is the excavator' — after a full rotation retail realized the mine needs every tool; the scarcest asset in their accounts is still patienceOriginal ↗
Institutional Views
华尔街共识 · 66 analysts, average target $313 (+49% vs. spot), 1.13 — strong-buy territory
-0.55%07-07“Its rival listed in glory and it dipped 1.2% — classic catalyst-day boat-switching, not abandonment. SKHY closing 13% over its offer price is the market notarizing the memory cycle; exit talk starts below $880, not before”
Score7/10RSI(14) 49.1Day move -1.2%SKHY debut +13% close at $168, opened higher and fadedStop $880 (under the 50-day at $899), unchanged a fourth day
Technicals — Down 1.2% to $979.30 in a low-volume consolidation (0.6x rvol) as SKHY's debut siphoned flows — a third coiling attempt below $1,000. The 50-day ($899) trend support is unthreatened, RSI 49 neutral, and the pattern remains a high flag.
Fundamentals — SKHY's +13% debut close above offer confirms the memory complex's public-market pricing; the 5.5x-versus-6.66x forward gap showed no convergence pressure — day-one buyers chose to lift the whole sector's anchor instead. No new fundamental variables; the HBM shortage narrative is intact.
News — SK Hynix's chairman told CNBC 'demand is enormous' — the rival's CEO just delivered the best roadshow MU bulls could ask for. WSB mentions held at #1.
Short-term · 1–3 weeks
LongBullish
Holding: SKHY's smooth landing removed the week's biggest unknown, and the low-volume coil is a bullish shape. The $880 hard stop stands a fourth day; no adds before a $1,000 break.
Entry Hold; raise the stop on a volume break of $1,000, with a $920 retest as the add-watchStop $880Target $1150
Long-term · months+
Accumulate
HBM sold out through 2027, the Anthropic pact, and SKHY-driven sector transparency — the memory supercycle's three-legged base is complete.
A later SKHY break below offer would bite the sector back
Supply response at the cycle top
Momentum decay after three failed runs at $1,000
Signal BacktestCumulative -0.55%price itself -5.13%profitable since 07-10
07-07Long$984.75 → open → $979.3-0.55%
Community Voices
WSB · 'SKHY up means memory bull market, SKHY down means MU scarcity' — bulls have scripted both outcomes for themselves; at times like this the stop is worth more than the opinionOriginal ↗
Institutional Views
华尔街共识 · 53 analysts, average target $1,576 (+61% vs. spot), 1.14 — near strong buy
CNBC · SK Hynix +13% on debut; chairman: AI memory 'demand is enormous'Source ↗
+6.97%07-02“Collateral damage in the unlock wave — it has no unlock of its own yet took two days of hits, with the HK$450 stop just two points underfoot. The poise earned by eight green candles gets spent Monday answering one question: is the trend still here”
Technicals — Down 2.0% to HK$460.20, -2.8% over two days, giving back a third of the eight-day run. The 50-day (HK$451.30) and the HK$450 stop nearly coincide — technical support and the discipline line stacked at one level; Monday's open is the answer.
Fundamentals — No company-level negatives — the decline is unlock-wave beta. The Hunyuan, games and Video Accounts ad narratives are unchanged and the southbound allocation case undamaged. Passive drawdowns like this are precisely why the stop exists: no forecasting, only responding.
News — It lagged with -2.0% on a day the tech index fell just 0.21%, as money rotated from big tech toward aerospace and biotech. Sector beta isn't the stock's fault — but position management reads price, not excuses.
Short-term · 1–2 weeks
LongLean bullish
Holding (~+7%): sector-beta drawdowns don't justify a discretionary exit, and the stop won't be lowered because 'it's unfair' — discipline earns its keep exactly when tested. With the 50-day and the stop coinciding, a break is a double signal.
Entry Hold; the HK$450 stop does not move — a break means out, a hold means the trend survivesStop HK$450Target HK$530
Long-term · months+
Accumulate
Its ballast status, AI application ecosystem and buyback engine are unchanged; the unlock wave is other companies' supply problem, not its demand problem.
Sustained beta drag from the sector's unlock wave
Game approvals and regulatory cadence
Macro sensitivity of ads
Signal BacktestCumulative +6.97%price itself +6.73%profitable since 07-04
07-02LongHK$430.2 → open → HK$460.2+6.97%
Community Voices
雪球 · 'Why does it fall with no unlock?' — because selling the flagship is the fastest way institutions cut tech exposure; deep liquidity is both a blessing and a taxOriginal ↗
Institutional Views
华尔街共识 · 54 analysts, average target HK$690.80 (+50% vs. spot), 1.14 — near strong buy
+10.17%07-02“The unlock storm blasted a 20-point crater in semis and it kept climbing — day four's independence is worth more than the first three days' gains. A stock like this is never endangered by others falling, only by itself rising too fast; fortunately the stop is already up”
Score8/10RSI(14) 55.3Day / week +2.0% / +17.0%Relative strength Closed green through the unlock double-kill — a second day of sector independenceStop HK$105 (raised yesterday)
Technicals — Up 2.0% to HK$110.20, green through both unlock-drag sessions, +17% for the week. RSI 55 is healthy, the 50-day (HK$117.50) remains first resistance, and the 20-day (HK$101.20) plus the HK$105 stop form a two-layer cushion.
Fundamentals — The unlock storm revealed the re-rating's flow base: money leaving tech didn't leave Hong Kong — it concentrated into the highest-certainty AI asset. Earnings previews, the H200 list and bank endorsements make it both the shelter and the spearhead.
News — Consecutive independent gains through a sector rout are the classic footprint of institutional accumulation. The next catalyst is the earnings window at month-end; the 50-day (HK$117.50) break decides the medium-term trend.
Short-term · 1–3 weeks
LongBullish
Holding (~+7%): relative strength through a sector storm is the highest-grade reason to stay. Stop holds at HK$105, rising to HK$110 on a 50-day break.
Entry Hold; a HK$105–107 retest remains the only dignified boarding zone for latecomersStop HK$105Target HK$130
Long-term · months+
Accumulate
Cloud-AI second curve, e-commerce profit repair and flow-concentration effects — the unlock storm strengthened, not weakened, its status as the China AI re-rating's core vehicle.
Technical give-back after a 17% week
H200 policy reversals
A soft earnings print interrupting the re-rating
Signal BacktestCumulative +10.17%price itself +17.11%0/1 closed trades wonprofitable since 07-09
07-02LongHK$94.5 → closed HK$94.107-04-0.42%
07-08LongHK$99.6 → open → HK$110.2+10.64%
Community Voices
雪球 · 'Others unlock, it rallies — that's what a leading theme looks like.' Retail finally learned relative-strength selection; unfortunately most learned it on day fourOriginal ↗
Institutional Views
华尔街共识 · 34 analysts, average target HK$185 (+68% vs. spot), 1.16 — strong-buy territory
Down 8.8% in a catch-down that ate most of the gains, through the 20-day with the 50-day ($31.60) still below; stop tightened to $30.50 — the gene-editing M&A dream doesn't merit an undefended wait
