Goldman Sachs’s HALO Framework for AI Plays

Goldman Sachs is urging investors to look beyond high-flying software and chip names and buy into a quieter kind of AI winner: heavy-asset companies. Under the label HALO—Heavy Assets, Low Obsolescence—the bank’s research identifies firms in energy, infrastructure, industry and semiconductors that own large, long-lived physical assets and face little risk of being made obsolete by new technology.

In its European screen, Goldman names BMW, ASML, Rheinmetall, Equinor and STMicroelectronics among potential beneficiaries, alongside telecoms groups Orange, Vodafone and Telefónica. The logic is straightforward: the AI build-out requires enormous investment in data centres, power grids, chip fabrication plants and high-capacity telecoms networks—expenditure that flows directly to the operators of that physical kit.

The structural bias favours Europe. According to RBC BlueBay portfolio manager Siddhi Purohit, about 38% of the European equity market consists of companies with a high share of tangible assets, against just 18% in the US. Ritholtz Wealth chief Josh Brown argues such businesses “are not only not displaced by AI, but actually benefit from the expansion” of supporting infrastructure. Goldman notes that since the start of 2025, European HALO stocks have outperformed their capital-light peers by around 35 percentage points; in emerging markets the gap since late 2025 is 108 percentage points.

Why Infrastructure-Heavy Stocks Are Outpacing Tech Rivals

The structural advantage of European equity

Europe’s higher concentration of asset-heavy industrials, energy and telecoms names means the region is naturally over-weighted in the HALO trade. While US indices are dominated by services and tech platforms, Europe offers a larger pool of companies potentially set to benefit from the same AI investment that has already sent US semiconductor stocks soaring. The outperformance data cited by Goldman—35 percentage points in Europe since early 2025—suggests the market has begun pricing this theme, but the gap could widen further if infrastructure spending accelerates.

Where ASML fits into the AI value chain

ASML, the Dutch semiconductor equipment maker, is the classic HALO example. It manufactures the extreme ultraviolet lithography machines that are essential for producing the world’s most advanced chips—the very chips that power AI models. Although ASML is not a software or AI platform company, it captures a toll on every leading-edge chip produced. Morningstar points to ASML as a standout HALO stock because its machines are not only expensive and difficult to replicate but also sit at the heart of an AI-driven capex cycle that is still in its early innings.

Earnings season may confirm the trend

Goldman expects the upcoming reporting season to reinforce the HALO thesis. It forecasts that capital-intensive companies in the energy, grid and chip sectors will deliver solid earnings, powered by the early stages of AI infrastructure deployment. If those numbers materialise, the relative attractiveness of heavy-asset plays versus software-only or low-capex services firms could sharpen, especially as the latter face greater AI disruption risk in their own business models.

What the HALO Theme Means for Portfolio Construction

  • Screen for durable, hard-to-replicate assets. Focus on companies that own the data centre power infrastructure, chip fabrication tools, defence platforms and telecom grids that the AI build-out cannot do without—exactly the sectors Goldman highlights.
  • Don’t abandon valuation discipline. ASML, for instance, already trades at a premium multiple. Even within the HALO universe, entry price, debt levels and earnings momentum remain decisive. The framework supplements, but does not replace, bottom-up analysis.
  • Watch the next earnings season for confirmation. If European heavy-asset firms report rising order books tied to data centre and grid expansion, the 35-percentage-point performance gap could widen, making current momentum a genuine investment signal rather than a one-off re-rating.

Risk & Opportunity Assessment

Commercial RiskMediumThe HALO trade depends on sustained AI capex. If corporate spending on data centres, power grids or chip plants slows, the earnings tailwind that underpins Goldman’s thesis could fade, especially for companies with high fixed-cost bases.
Competitive RiskLowThe whole premise of HALO is low obsolescence and high barriers to entry. ASML’s lithography machines, Rheinmetall’s defence systems and existing grid infrastructure have few rapid substitutes, limiting competitive pressure in the near term.
Regulatory RiskMediumUtilities and telecoms named—Orange, Vodafone, Telefónica—operate in price-regulated segments. Any tightening of allowed returns on grid or spectrum assets would cap the benefit of AI-driven infrastructure spending for these names.
Reputation RiskLowThere is no immediate reputational angle for heavy-asset industrials. The narrative is anchored in tangible assets meeting infrastructure demand, not in ESG or governance controversies specific to this theme.
Technology DisruptionLowThe thesis explicitly selects assets with low technological obsolescence. ASML’s EUV tools and the physical grid are not at risk of being made redundant by a new software platform; they are prerequisites for digital expansion.
Commercial OpportunityHighGoldman reports a 35 percentage point outperformance by European HALO stocks versus capital-light peers since early 2025. Continued AI infrastructure build-out could extend this advantage, creating significant return potential for holders of the right heavy-asset names.