Artificial intelligence is driving the largest capital reallocation in business history. That capital does not materialize as software in the cloud — it materializes as power plants, grids, data centers and networks. And infrastructure has owners.
The five largest technology companies are projected to spend nearly $700 billion on capital expenditure in 2026 — roughly 60 percent more than in 2025. As AI inference costs fall about tenfold a year, the cost of cognitive work is collapsing. But that spending does not stay in the cloud.
It materializes as power plants, electrical grid upgrades, data centers, cooling systems, and transmission lines. Every dollar of AI investment creates a corresponding demand for physical infrastructure — and the companies that own that infrastructure are among the principal beneficiaries of the buildout. For three years investors captured the AI story through semiconductors, hyperscalers and software. That was logical and profitable, but it captured only the part that runs on code. The part that runs on concrete, copper and kilowatts has been systematically undervalued.
HALO describes the category of businesses that own and operate the physical foundation of the AI economy: regulated utilities generating the power, grid operators moving it, data-center REITs housing the compute, and the transport and water networks that serve the broader economy.
These businesses share three characteristics. Their revenues are contracted or regulated. Their assets are irreplaceable. And their relevance to the AI economy is growing, not shrinking. Critically, the opportunity is global — the buildout is driving infrastructure investment across Europe, Asia-Pacific and the Americas, opening a universe of high-quality real-asset businesses that domestic-focused allocations typically miss.
A single large-scale data center consumes as much power as a small city. Electricity — not chips — is the gating factor on the AI buildout, and every dollar of AI capital expenditure creates a corresponding demand for baseload generation and reliable grid infrastructure.
The opportunity spans the full energy value chain and the essential infrastructure around it. These are not glamorous growth companies. They are regulated or semi-regulated businesses with long-duration contracted revenues, inflation-linked tariffs, and capital-return profiles that have compounded through multiple economic cycles. The AI transition does not disrupt them — it increases their relevance.
Integrated utilities generate and distribute the power; transmission operators build the high-voltage networks; renewable generators supply the clean baseload hyperscalers increasingly require.
Data-center REITs house the compute; transport networks move the physical economy; water and healthcare REITs serve demand that is demographic and acyclical, not discretionary.
Real assets offer what most technology stocks do not: reliable, growing income. Regulatory frameworks that link returns to invested capital reward exactly the infrastructure the AI transition demands.
AI’s most binding constraint is not compute — it is electricity.
An investment that benefits from the AI transition but is also vulnerable to its consequences is a more complicated bet than it appears. Real assets are different: the same characteristics that make them essential to the AI economy also make them resilient to its risks.
In the workplace, AI is already improving productivity — fewer employees are needed to get the job done. When AI replaces a $75,000 worker, the margin improvement is real and immediate. But that $75,000 was also someone’s spending. Aggregate the displacement across an economy and you simultaneously improve the supply side and erode the demand side. That consumer-demand question will eventually be resolved — through policy, new distribution mechanisms, and adaptation — but the timing is uncertain. In the interim, the best-positioned companies are those least dependent on discretionary consumer spending.
Utilities earn regulated returns on invested capital; infrastructure operators hold long-duration concessions; REIT leases run for years and decades — not quarterly subscription renewals.
The customer base — governments, municipalities, hyperscalers, institutional operators — spends on demographic necessity, regulatory mandate and long-term capital planning, not on consumer confidence.
People need electricity, water, healthcare and transportation regardless of how the economy is restructured. Demand for these services is not cyclical the way consumer spending is.
This does not mean real assets are immune to all risk — rising rates can pressure valuations, regulation can shift returns, construction costs can overshoot. But the specific risk the AI transition may create — an erosion of consumer purchasing power that threatens discretionary businesses — is precisely the one these companies are structurally built to withstand.
Investors have taken to calling this the HALO trade. It is a misnomer. A trade is something you enter on a catalyst and exit when it resolves. But AI power demand compounds annually regardless of what oil is doing; decades of infrastructure underinvestment do not reverse when a crisis de-escalates; the energy transition requires trillions in real-asset deployment that no ceasefire or rate cut will cancel. What looks like a trade is a structural reallocation — capital recognizing, perhaps belatedly, that the physical economy cannot be arbitraged away.
GARA is not a bet on which technology wins or how quickly automation displaces workers. It is a bet on the physical infrastructure the AI economy cannot function without — and on the durability of the income those assets generate through whatever transition follows.
The AI transition is not just a technology story. It is an infrastructure story. The question is whether your portfolio is positioned for both.
Concise notes on the themes shaping our funds, delivered when we publish.