Score9/10RSI(14) 48.5Off 52w high -22.3%P/E (TTM) 22.1Revenue YoY +167%
Technicals — Broke the $1,000 handle to close at $976, down 20.9% on the week; the 20-day ($1,043) now caps rallies and the next real support is the 50-day at $852. RSI at 48.5 is merely neutral — at this volatility, no oversold reading means no bottom yet. Daily ATR is a wild $97 (~10%); stress-test any position for 10% daily swings.
Fundamentals — Fundamentals are intact: $41.5B revenue last quarter (4x YoY), management guiding memory tightness through 2028, and just 22x earnings against +167% revenue growth. What changed is the market pricing 'cycle peak' — Samsung and SK Hynix announced supply additions, and the 'AI capex peaks in 2026' thesis is gaining the upper hand.
News — Institutional profit-taking after an 80% first-half run is the main driver — no single fundamental break. Piling on: a California class action alleging Samsung, SK Hynix and Micron colluded to restrict DRAM supply (prices up 700% in four years). The selling spread to Korea, with Samsung and Hynix down 7–9% in a day.
Short-term · 1–3 weeks
NeutralSidelines
Momentum is broken and the 20-day caps at $1,043. Wait for volume dry-up plus a reclaim of the 5-day before entering; a straight slide toward the 50-day ($852) means systematic exit from the memory trade — don't catch that knife.
Entry Scale in at $900–950 after volume dries upStop $850Target $1150
Long-term · months+
Accumulate
The AI memory supply gap hasn't gone anywhere: HBM capacity is locked under long-term agreements and management guides tightness into 2028. 22x earnings against triple-digit growth makes pullbacks a gift for the long book — provided you scale in and keep dry powder.
Samsung/Hynix supply lands and memory pricing peaks early
The DRAM antitrust suit turns into real fines or pricing constraints
If AI capex truly peaks in 2026, cyclicals get de-rated hard
Community Voices
WSB · Still #1 on the board, but mentions collapsed from 1,179 to 227 in 24h — retail gunfire stopped before the decline didOriginal ↗
StockTwits · Dip-buyers posting fills, top-sellers posting gains — each side convinced the other is the bagholderOriginal ↗
Institutional Views
华尔街共识 · 52 analysts, average target $1,574 (+61% vs. spot), rating 1.13 — near strong buy
加州集体诉讼 · Alleges Samsung/Hynix/Micron colluded to restrict DRAM supply and inflate prices 700% over four years — a legal overhang on the whole groupSource ↗
Score8/10RSI(14) 46.8Off 52w high -25.9%P/E (TTM) 60.7This week -22.0%
Technicals — Down 14.1% on the day to $1,745 and 22% on the week, breaking the 20-day ($1,952) with the 50-day at $1,610 below. Daily ATR of $211 (~12%) makes it the most volatile mega-cap on the board. The 52-week range is $40–2,354 — on this chart 'support' is a psychological concept, not a technical one.
Fundamentals — The NAND shortage thesis holds: revenue +83% YoY, and BofA argues the supply-demand imbalance and firm pricing persist through 2027, lifting its target from $2,100 to $2,500. But at 60.7x earnings the 'shortage' is priced well past 2027 — the margin for error is razor thin.
News — Led the memory decline two days running (-10%, then -14%), on the same triggers as Micron: profit-taking, Samsung/Hynix supply-add chatter, and 'AI storage glut ahead' notes. BofA raised its target to $2,500 into the selloff — a rare head-on bull-bear collision.
Short-term · 1–3 weeks
NeutralSidelines
A 12% daily ATR means any stop can get swept by noise — size it like a wager, not an investment; chips no bigger than dinner money. Lose the 50-day ($1,610) and the cycle story goes into triage.
Entry Small size at $1,600–1,700 (upper band of the 50-day) after stabilizationStop $1500Target $2000
Long-term · months+
Neutral
The purest NAND torque on the board, but a single product line at 61x earnings after a 38x one-year run leaves long-term odds against the buyer. For memory supercycle exposure, Micron's HBM story offers better value per unit of risk.
If NAND pricing peaks before 2027, it's a double de-rate
Samsung/Hynix/Kioxia add supply faster than expected
The volatility itself is a risk: a 12% daily ATR amplifies every mistake
Community Voices
WSB · #2 on the board with mentions down from 262 to 71: half shouting 'NAND tight through 2027,' half posting -22% loss pornOriginal ↗
Institutional Views
华尔街共识 · 29 analysts, average target $2,174 (+25% vs. spot), rating 1.33 — buy-to-strong-buy
美银 · Raised its target to $2,500 (from $2,100) into the selloff, keeping a Buy: NAND imbalance and firm pricing persist through at least 2027Source ↗
Technicals — At $140.27, just 4% above the $134.57 52-week low after a 42% monthly collapse. RSI 26.7 with RSI7 at 12 — deeply oversold but with zero stabilization signals, and the 20-day is way up at $178. ADX at 30 says the downtrend is still accelerating; every 'cheap' print can get cheaper.
Fundamentals — Bull and bear cases share one source: RPO exploded $85B in a single quarter to $638B — contracted revenue, not a pipeline. The cost: FY26 free cash flow of -$23.7B, total debt of $124B (vs. $89B a year ago), and FY27 plans for $70B capex funded by another $40B raise including a $20B ATM. At 24x earnings you're buying a leveraged AI infrastructure bet.
News — Down nearly 9% more this week: chatter of OpenAI delaying its IPO hit the credit story of Oracle's biggest customer, promoting 'concentration risk' from research footnote to headline. Bulls are surfacing too — 24/7 Wall St calls it a 'golden buying opportunity,' arguing the contracts are signed and the market simply refuses to value a funding story.
Short-term · 1–3 weeks
NeutralSidelines
The $134.57 52-week low is the line everyone is watching: hold it and the oversold bounce targets the $170 gap zone; break it and systematic stops amplify the stampede. Headline risk (OpenAI IPO, financing announcements) can blow through any technical level — size for a 10% overnight gap.
Entry Light size only after $134 holds — shrinking volume and no new lowsStop $130Target $170
Long-term · months+
Neutral
If it works, it's a triple play: $638B of backlog converting to revenue, FCF inflecting positive, and a re-rate. If it doesn't, the bleeding continues — fronting datacenter capex for customers who aren't yet profitable themselves. This is an option, not a bond: worth a small position, not a conviction bet.
OpenAI concentration: the biggest customer's ability to fund itself dictates Oracle's revenue recognition
$124B of debt in a high-rate world; refinancing costs eat earnings
The $20B ATM overhang makes dilution a standing pressure
Community Voices
StockTwits · 'The telecom bubble of AI' vs. 'a mispriced cash cow' — the shouting usually bottoms before the stock doesOriginal ↗
Institutional Views
华尔街共识 · 45 analysts, average target $254.8 (+82% vs. spot), rating 1.26 near strong buy — a rare chasm between consensus and tape
24/7 Wall St · Calls the crash a 'golden opportunity': the $638B RPO is inked contracts; the market is pricing funding panic, not the businessSource ↗
Score8/10RSI(14) 46.8Off 52w high -21.1%P/E (TTM) 359Q2 deliveries YoY +25%
Technicals — Down 7.5% to $393.45, swallowing most of the 12% early-week run, with $29.1B traded — the most active name on the board. Price fell back below all three moving averages ($399/$406/$419), which now form a resistance band. The stock has dropped on each of the last three delivery days — pattern traders literally short the print now.
Fundamentals — Q2 deliveries of 480,126 (+25% YoY) crushed the 406K consensus and even Goldman's bullish 420K; two straight quarters of growth officially end the two-year decline. Deliveries outpaced production by 28K units, clearing the inventory worry. But at 359x earnings, deliveries are just the ticket in — margins and ASPs on the 7/22 call are judgment day.
News — The market is digesting two doubts: how much high gas prices flattered the number, and how much demand was pulled forward ahead of price hikes. The 12% run-up earlier in the week had fully priced the beat — the news landing was the top. Next catalysts: the 7/22 print and robotaxi city expansion.
Short-term · 1–3 weeks
NeutralSidelines
The $399–419 moving-average band overhead needs time to clear; $370 is the June platform below. With IV elevated pre-earnings, directional bets are worse than premium-selling. The delivery catalyst is spent — 7/22 margins are the real watershed.
Entry Light size on a stabilized $370–385 retest; add only after the 7/22 printStop $355Target $420
Long-term · months+
Neutral
Deliveries back in growth, energy storage scaling, and the robotaxi option are a real growth story — but 359x earnings prepays most of it. The right long-term posture: keep a core position for the optionality, and save the add for a valuation air pocket.
If margins miss on 7/22, the valuation floor gives way
Demand fades once gas-price tailwinds and pull-forward buying pass
BYD and peers' overseas push eats directly into Europe and EM share
Community Voices
StockTwits · Cheers lasted three seconds before turning into 'who's distributing' — are you the flock or the lead wolfOriginal ↗
Institutional Views
华尔街共识 · 50 analysts, average target $403.6 (just +2.6% vs. spot), rating 1.77 — even Wall Street thinks it's fully priced
交付 vs 预期 · The 480K print steamrolled every raised estimate: Goldman 420K, Barclays 418K, Morgan Stanley 413KSource ↗
0.00%07-02“Up 150% in three months — analysts can't raise targets fast enough; driving in uncharted territory, the seatbelt matters more than the gas pedal”
Score7/10RSI(14) 52.3Off 52w high -11.4%P/E (TTM) 1703-month gain +154%
Technicals — Down 4.3% to $517.82, sitting exactly on the 20-day ($516.8), with the 50-day at $460 and the 200-day way down at $278 — the uptrend is fully intact, just too steep. RSI 52 is neutral, ATR $36 (7%). The $500 round number plus the 20-day form the bulls' first must-hold line.
Fundamentals — The MI lineup keeps taking AI inference share, with revenue +35% and EPS +123% — but 170x trailing earnings means the market is paying on 2027's delivery. The Street's $511.8 average target sits below spot — not bearishness, just target-raising that can't keep pace with the tape.
News — No stock-specific negative — this is sector profit-taking. In early June's Broadcom-triggered AI chip squeeze, AMD dropped 10.9% in a day and fully recovered — money still treats it as the top offensive play on the AI inference era.
Short-term · 1–3 weeks
NeutralSidelines
A trend this steep isn't for fresh heavy positions — 7% daily ATR on a 170x multiple amplifies any sector panic. Holders should trail stops to protect gains; those flat should wait for the 50-day squat.
Entry Light momentum size if $500 holds; the better entry is a deep retest of $460 (50-day)Stop $460Target $585
Long-term · months+
Accumulate
Inference is a bigger, longer-lasting market than training, and AMD's price-performance play is structurally favored there. With datacenter and client both firing, 2027 earnings can grow into today's multiple. Pullbacks toward the 50-day are long-term add zones.
170x leaves zero tolerance — one guidance stumble is -20%
NVIDIA price cuts or CUDA lock-in fight back and stall the share-gain story
If the AI capex peak thesis lands, its high beta amplifies the downside
Community Voices
WSB · Mentions at 27, a monthly low: too high to pound the table, too strong to bottom-fish — it's always quiet near the summitOriginal ↗
Institutional Views
华尔街共识 · 59 analysts, average target $511.8 (-1.2% vs. spot), rating 1.26 — the tape has lapped the targets; a wave of raises is pending
板块观察 · If MI300X/MI400 share keeps expanding it directly compresses NVIDIA's premium — AMD is the biggest beneficiary of the AI narrative rebalancingSource ↗
New today“The miner-turned-cloud story is seductive, but Meta crashing the party means the script gets rewritten — no conviction talk until the dust settles”
Score7/10RSI(14) 30.6Off 52w high -49.5%1-month move -39.9%Revenue YoY +110%
Technicals — Down 10.4% to $38.82 — off 25.7% on the week and 39.9% on the month, halved from the 52-week high. RSI 30.6 sits on the oversold line with the 200-day ($49.1) now overhead. Dual hot+oversold channel status means attention and selling pressure peaked together — a classic narrative-collapse chart, with $35 the nearest round-number support.
Fundamentals — Revenue +110%, with Microsoft/NVIDIA-linked AI cloud contracts underpinning the valuation — but the market has started asking how fast they convert. Bernstein bluntly calls it 'behind on scale and enterprise capability' versus CoreWeave and Nebius. Meanwhile legacy bitcoin-mining cash flow shrinks through the pivot — being between two stools is the most dangerous phase.
News — The trigger: Meta announcing its entry into AI cloud infrastructure — a hyperscaler morphing from customer to competitor, shaking the pricing-power assumption of the whole neocloud group. IREN, one of the smallest and most aggressively pivoting names, got sold the hardest.
Short-term · 1–3 weeks
NeutralSidelines
A stock down 40% in a month can bounce violently, but Meta's entry is structural, not sentimental — it takes time to digest. If you play the bounce, size it like a lottery ticket; below $32 the market is pricing 'failed pivot.'
Entry Consider a left-side position only after $35 holds — two straight days without new lowsStop $32Target $50
Long-term · months+
Neutral
The power capacity and site pipeline are real assets, and AI compute leasing demand is real too. But the competitive map changed overnight — the endgame for smaller neoclouds may be acquisition rather than becoming giants. That makes this an M&A option, not a growth stock.
Score8/10RSI(14) 29.4Off 52w high -49.5%P/E (TTM) 15.6YTD -34.3%
Technicals — Closed down 0.4% at HK$94.1 on a day the Hang Seng rallied 1.3% — glaring relative weakness. RSI 29.4 is oversold, just 6% above the HK$88.65 52-week low, with the 20-day (HK$105) as the first reclaim threshold. The +2.3% week is a faint stabilization signal, but a 27.6% monthly wound needs time.
Fundamentals — 15.6x earnings with a HK$186 average target across 33 analysts — absurdly cheap on paper, but the reasons are real: a bottomless instant-commerce subsidy war, regulators publicly rebuking inflated discounts, and heavy AI capex with fuzzy near-term monetization. The Qwen full-stack AI play (including robotics) is the one option institutions keep flagging.
News — Regulators called out Alibaba, JD.com and peers for misleading subsidy claims, surfacing policy risk in the instant-commerce price war; the U.S. ADR made new 52-week lows below $96 in tandem. Consensus remains 'moderate buy,' but left-side money is visibly waiting for a right-side signal.
Short-term · 1–3 weeks
NeutralSidelines
Relative weakness hasn't lifted; anything short of the 20-day is a dead-cat bounce. A break of the HK$88.65 low triggers fresh technical selling — no left-side heroics on a trade.
Entry For a trade, wait for HK$105 (20-day) to be reclaimed; HK$88–95 is for long-term scaling onlyStop HK$86Target HK$115
Long-term · months+
Accumulate
Cloud plus commerce at 15.6x, with the Qwen ecosystem and instant-commerce user asset marked at zero — a textbook 'all the bad news is in the price' setup. Subsidy wars end; when this one does, margin recovery and a re-rate arrive together. Scale in, think in years, no leverage.
The subsidy war drags on and margin recovery keeps receding
Regulatory risk re-prices the entire platform economy
AI capex fails to monetize and cloud growth disappoints
Community Voices
StockTwits · Retail is bottom-fishing while earlier bottom-fishers get liquidated — value-faith posts and capitulation posts share the same feedOriginal ↗
Institutional Views
华尔街共识 · 33 analysts, average target HK$186 (+98% vs. spot), rating 1.17 — doubling upside on paper against a relentless downtrend; a textbook case of broken consensus
监管动态 · Regulators rebuked platforms for inflated subsidy claims; the instant-commerce war faces policy disciplineSource ↗
