Record TSMC Quarter Sold, Chips 1% From the Crash Line
TSMC reported a record quarter and chips sold off 4.5% anyway. SOXX closed 19.1% below its June peak, about 1% above the line I call a crash.
The risk read for this session
These are model estimates, not forecasts anyone can guarantee, and not investment advice. The dot before each probability is its band: 🟢 low · 🟡 moderate · 🟠 high · 🔴 very high, with thresholds set separately for each scenario. SOXX is the main US semiconductor ETF, the basket I track; from peak is the fall from the highest close of this cycle; VIX is the market’s fear gauge; the high-yield spread is the extra interest riskier companies pay to borrow, my early-warning line for funding stress. My internal alert is at Level 3, on a scale where 3 is the highest.
📌 THE DAY IN THREE LINES
The market sold its best news: TSMC reported a record quarter with spending plans pointed higher, and the chip sector fell 4.46% anyway. The SOXX semiconductor ETF closed at $530.50, 19.1% below its June peak — about 1% above the crash line. My one-month crash odds nearly doubled, from 34% to 65%.
Today my internal alert stands at Level 3, the highest of my three levels, activated for the first time.
📰 WHAT HAPPENED THURSDAY
TSMC, the company that manufactures most of the world's advanced AI chips, reported Q2 revenue of $40.2 billion, up 33.7% from a year ago and at the top of its own guidance, with a 67.7% gross margin and earnings per share up 77% (official SEC filing). On the earnings call, management pointed 2026 capital spending toward the high end of its $52-56 billion range, as reported by industry press. The stock fell 2.3%. Commentators read the higher spending as more risk, not more growth — a sign of how the mood has changed.
The selling was broad and concentrated in AI: memory stocks fell 10.2% as a group for the second straight rout (SK Hynix's US listing −13.7%, Sandisk −12.6%), pressured by reports of a planned $8.5 billion IPO by Chinese memory maker CXMT and by unconfirmed reports of possible new US export limits on advanced memory. Oracle fell 6.3% to a one-year low. Speculative cloud names fell 6.2% as a group. Meanwhile the rest of the market absorbed the blow: the S&P 500 slipped just 0.51% to 7,533.77, and the average US stock actually rose (the equal-weight S&P was up 1%). Money rotated out of AI into defensives — an orderly exit so far, not a panic. The VIX fear gauge rose 7% to 16.73, still far below the 30 panic line.
🚦 OUR NINE SIGNALS (traffic lights)
▪️ 🟠 Moving averages: SOXX closed below its 50-day "waterline" (the average of the past 50 days' prices — my trend line) for a second day, on above-average volume. Nvidia closed below its own waterline for the first time since mid-June.
▪️ 🔴 Structure (worsened from 🟠): the index broke its July 7 floor at $536.12 after a series of lower highs — the textbook bearish sequence. This is my first fully red price signal.
▪️ 🟡 Momentum: still negative on the daily chart, no confirmed weekly breakdown yet.
▪️ 🟠 Breadth: only 3 of my 55 US names rose; just 24% remain above their own waterline.
▪️ 🟠 Rotation and canaries (worsened from 🟡): money is rotating out of AI in an orderly way; my five "canary" names trail the sector by 7.4 points over 20 days, and Oracle hit a one-year low.
▪️ 🟠 Distribution: two new "distribution days" (sessions of heavy institutional selling) — seven now active.
▪️ 🟢 Volatility: VIX at 16.73 with a normal term structure. No panic — which, with my red euphoria signal, still reads as complacency.
▪️ 🟢 Credit: high-yield spreads steady at 271 basis points (as of July 15, FRED). The debt tap is still open.
▪️ 🔴 Euphoria: the sector still sits 36% above its 200-day average, and 45% of fund managers in BofA's July survey call the AI bubble the top tail risk.
Bottom line: one family of signals is now broken (price structure), two more are strained. The overall regime stays at ALERT, but my operational alert rises to Level 3 — the maximum — because the sector has spent two days below its waterline on volume and because the one-month odds crossed my 45% action threshold.
📏 KEY LEVELS
▪️ SOXX: $530.50, −19.12% from its June 22 peak of $655.95. The −20% crash line is $524.76 — 1.08% below Thursday's close. The deeper-crash line (−30%) is $459.17.
▪️ Nvidia: $207.40, −12.3% from its May peak of $236.54; its −20% line is $189.23.
▪️ S&P 500: 7,533.77, just 1.0% below its June 2 record close of 7,609.77.
🏭 THE SUPPLY CHAIN, GROUP BY GROUP (Thursday's average moves)
Memory and storage −10.2% (worst: SK Hynix ADR −13.7%). Server makers −6.2%. Speculative cloud −6.2% (worst: Nebius −13.9%). Networking −5.7% (Corning −9.2%). Chip designers −4.8% (best: TSMC −2.3%; worst: Marvell −8.7%). Equipment −3.6%. Data-center power −3.3%. EDA software −1.9%. Hyperscalers −1.8%, the most resilient group: Apple +1.8% and Microsoft +1.4% rose while Alphabet fell 4.4%.
📊 TODAY'S ODDS (same three questions every day)
▪️ 🔴 Chip-sector crash within one month (SOXX closing 20% below its peak, at or under $524.76): 65%, up sharply from 34% yesterday.
▪️ 🟠 Deeper crash within three months (30% below the peak): 24%, up from 14%.
▪️ 🟡 Broad-market crisis within three months (S&P 500 down 20% with AI at the epicenter): 6%, up slightly from 5%.
Why such a big jump? Mostly mechanics: the crash line is now just 1.08% away, and this market has produced four daily drops of 2% or more in the last two weeks alone. One ordinary bad day crosses it. On top of that: a broken price floor, collapsed breadth, seven distribution days, and a market that sells record earnings. Pulling the other way: credit is calm, the broader market is rotating rather than fleeing, and the Fed is priced at ~90% odds of no move in July. I flag the size of this revision openly — my rules require an identifiable trigger for any jump this large, and today there is one.
💳 CREDIT & MACRO
High-yield spreads at 271 bps — no stress signal from the debt market, though Oracle's bond spreads have been widening in secondary trading. Treasury yields edged up: 10-year 4.57%, 2-year 4.16%. Futures markets price roughly 90% odds that the Fed holds rates steady on July 28-29 — a volatile week for those odds, which briefly priced meaningful hike risk before Chair Warsh's testimony.
👀 THE NEXT TWO WEEKS
SK Hynix earnings July 22 — a key test with memory stocks in rout. The Fed decision July 28-29. Then the main event: Microsoft, Meta, Alphabet and Amazon report July 29-30, and their AI spending guidance is the single most important variable for this market — especially after it just sold TSMC's higher capex. Samsung follows July 30; Nvidia reports August 26.
🛡️ CAPITAL PRESERVATION NOTE
With my top alert level active and the crash line one session away, I repeat what this observatory is for: these are informational estimates, not instructions. None of my four de-escalation conditions is met (no new distribution days, sector back above its waterline, a valid follow-through day, improving breadth: 0 of 4). Anyone with real money at stake should be talking to a licensed advisor, not a dashboard.
📚 SOURCES
Market data: Polygon, FMP and Alpha Vantage end-of-day APIs (session of July 16). TSMC results: official SEC filing; call details via industry press (Data Center Dynamics, TradingKey). Session chronicle: CNBC, Motley Fool. Credit: FRED (high-yield spreads), Investing.com. Fed odds: CME FedWatch as cited by financial press.
Model estimates — not certainties, and not investment advice. I am not a licensed advisor. All investment decisions belong to you, ideally with professional guidance.