Why the Market Sold Off: A Five-Layer Anatomy
This week’s decline is not a single catalyst but a sequential accumulation of sentiment shocks, each amplifying the next. Understanding the layering matters because each thread has a different duration and severity — some are now resolved, some are not.
Layer 1 — Broadcom’s guidance reset (June 3–4). The origin point for the broader AI sentiment correction was Broadcom reporting a technically strong quarter but declining to raise its full-year AI chip target beyond $100 billion. Q3 AI chip guidance came in at $16 billion, short of the $17.2 billion that institutional whisper expectations had priced in. The stock fell 14–15%, wiping nearly $300 billion in market value and breaking the market’s assumption that AI semiconductor guidance would only ever go up. That expectation reset seeded the vulnerability for everything that followed.
Layer 2 — Korea circuit breaker (June 22–23). South Korean stocks triggered circuit breakers twice in a single session. SK Hynix and Samsung Electronics both plunged 12%, dragging SanDisk, Western Digital, and Micron down 9–13% in global sympathy. This was the most acute single-session event of the week, driven by the KOSPI’s heavy weighting toward semiconductor firms and the fragility of crowded long positions in memory names that had surged 200–300% year-to-date.
Layer 3 — Alphabet’s AI talent exodus (June 22). Gemini co-lead Noam Shazeer — who co-authored the foundational “Attention Is All You Need” transformer paper and for whom Google had paid $2.7 billion in a 2024 acquisition — departed for OpenAI within two years of returning. Within 48 hours, Nobel Prize-winning DeepMind scientist John Jumper announced his move to Anthropic. Alphabet’s stock fell roughly 7%, erasing approximately $250 billion in market capitalisation. This was an entirely company-specific event — the broader tech sector ETF actually rose that same day — but the symbolic weight of two marquee researchers leaving simultaneously for rivals raised structural concerns about Google’s frontier AI competitiveness.
Layer 4 — Hyperscaler capex anxiety. Combined 2026 capital expenditure guidance across the major U.S. cloud platforms exceeded $452 billion, with Alphabet projecting $175–190 billion and Amazon separately flagging approximately $200 billion. Alphabet’s Q1 free cash flow fell 47% year-over-year to $10.1 billion; Amazon’s trailing free cash flow collapsed 95%. The market is now asking a question it was not asking six months ago: can these companies monetise their AI infrastructure fast enough to justify the investment scale? That is a legitimate concern without a clean near-term answer.
Layer 5 — Federal Reserve and technical breakdown. The Warsh Fed’s hawkish June meeting, in which nine of eighteen policymakers projected a 2026 rate hike, combined with a Nasdaq break below the 7,370–7,380 gap-fill zone to trigger systematic de-risking. Defensive names — Microsoft’s software segment, Walmart, Procter and Gamble, Johnson and Johnson — saw inflows, while IBM bucked the tech selloff with a 5% gain on a JPMorgan upgrade tied to AI software acceleration.
The anatomy of this selloff is important for supply chain positioning. Layers 1 and 2 (AVGO guidance + Korea circuit breaker) were sentiment-driven, not fundamental. Layers 3 and 4 (Alphabet talent + hyperscaler capex) carry genuine structural questions but apply primarily to the hyperscaler layer — not the hardware supply chain. The supply chain names in USINO.AI’s core tracking universe — TSMC, SK Hynix, Sphere Corp, Filtronic, Marvell — benefit from AI infrastructure demand regardless of which hyperscaler wins the model war. This distinction matters when assessing which names represent dip opportunities versus which face genuine impairment.
Micron Reports the Strongest Quarter in Its History
Micron Technology reported its fiscal third-quarter 2026 results after the close on June 24 and delivered what its own management described as the strongest quarter in the company’s history. Revenue reached $41.456 billion, up 346% year-over-year and approximately $6 billion above the Wall Street consensus of $35.59 billion. Adjusted earnings per share of $25.11 compared to a consensus estimate of $20.28 — a beat of nearly 24%. Net income was $28.24 billion, versus $1.89 billion in the same period a year earlier. Cloud memory alone grew more than 300% year-over-year to $13.77 billion. Data centre revenue climbed more than sevenfold to $11.5 billion from $1.53 billion one year ago. The company also reported over $5 billion in data centre solid-state drive revenue.
The Q4 guidance was the figure that moved markets in after-hours. Micron guided fiscal fourth-quarter revenue at $50 billion, plus or minus $1 billion, against a Wall Street consensus of $42.915 billion — nearly $7 billion above expectations. Q4 EPS guidance of $30 to $32 per share compares to the $24.80 consensus. The company separately announced its first-ever Strategic Customer Agreements — long-term supply contracts with key AI infrastructure partners — and declared a quarterly dividend of $0.15 per share payable July 21. The stock rose approximately 10% in after-hours trading, with the VanEck Semiconductor ETF (SMH) up 3% and the Roundhill Memory ETF (DRAM) surging 10% in sympathy.
Micron’s Q4 guide of $50 billion is the single most important data point for Asian supply chain investors this quarter. It confirms that AI memory demand is not decelerating — it is accelerating. For the USINO.AI coverage universe: SK Hynix (000660.KS) is Micron’s most direct competitive analogue and should see a sharp relief rally when the KOSPI opens Thursday. TSMC’s CoWoS capacity expansion and the entire advanced packaging chain are validated. Sphere Corp (347700.KQ) and the Korean OSAT complex benefit from confirmed HBM demand continuity through at least calendar Q4 2026. The Cantor Fitzgerald thesis — that 2027 supply will be even tighter than 2026, extending earnings growth into 2028 — is now far more credible after tonight’s result.
AI Stock Triage: Decline Risk vs. Dip and Short Squeeze Opportunities
The sell-off has been indiscriminate. USINO.AI separates the AI names into three buckets based on fundamental impairment, short interest dynamics, and MU’s read-through catalysis.
GOOGL (~$344). The most structurally impaired of the major AI names right now. Three concurrent headwinds are non-trivial: the departure of two of the most consequential AI researchers in the industry within 48 hours; a $84.75 billion secondary equity offering that suspended buybacks and diluted shareholders; and free cash flow that collapsed 47% year-over-year due to a $175–190 billion capex programme that generates no immediate revenue. The DJIA inclusion on June 29 creates a mechanical passive-fund bid that may temporarily support the price, but it is not a fundamental catalyst. A sustained recovery in GOOGL requires demonstrable evidence that the Gemini product roadmap remains on track without its key architects, which the market cannot verify for at least two quarters.
SPCX (~$150). Down roughly 33% from its post-IPO peak of $225.64. The first bond issuance adds a new layer of capital structure complexity to an equity that already traded on pure sentiment rather than earnings. The August unlock of approximately 10% of insider shares — roughly doubling the free float — remains the dominant structural overhang. No earnings record, no float normalisation, and now bond dilution: the decline risk remains asymmetric until August earnings.
AVGO (~$380). Down approximately 23% from its $495 peak. Gross margin compression from mix shift to lower-margin ASICs, growing competitive pressure from MediaTek in custom silicon, and Google’s projected share of Broadcom’s TPU revenue declining from 95% to 65% by 2028 are all medium-term headwinds. The MU beat may produce a sympathy bounce today, but the next hard catalyst — Broadcom’s own fiscal Q3 earnings in early September — is ten weeks away. Analysts are nearly uniformly bullish (42 buys, average target $509), but near-term valuation compression is not yet complete at 65 times trailing earnings.
MU — Highest conviction in the complex. Delivered a beat so large it reframes the entire sector narrative. The $50 billion Q4 guide is nearly $7 billion above consensus. Pre-earnings implied volatility was near 120%, with heavy put concentration across the $900–$970 range during this week’s selloff. Those positions are now deeply wrong-way. Covering pressure on short sellers and put holders will amplify any gap-up. The next earnings date is September 22, providing a long runway for the institutional rotation back into the name. Key support zones that held: $970–$1,090.
NVDA (~$198) — Strong dip candidate. Trading 16% below its all-time high of $236.54 set May 14. Bank of America reiterated Buy with a $350 target following CFO Colette Kress’s appearance at the BofA Technology Conference, where she laid out an addressable market expansion roadmap: $40 billion per gigawatt for Blackwell Ultra, rising toward $60–80 billion for Vera Rubin and approximately $100 billion for Feynman. Bernstein explicitly called NVDA “absurdly cheap” at current levels. The next earnings are August 26 — MU’s blowout is the most direct read-through signal available for Nvidia’s data centre demand picture, and it is unambiguously positive. Short interest has been building during the pullback; a reversal could be sharp.
SK Hynix (000660.KS) — Asian squeeze candidate. The Korea circuit breaker event was technically extreme and sentiment-driven. Hynix’s HBM business is structurally aligned with Micron’s — both supply the same hyperscaler demand base. Micron’s blowout is a direct proxy validation for Hynix’s own demand picture. The circuit-breaker selloff created a mechanical oversell that MU’s result now directly counters. Watch for an outsized gap reversal when the KOSPI opens.
MRVL (~$289 pre-pullback) — Watch closely. Jensen Huang’s public characterisation as a “next trillion-dollar company” candidate attracted both institutional buyers and short sellers during the record run. The pullback into the broader selloff creates an entry if MRVL can reclaim $289. Custom silicon demand for AI networking — Marvell’s core business — is directly validated by MU’s data centre revenue acceleration. Bernstein upgraded MRVL alongside NVDA and AMD. Next catalysts are dependent on hyperscaler capex confirmation, which is exactly what MU’s Q4 guide provides.
SpaceX Back to Earth: What the −33% Crash Means for Asian Supply Chain Stocks
SpaceX completed the largest IPO in history on June 12, 2026, raising approximately $75 billion at $135 per share and debuting at $160.95 on Nasdaq — a 19.2% first-day gain. The stock continued climbing to an all-time high of $225.64 on June 16 before reversing sharply. As of June 24, SPCX was trading at approximately $150–155, down 33% from peak and back near IPO-day closing levels. The pattern is textbook post-IPO momentum collapse: pre-IPO supply chain run-up on contract narrative, IPO-day spike on retail euphoria, a brief peak as momentum peaked, then a sustained de-rating as the float begins normalising. USINO.AI’s SpaceX Asian Supply Chain report identified this risk explicitly. The question now is which supply chain names are experiencing a price correction versus a thesis break.
The SPCX de-rating has pulled all three primary supply chain names covered in the USINO.AI SpaceX report lower in sympathy. The critical distinction: none of the underlying contracts has been amended, cancelled, or placed under review. The price declines are sentiment-driven, not fundamental. SPCX’s August insider unlock — which will roughly double the free float — and the first bond issuance since IPO are the dominant structural overhangs on the parent stock, and both are equity-specific, not supply-chain-specific concerns.
Supply Chain Name-by-Name Triage
Sphere Corp (347700.KQ) — ~35,200 KRW, ATH 57,300 KRW (March 17). Sphere ran hard in the March–May pre-IPO window on its confirmed 10-year, $1.05 billion superalloy supply agreement with SpaceX — the purest rocket hardware contract in the small-cap space supply chain. The stock is now down approximately 39% from its March ATH and roughly 18% over the past month. However, over the past twelve months it remains up 270%, reflecting the transformational nature of the SpaceX contract relative to the company’s prior revenue base. The $55.48 million 2026 delivery commitment is unchanged. This is price mean-reversion after an extreme run, not fundamental deterioration. Watch for stabilisation around the 30,000–32,000 KRW range as the SPCX floor firms.
Filtronic (FTC.L) — ~381p, ATH 480p. Filtronic’s RF and GaN amplifier relationship with SpaceX — covering more than $115 million in cumulative Starlink satellite contracts — remains intact. The stock is trading at approximately 381p, well above its pre-SpaceX-IPO base, and its 52-week range of 1.64p to 480p illustrates the scale of the re-rating the SpaceX relationship generated. Analyst consensus sits at approximately 287p, meaning the stock continues to trade at a meaningful premium to fundamental targets even after the pullback — indicating the SpaceX supply premium is still being priced in. The next earnings date is August 4, which will be the first post-IPO financial read on Starlink contract revenue recognition.
WNC Corp (6285.TW) — ~272 TWD post-spike, pre-IPO base ~153 TWD. WNC experienced one of the most dramatic single-session moves in the Taiwan supply chain complex — surging from a prior close of 153.50 TWD to 272.50 TWD on June 12 (IPO day), a near-77% intraday gain. As the dominant manufacturer of Starlink user terminals and ground-station hardware, WNC’s relationship with SpaceX is structural and long-duration: SpaceX has actively directed WNC to expand its Vietnam manufacturing footprint, locking in supply-chain geography ahead of any potential tariff escalation. Even at current post-spike levels, WNC trades above its pre-IPO base. However, the 52-week high of 336.50 TWD represents meaningful further upside if SPCX stabilises.
The SPCX post-IPO collapse follows a well-documented pattern: the parent equity de-rates as IPO euphoria fades and float normalises, and supply chain names de-rate in sympathy even when their contract positions are unchanged. The key Munger inversion here is that the price decline is not evidence the thesis is wrong — it is evidence the market over-priced the IPO halo. The SPCX supply chain floor will likely be set by two events: first, whether SPCX holds the $147–155 technical support zone; and second, the August earnings release, which will be the first disclosure of actual revenue against the Starlink and Starship supply commitments. Until August, Sphere Corp, Filtronic, and WNC remain in a sentiment-driven holding pattern. The contract fundamentals — Sphere’s $1.05B agreement, Filtronic’s GaN amplifier relationship, WNC’s terminal manufacturing lock-in — are intact. Investors with a six-to-twelve-month horizon who bought into the USINO.AI SpaceX report thesis have not had that thesis invalidated. They are experiencing post-IPO normalisation, not thesis failure. Tickers: SPCX · 347700.KQ · FTC.L · 6285.TW · 2313.TW.
Alphabet’s Structural Risks and the June 29 DJIA Inclusion
Alphabet faces a rare convergence of talent, capital, and competitive pressures in 2026. The departures of Noam Shazeer — who co-authored the 2017 transformer paper and returned via a $2.7 billion acquisition in 2024 — and John Jumper, the Nobel Prize-winning lead researcher on AlphaFold2, within 48 hours of each other represent the most concentrated loss of frontier AI talent from a single institution in the industry’s short history. Separately, Alphabet raised $84.75 billion in a secondary equity offering, the largest in U.S. corporate history, to fund AI infrastructure at a level that has already compressed Q1 free cash flow by 47% year-over-year and is guided to “significantly increase” further in 2027.
One counterbalancing mechanical factor: S&P Dow Jones Indices confirmed that Alphabet will join the Dow Jones Industrial Average effective June 29, replacing Verizon. DJIA-tracking funds will be required to purchase Alphabet shares in proportion to their AUM before the open on that date. This creates a passive bid that is structurally time-limited — the rebalance happens once — but may provide near-term price support. It is not a fundamental catalyst and should not be confused with renewed institutional conviction in the stock’s direction.
USINO.AI Supply Chain Watchlist — June 25 Triage
The USINO.AI Daily Brief is complimentary market intelligence for informational purposes only. Nothing herein is investment advice, a solicitation, or a buy/sell recommendation. USINO.AI does not provide regulated financial advisory services. Conduct independent research and seek professional advice. Past performance is not indicative of future results. © 2026 USINO AI PTE. LTD. All rights reserved.