TSMC shatters records and ASML upgrades outlook — long the AI chipmakers

Verdict: the AI-chip thesis is real, but the entry isn't — wait for ADX to confirm

The thesis has genuine fundamental support: TSMC's Q2 profit rose over 77% (per CNBC, July 16, 2026), ASML raised its 2026 guidance a second time with a Q2 beat (per Yahoo Finance, July 15), and AMD's net margin jumped 6.4 points to 19.9% in the quarter ended June 27, 2026. But the trade isn't live: TSM closed at $433.24, above its 50-day EMA of $419.93, yet its ADX of about 11.0 sits well below the 20 trend-strength gate — the binding constraint. The strongest point for the idea is recognized revenue and margin expansion across all three names, not just guidance. The strongest point against is ownership posture: in the June 30, 2026 institutional and insider filings, AMD insiders were net open-market sellers of roughly $153.2 million while TSM insiders bought only about $370,000, and no robust parameter setup was validated because the sensitivity evaluation exceeded its time budget. What flips the verdict: TSM's ADX rising to 20 or above while price holds above the 50-day EMA would confirm the entry — or a close back below roughly $420 would invalidate the setup entirely.

Trade now: TSM entry is close, but not armed — three conditions remain outstanding

The TSM long setup is close to trigger, but it is not there yet. TSM last closed at $433.24, already 3.2% above its 50-day EMA of $419.93, so the price condition is effectively in place. However, the entry rule requires a fresh cross above that average plus an RSI (14) cross above 45 and an ADX (14) above both 18 and 20. RSI sits at 60.1 — already above 45 but with no fresh cross signal — and ADX reads about 11.0, roughly 9 points below the 20 threshold. In plain terms: the trend-strength gate is the binding constraint today. "Wait" means letting ADX build to at least 20 while price holds above the 50-day EMA; if price instead falls back below roughly $420, the entry resets and you should stand aside rather than anticipate. Risk framing is fixed by the rule set: a hard stop at a 2.4% loss from entry and a take-profit at a 4.8% gain, giving an effective reward-to-risk of about 2-to-1 on every trade, with positions capped at 25% of capital and sized to roughly 2.4% risk per trade. The backtest evidence supports the framework on the primary TSM pair: over a 60-month window the strategy traded 7 times, won 57.1% of them, and returned 89.6% with a 24.8% maximum drawdown. Note the exit-fill caveat: stops and targets were filled on daily bars, not intrabar data, so realized fill quality in live trading may differ. Do nothing today. Set alerts at ADX 18 and 20 on TSM, and at a close below $420 to invalidate the setup. The same conditions apply to ASML (last close $1,698.30, 50-day EMA at $1,725.85 — price is 1.6% below the average, RSI 45.9) and AMD (close $516.13, 50-day EMA at $484.78, RSI 66.2, ADX 32.1); AMD's ADX already clears the trend filter, but it needs a fresh RSI cross above 45. No parameter optimization was performed for this publication — the sensitivity evaluation exceeded its time budget, so no robust alternative setup was established; the strategy runs as designed.

AI demand is showing up in reported revenue, not just guidance

The idea's core claim — that AI hardware demand is showing up in actual reported results, not just guidance — checks out in the numbers. TSMC's fiscal 2024 revenue grew 33.9% year over year to NT$2.89 trillion, with operating margin up 3.05 points to 45.7% and net margin at 40.0%. That momentum has continued into 2026: per the CNBC report from July 16, 2026, TSMC's second-quarter profit jumped over 77% and beat estimates, and MarketWatch noted a record month of revenue ahead of that print. This is not a story about promised demand; it is a story about recognized revenue at near-46% operating margins. ASML confirms the same picture one step up the supply chain. Its fiscal 2025 revenue grew 15.6% to $32.7B, gross margin expanded 1.55 points to 52.8%, operating margin rose 2.67 points to 34.6%, and net margin hit 29.4% — a 92.9th-percentile return on equity of 49.0% among industrials. Critically, per the Yahoo Finance piece from July 15, ASML raised its 2026 guidance for the second time alongside a Q2 beat. Equipment suppliers raising outlooks twice in a year is a leading indicator for foundry capital spending, not a lagging one. AMD rounds out the trio with accelerating fundamentals: fiscal 2025 revenue grew 34.3% to $34.6B with free cash flow of $6.7B (99th percentile among IT peers), and the most recent quarterly sequence shows net margin jumping 6.42 points to 19.9% in Q2 2026 (period ended June 27) from 13.5% in Q1, with operating margin up 2.85 points to 17.3%. Per BofA's view cited by Yahoo Finance on July 14, AMD looks set for a beat-and-raise quarter on surging AI server demand — which the Q2 margin data already corroborates. The completed backtest supports the direction empirically. On daily bars over the 60-month window, the TSM pair produced an 89.6% total return across 7 completed trades with a 57.1% win rate, with all three symbols' data loaded and 1,235 bars evaluated. Shorter windows also finished positive: 34.3% over 24 months and 15.8% over 12 months. That means the strategy has been tested on completed historical trades, not just asserted. The stack is also structurally sound financially. ASML cut shares outstanding 2.0% year over year while growing profits; TSM carries a 2.36 current ratio and $2.13 trillion in cash against modest long-term debt; AMD holds $5.5B of cash with $6.3B of long-term debt against a $63.0B equity base. None of the three depends on leverage to generate the growth the thesis relies on.

A 25% drawdown history, insider selling, and valuation of an AI premium

Start with the volatility cost of this trade. In the 60-month backtest, the strategy's maximum drawdown was 24.8% — the equity curve sat underwater for long stretches, spending most of 2022 between roughly 15% and 27% below high water before recovering. In plain terms: this setup has regularly given back a quarter of its gains in a bad stretch, and the 89.6% return came from only 7 completed trades over five years, so the sample is thin. A single bad trade moves the result materially. The drawdown figures should also be treated as coarse — exits were filled on daily trigger bars rather than intraday data, so stop quality in fast markets may be worse than reported. Ownership posture is the clearest red flag. As of the June 30, 2026 institutional filings…

Scores

• Conviction score breakdown: 59
• Thesis support: 75
• Trade readiness: 40
• Risk quality: 55
• Backtest evidence: 45
• Fundamentals trend: 80

Watch items

• TSM — ADX (14) trend strength
• TSM — Close vs 50-day EMA
• TSM — RSI (14) fresh cross
• AMD — ADX (14) trend strength
• ASML — Close vs 50-day EMA
• AMD — Insider open-market activity