The Industrials of the Buildout

Which industrial companies benefit from the AI data-centre buildout — and which still offer attractive economics?

The Artificial Intelligence infrastructure cycle extends far beyond semiconductors and hyperscalers.

Every new gigawatt of computing capacity requires electrical equipment, cooling systems, power generation, cables, engineering and construction. This has created a second group of potential beneficiaries: the industrial companies building the physical infrastructure behind AI.

But exposure alone does not make an attractive investment.

In The Industrials of the Buildout, LFG+ZEST examines 239 companies across the industrial universe to identify where AI datacenters represent a meaningful source of revenue — and whether current valuations, margins and growth expectations still justify that exposure.

The analysis identifies 41 companies with material datacenter exposure, but only five currently derive more than 40% of revenue from the theme. Among the 28 names with at least 10% exposure, business models range from electrical and cooling equipment to grid infrastructure, on-site power and mechanical, electrical and plumbing contractors. 

The first important finding is that the market already treats this group increasingly as one common trade.

A principal-component analysis across thirty operating and market variables finds that one common pattern explains approximately 28% of the differences across the exposed companies and is strongly correlated with their datacenter exposure.

In practical terms, holding several AI-industrial beneficiaries does not necessarily provide as much diversification as the number of individual securities might suggest. 

Unlike the semiconductor supply chain, however, the economic fingerprint of the AI boom is not primarily visible in margins.

Across industrials, greater datacenter exposure has little relationship with the margin expansion analysts expect. Contractors largely benefit through higher volumes while maintaining relatively normal profitability. Selected equipment manufacturers with scarce products have captured more pricing power, but the broader effect of the buildout appears mainly in growth expectations and valuation multiples.

That distinction produces a counterintuitive valuation result.

Companies exposed to the buildout currently trade at approximately 1.44 times EV/EBITDA for each point of expected growth, compared with 2.05 times for the wider industrial universe.

The market therefore appears to believe the growth forecasts while simultaneously discounting their duration.

A slowdown in gigawatt construction over the next two to three years is already partially reflected in prices. 

But not all apparent cheapness is equally robust.

When current valuations are recalculated using each company's own five-year average margin, some of the most visible AI beneficiaries become materially more expensive. Other names remain relatively attractive even after profitability is normalized.

This distinction is central to the analysis: a low multiple during peak profitability is not necessarily the same thing as genuine value.

Crowding creates an additional complication.

Historical evidence suggests that during periods of exceptionally strong capital-expenditure intentions, the most controversial and crowded stocks tend to underperform even when the underlying investment cycle remains strong.

LFG+ZEST therefore developed its own Controversy Score, extending the analysis beyond US large caps to European and mid-cap industrial companies. Thirteen of the 41 exposed names currently sit in the most crowded fifth of the industrial universe. 

The final ranking combines five dimensions: valuation relative to expected growth, margin normalization risk, how far valuation has moved ahead of actual exposure, crowding and free-cash-flow yield.

The objective is not simply to identify the companies with the greatest AI exposure. It is to distinguish between exposure that remains economically attractive and exposure for which the market may already be assuming too much. 

Timing is critical.

The announced datacenter buildout appears sufficient to fund most supplier expectations through 2027. Beyond that point, growth in physical capacity slows significantly.

The next important evidence should therefore come from 2027 orders and backlogs.

Industrial suppliers generally have stronger balance sheets than the customers financing the buildout, meaning that financial stress may not provide an early warning. A slowdown is more likely to appear first in new orders, backlog conversion and valuation multiples.

The conclusion is therefore not a blanket call against AI-exposed industrials.

Part of the future slowdown is already reflected in market pricing. The more important risk is where current margins, valuations or shareholder positioning implicitly assume that today's exceptional buildout continues indefinitely.

The full video presents the principal findings of the analysis.

For readers who wish to explore the methodology, company-level screening and supporting data in greater detail, the complete research report is available upon request.

This material is provided for informational and research purposes only and does not constitute investment advice, a recommendation, an offer or a solicitation.