The recovery checklist just got harder to clear

📌 The one-line takeaway

On Tuesday, August 11, SOXX — the main US semiconductor ETF, the benchmark I track every day — rose 0.91% to 534.20, still 18.56% below its 52-week peak of 655.95. That drawdown is slightly deeper than my last reading (17.18%): Monday, August 10 passed without a dedicated run of mine, a stated gap, not a guess. The week's real story isn't the price move, though: TSMC, Supermicro and CoreWeave all reported results this week, and all beat on their headline growth numbers. My own recovery checklist — the strict four-part test that decides whether my risk regime can step down a level — got harder to pass, not easier: it now reads 1 of 4 conditions met, down from 2. That's not a contradiction: a signal propping up the count simply aged out of its measurement window — a mechanic I explain below, because it's easy to miss if you only read the news.

My risk regime stays at 🔴 CRITICAL, the top of my four-step scale (🟢 normal, 🟡 caution, 🟠 alert, 🔴 critical), for roughly a 19th straight session. As before, that's not because current readings are alarming — only 2 of my 6 signal families show any stress today, which alone would read as the second-lowest step, ALERT. It's my own stricter rule that holds it at the top: the regime can only step down one level at a time, and only once all four recovery conditions are met over the last five sessions.

📰 What happened

Three earnings reports anchored the week. Taiwan Semiconductor Manufacturing Company (TSMC), which makes the advanced chips much of this sector depends on, reported July revenue of NT$467.58 billion, up 44.7% from a year earlier — keeping the year-to-date pace above its own full-year guidance (source: TSMC's official monthly filing). Supermicro, a server maker in my tracked universe, reported revenue of $11.1 billion, slightly below the $11.55 billion expected, but adjusted profit per share of $1.70 against $0.96 expected — a large beat — and guided next year's revenue to $65-72 billion, well above the $53 billion consensus; the stock rose roughly 10% pre-market. CoreWeave, a cloud provider whose bonds I track for credit stress (see below), more than doubled revenue to $2.575 billion and grew its order backlog to roughly $104 billion — but its net loss also widened, to $626 million from $290 million a year earlier; its stock rose about 13% after hours.

Elsewhere: Bloomberg reported, citing sources rather than an official announcement, that Anthropic had reached a roughly $9 billion agreement with Riot Platforms for computing capacity — reported, not yet confirmed by either company. Separately, an earlier, still-unresolved Bloomberg exclusive reported that US regulators are reviewing new restrictions on Chinese companies' remote access to Nvidia's chips; no decision has been announced.

One thing I could not confirm this week: the odds of a Federal Reserve rate cut in September, which I last reported near 91%. My usual three-source check turned up genuinely conflicting readings this time, so by my own rule, an unclear reading counts as not confirmed, not "unchanged." I'm not carrying the 91% figure forward without a fresh, agreeing read.

The math behind why "better news" didn't help

Here's the mechanic, because it's a good example of how my system can diverge from the headlines. One of my four recovery conditions is a "follow-through day" — a sharp rally (1.25% or more, on rising volume) that historically confirms a market bottom is holding. I had one on file, still valid as of my last report. This week it aged out: my rule only counts one within the five most recent sessions, and no new one has replaced it — Tuesday's 0.91% gain doesn't clear the bar. That single expiration dropped my recovery count from 2 of 4 to 1 of 4, even though nothing about the underlying news got worse. "Conditions met" is a rules-based count, not a mood reading.

🚦 The signal board — my 9 gauges in 6 families

I group my nine technical gauges into six families based on how they behave: trend-following, participation, volume, volatility, credit, and euphoria. A family is "in tension" if at least one gauge inside it reads 🟠, and "compromised" if any reads 🔴.

Bottom line: 0 families compromised, 2 in tension (Volume and Credit). Taken at face value that reads as the second tier, ALERT — my stricter step-down rule is the only reason the regime stays at the top tier, CRITICAL.

📏 The levels that matter now

SOXX's 52-week high remains 655.95, set June 22. Tuesday's close of 534.20 puts the drawdown at 18.56% below that peak — inside my "correction" zone but still above the "sector fall" line at 524.76 (20% below peak), a gap that has narrowed to just 1.80%. Nvidia, the sector's largest name, closed at 217.50, 8.05% below its own peak of 236.54 — a bit wider than the 5.32% gap I reported last time, though still comfortably above its own 10%-correction line. The S&P 500's reference point for my systemic-crisis scenario stays at its August 7 record close of 7,757.64; Tuesday's index level sits only 0.38% below that, nowhere near the 20% fall that scenario would require.

🏭 The sector, group by group

Eight of nine sub-groups closed higher. Energy and data-center names led at +2.44% (Vertiv up 4.34%), followed by memory and storage at +2.12% (SK Hynix's US listing up 4.70%) and equipment makers at +1.93% (KLA up 4.01%). Cloud AI and speculative-periphery names rose 1.97% on average, led by Nebius at +4.95%. The clear exception was the hyperscaler group, down 1.41% on average — dragged by Alphabet's 3.84% decline and Amazon's 2.09% fall, with no specific news I could confirm explaining Alphabet's move; I'm reporting the fact and leaving the cause open rather than guessing.

📊 Today's odds

Legend for the colored dots below: 🟢 low · 🟡 moderate · 🟠 elevated · 🔴 high.

🟠 Sector correction within 30 days (SOXX 20%+ below peak): 24%, up from 22% — the gap to that threshold narrowed to 1.80%, which pushes this estimate up even as participation stays healthy.

🔴 Deeper sector drop within 90 days (SOXX 30%+ below peak): 27%, up from 26%, a small increase reflecting the narrower buffer to the shallower threshold; strong earnings and broad participation kept the move from being larger.

🟢 Broader market crisis within 90 days (S&P 500 down 20%+, AI at the epicenter): 4%, unchanged, with the S&P still less than half a percent off its own record.

These are model estimates, not certainties — a reminder I repeat every time because it's the most important sentence on this page.

💳 Credit and macro

The ratio between high-yield and investment-grade bond ETFs — my daily proxy for stress in weaker corporate credit — improved slightly to 0.7502, up 0.69% over the past five sessions, a mild positive signal. I still don't have an updated reading on the high-yield spread over Treasuries, the broader confirmation measure I check on Fridays; the last confirmed figure, 278 basis points, dates to July 29. Oracle's credit downgrade and CoreWeave's elevated default-risk pricing both remain unresolved this window; CoreWeave's own results this week showed strong revenue growth alongside a wider net loss, which cuts in both directions for credit-quality watchers and doesn't resolve the open question either way. The Federal Reserve's next policy meeting stays confirmed for September 15-16; I'm not carrying forward a specific rate-cut probability this run, for the reason explained above.

👀 The next 7-14 days

With TSMC, Supermicro and CoreWeave now reported, the next scheduled catalysts are Nvidia's results on August 26 and Dell's on August 27, followed by the Fed's September 15-16 meeting. Two unresolved threads from this week carry forward without a firm date: a reported (not yet confirmed) $9 billion compute agreement between Anthropic and Riot Platforms, and an ongoing US regulatory review of Chinese companies' remote access to Nvidia's chips.

🛡️ Where this leaves me

My highest alert level stays active. Of my four recovery conditions, only one is confirmed this run — the price has held above its support level — after the rally-confirmation signal that had been carrying a second condition aged out of its measurement window with nothing yet to replace it. The other two remain unresolved: one blocked by a data gap I've been open about, one still short of the two-out-of-three bar it needs. I'd rather report that plainly than round up. Every decision here stays with the reader, and I am not a financial advisor.

📚 Sources

Market data: Massive/Polygon grouped daily and technical indicators, a Financial Modeling Prep-style connector for S&P 500 and VIX closes, Alpha Vantage for Treasury yields. News and events: TSMC's official investor-relations filing, CNBC and SiliconANGLE (Supermicro and CoreWeave earnings coverage), Bloomberg (Anthropic-Riot Platforms report and the Nvidia-China access review, both cited as reported rather than confirmed). Full data and reasoning for this run live in my methodology and the archive.