Semiconductors Are the Market's Load-Bearing Wall — No Short Thesis Until This Breaks

The question I get most in a bull run is this. "We've already come this far — where do I cut?" My answer is simple. The moment semiconductors start weakening. Before that, there is no strong short thesis to build.

SMH started the post-2022 bottom at $130, ran to $280, pulled back to $160 in the tariff shock, and now trades at $470. That is a parabolic chart on a sector ETF. Sector ETFs do not usually look like that. And the move is still active — that's the point.

Why Semiconductors Are the Market's Load-Bearing Wall

In construction, a load-bearing wall is the wall that holds the building's weight. Remove it and the building collapses. In the market, semiconductors play a similar role. Almost all of the NASDAQ's last two years of move comes from semiconductor leverage. When semis are strong, NASDAQ is strong. When semis break, NASDAQ breaks. It's not a formula — it's an empirical rule.

How strong is the sector right now? SMH fell from $280 in mid-2024 to $160 in April 2025 — roughly a 43% drawdown. From that low it's now at $470. Call it a 3x off the bottom. And it's about 70% above its prior all-time high. That is not the behavior of a normal sector ETF — it's the behavior of a high-growth individual stock.

Why the Data-Center Narrative Still Works

Data-center capex is nearly the only category where companies raised guidance repeatedly throughout 2025. The four hyperscalers are committing hundreds of billions of dollars annually, and a large share of that spend ends up in AI chips, HBM, and networking silicon. The debate is about when this cycle ends.

What I see is simple. There is no concrete evidence the capex cycle stops in the next few quarters. Instead, the bottleneck is shifting. If 2024 was GPU-supply-constrained, 2025–2026 is constrained by HBM and foundry capacity. A bottleneck means pricing power sits with suppliers — and that's what sustains semiconductor sector margins.

Multiple Sectors Are Rotating In Together

Semis are not running alone. This week I'm watching the re-breakouts in XLC (communications) and XLK (tech):

• XLC — Meta and Google are each attempting breakouts. Meta reclaimed its 200 DMA and Google is knocking on box-range top.
• XLK — Apple is knocking on all-time highs again. It's been a laggard for several quarters.

Multiple sectors printing new highs at the same time is the strongest form of signal. A single-sector rally is easy to break; multi-sector participation thickens the flow.

What Happens in Price Discovery

The zone where an asset prints its first break of all-time highs is what I call a price-discovery zone. There is no prior inventory, resistance is thin, and once a trend sets, extension happens fast.

The most recent example is July–September 2025. SMH consolidated in the $280 box, then broke out — and ran to its current zone essentially uninterrupted. XLK and XLC are now in a similar posture. A successful break-and-retest here can produce a trajectory similar to H2 2025.

What Breaks This Scenario

I keep the opposite scenario ready. If semis weaken, all of this logic flips. The specific conditions I watch:

1. SMH falling more than 5% from its recent high and losing the 20 DMA
2. NVDA losing $200 and closing below
3. At least one of AVGO or AMD losing its one-month low

If those three trigger together, the thesis in this article is invalid. Until they trigger, I trust sector leadership.

Positioning — What I'm Actually Doing

In this environment I don't carry a short bias. I also don't lever up. The reason is the non-zero probability of geopolitical escalation. While headlines remain words only, the market ignores them. But if physical action arrives, thin-liquidity zones fill fast.

So for day trades I respect breakdown levels — SPY 697–698, QQQ 636–637, SMH beneath $450. For swings I hold sector leaders (NVDA, AVGO, META). Stops are level-based. Profits are trend-based.

FAQ

Q: Can I still open a fresh semi position here?

A: For ETFs like SMH, the prior high ($280) is structural support and price is about 70% above that. Risk/reward is not compelling. A pullback that retests the 20 DMA is a much better entry.

Q: Why is NVDA still the leader?

A: It owns more than 90% of the data-center GPU market, and no competitor is taking material share. AMD is chasing but the absolute scale gap is large. With Blackwell-to-Vera-Rubin roadmap already disclosed, growth visibility is high.

Q: Why do people keep saying "sell now"?

A: Most of those arguments are valuation-based. They may not be wrong. But "expensive" becoming "breaks" requires a catalyst, and right now there isn't one. Sell calls without a timing catalyst create opportunity cost until the trend actually flips.

Related: Meta Is a Dormant Volcano — My Top Watch This Week