Industrial stocks moving with semiconductor ETF: the $2 trillion repricing Wall Street’s classifications missed

On May 28, 2026, Caterpillar joined a 2 gigawatt power alliance with American Intelligence and Power to build dedicated natural-gas generation for the Monarch Compute Campus, with first deliveries booked for September 2026 (Caterpillar IR release). That contract is one data point in a broader pattern. A one-year daily-return correlation study by Renaissance Macro’s Neil Dutta, surfaced on TCAF 245 (episode link) and reported by (Equipment Finance News), found 15 non-tech S&P 500 companies, worth roughly $2 trillion combined, trading with the VanEck Semiconductor ETF SMH at correlations of 0.5 or higher. Twelve are industrials.

Industrial stocks moving with semiconductor ETF SMH is not a curiosity. It is a quiet repricing of what the U.S. industrial complex actually sells, and a stress test of the sector framework most index funds still rely on.

Key facts at a glance

• 15 non-tech S&P 500 names trade at 0.5+ one-year daily-return correlation with SMH, per Renaissance Macro Research, surfaced on TCAF 245 (Equipment Finance News).
• 12 of the 15 are industrials, with CAT, VRT, ETN, and CMI named on the podcast (TCAF 245).
• Caterpillar’s Q1 2026 Power Generation revenue: $2.817B, up 41% year over year, per (24/7 Wall St.).
• Hyperscaler AI capex tracked at more than $500B in 2026 (Goldman Sachs) and trending toward $1T by 2027 (CNBC).

The bull case

The market is doing what GICS has not. Caterpillar sells backup generator sets and engines into hyperscaler data centers, Vertiv sells cooling and power management, and GE Vernova supplies the turbines that power the campuses. Yet all three remain classified as ordinary industrial cyclicals tied to mining, construction, and traditional capex. When the order book changes, the share price follows the new buyer, not the old label. Neil Dutta’s framing on TCAF 245 captures the point bluntly: these names “trade like semis because their order books have become AI capex order books” (TCAF 245).

The mechanical implication shows up at the index level. Equal-weighted RSP carries roughly twice the industrials weight of cap-weighted SPY, so the diffusion of AI capex into power and components is captured more fully outside the megacap tech cohort. That helps explain why RSP has led SPY in 2026 even as the headline AI story still gets framed around chips (24/7 Wall St. equal-weight read). The same broadening pattern is showing up in active manager commentary: a CNBC roundup flagged a single industrial name as Josh Brown’s Best Stocks pick precisely because of its AI buildout exposure (CNBC industrial pick), and (LongYield) frames it as the S&P 493 converging on the AI infrastructure trade.

The bear case

Diffusion of revenue is not diffusion of profit. NVIDIA captures the overwhelming share of AI accelerator spend at gross margins north of 75%, and the four hyperscalers driving the buildout earn software-like operating margins. Caterpillar and the rest sell commodity hardware into a cyclical order book at mid-teens operating margins, and those orders can be deferred or cancelled. The io Fund’s read on the buildout sizes the AI capex pool aggressively but flags the same concentration of profit at the top of the stack (io Fund analysis).

Cresset Capital’s 2026 outlook makes the second-derivative point sharper: if AI revenue disappoints, whether through OpenAI burn, inference-cost collapse, or in-house custom silicon at the hyperscalers, the first repriced names are the derived-demand suppliers, not the chip designers, because turbine and generator backlogs have no software gross margin to absorb the cut (Cresset 2026 outlook). Capital Group’s framing is more measured but agrees that breadth alone does not solve concentration risk in the cap-weighted index (Capital Group).

In short, equal-weight’s industrial overweight may be higher beta to an AI capex air pocket, not safer diversification away from it.

What the data shows

A quick comparison of the two index structures against the AI-capex flow.

Source for weights and 2026 leadership pattern: (24/7 Wall St. equal-weight read).

A second cut, sized in dollars, sharpens the gap between revenue and profit.

Sources, in order: (Goldman Sachs), (CNBC capex), and (24/7 Wall St. CAT analysis).

What would change our mind

Three observable signals would invalidate the broadening read.

1. A cancellation wave in CAT’s Power Generation backlog, or an explicit guide-down on data-center engine orders in a subsequent quarterly print.
2. A sustained reversal of RSP leadership over SPY, paired with industrials weight in RSP contracting back toward parity.
3. Direct hyperscaler capex guidance cuts in the next earnings cycle, especially if paired with a public shift toward more aggressive custom-silicon substitution.

If any two of those land in the same quarter, the second-derivative names should be expected to reprice harder than the chip stack, not in line with it.

What this does not tell you

The original correlation study has not been fully disclosed in the public reporting reviewed here. The full list of 15 names, the rolling window, the sector definitions, and the lookback methodology are not documented in the cited episode or in the secondary press write-up (Equipment Finance News). Vertiv, Eaton, Caterpillar, and Cummins are confirmed by name on TCAF 245; the other names referenced on the podcast were spoken phonetically and have not been independently verified.

Correlation also is not causation. A 0.5+ daily-return correlation with SMH is consistent with shared AI capex exposure, but it is also consistent with a common factor like rates, dollar moves, or risk appetite that lifts both cohorts together. The argument that AI capex is the causal driver rests on the order-book evidence at the company level (CAT Power Generation up 41% year over year, the 2 gigawatt Monarch alliance), not on the correlation number alone.

None of this is investment advice, and nothing here is a recommendation to buy or sell any security or index fund.

FAQ

What does “industrial stocks moving with semiconductor ETF” actually mean?

It refers to the one-year daily-return correlation between specific industrial company shares and the VanEck Semiconductor ETF (SMH). When two stocks rise and fall together day after day at 0.5 or higher, the market is treating them as exposed to a common driver. In this case, that driver looks like AI capex.

Why hasn’t sector classification caught up?

GICS sector buckets update slowly and rely on a company’s stated revenue mix. Caterpillar still reports as construction, mining, and energy. The data-center Power Generation segment is growing fast but is not yet reclassified, so index providers and most passive funds still treat it as a traditional industrial.

Is equal-weight the obvious way to capture this?

Equal-weight indexes like RSP carry a structural overweight to industrials versus cap-weighted SPY, which mechanically captures more of the AI infrastructure diffusion. The bear case is that the deepest profit pool stays with the cap-weighted top of the stack, and equal-weight is therefore higher beta to AI capex disappointment, not lower risk. The honest answer is that the trade-off is a barbell question, not a switch.