The first part of this story is now well-understood.

AI needs chips, power, cooling, and data centers. It needs copper wire, concrete, and grid capacity. The market began pricing this in two years ago. Semiconductors, utilities, and infrastructure stocks have run hard, driven by figures that are difficult to ignore: Big Tech is planning roughly $635 billion in AI infrastructure spending in 2026 alone, and the IEA expects global data-center electricity consumption to double by 2030. Nobody is getting paid to tell you that AI needs atoms anymore. That trade has happened.

The second part of the story arrived more recently and has been painful.

Software got repriced. The old assumption was that software companies were automatically safe because they were asset-light and high-margin. AI broke that logic. U.S. software and data-services stocks lost roughly $1 trillion in market value during the February 2026 selloff. IGV, a broad software ETF, is down 20% from the start of 2025 through March 31, 2026. This is a violent reversal from the early phase of the AI story, when software was treated as a likely beneficiary (note: IGV doubled in the first 2 years after the launch of ChatGPT).

While the Iran war and its implications have taken some recent focus, we are beginning to see the next AI trade. The market is seeking to find: Which businesses have AI-proof moats, and which are just expensive wrappers around workflows that AI can now perform?

The Collapse of Terminal Value

In mid-March, Chamath Palihapitiya argued that we are witnessing the "Collapse of Terminal Value." The argument is direct: AI lowers the cost of disruption so dramatically that many companies can no longer project cash flows beyond five years with any confidence. If a business moat is built on human expertise, proprietary information, or a process advantage that exists simply because "that’s how it’s always been done," AI can attack it.

This moat problem extends well beyond SaaS. Think about professional services. The entire business model of consulting - strategy firms, management consultants, IT advisory shops - is built on selling human expertise by the hour. When a client can get a 90% quality answer from an AI agent in ten minutes, the willingness to pay $500 an hour for a junior analyst starts to look different.

Architecture firms, design studios, and marketing agencies (just to name a few) are in the same crosshairs. These businesses sell a combination of taste, judgment, and specialized knowledge. AI doesn’t need to replace a great architect to pressure firm margins; it just needs to make a good-enough architect 5x more productive. If the same output requires significantly fewer billable hours, the moat evaporates and valuations compress.

What AI Cannot Build (Yet)

This is where capital moves toward assets that AI cannot disrupt. We are looking for things with physical defensibility and inelastic demand - things a better large language model cannot unbundle overnight.

• Fixed Physical Scarcity: AI can optimize a booking engine, but it cannot create more beachfront. In mountain towns like Aspen or Jackson Hole, the geography is fixed, the zoning is restrictive, and the buildable land is finite. The land is the moat.

• High Capital Intensity: A cruise line or an oil refinery requires years of permitting, massive financing, and immense physical labor. AI can improve itinerary planning, but it cannot produce a cruise ship or a pipeline network in a fiscal quarter.

• Geological Constraints: Gold remains the original hard-to-replicate asset. Its supply grows at roughly 1.5% a year, a rate set by geology and physics, not by the speed of a GPU cluster.

The Caveats: To be sure, this thesis isn't without risk. "Embodied AI" - humanoid robotics and autonomous construction - is a next stage wildcard that could eventually shrink build times and erode physical moats. Furthermore, as AI-driven virtual experiences (VR/AR) improve, some luxury demand may bifurcate into the digital realm. But for the next 3 - 5 years, and maybe much longer, the friction of the physical world remains the ultimate defensive barrier.

The Investor Playbook: A Framework for Alpha

The transition from "AI-driven growth" to "scarcity-driven value" requires a different lens. Investors should start identifying the "physical wall" - assets that gain value because they cannot be digitized, simulated, or easily permitted into existence.

1. The Positional Moat: Ultra-Luxury and Social Scarcity

The first layer of this trade exists outside the public markets in the realm of the 0.1%. As AI drives massive productivity gains, wealth will continue to concentrate at the top. The items that define this lifestyle are protected by social scarcity: they are used both for enjoyment and as a definitive signal of status.

• The Assets: Beachfront estates in land-constrained markets, private ski-mountain allotments, and timeless artwork.

• The Logic: You can use AI to generate a perfect "new" masterpiece, but you cannot use it to manufacture the provenance of a 500-year-old canvas. These assets are "positional" - their value is derived specifically from the fact that others cannot have them.

2. The Atomic Anchor: Hard Commodities

While software is now infinitely replicable, the elements required to power and build the physical world are governed by physics, not code. Investors looking for a pure play on "atoms" should look toward commodities that AI and robotics cannot easily synthesize or extract.

• The High End: Gold remains the ultimate store of value in a world of digital abundance. It has no counterparty risk and no disruption risk.

• The Industrial End: Copper is the literal nervous system of the AI age. Even if AI designs a better grid, it still needs the copper to build it.

• The Strategy: Our experience suggests that mining companies often struggle to capture the full upside due to operational risks and capital mismanagement. For this trade, the most effective path is often owning the direct commodity itself rather than the producers.

3. The Permissioned Infrastructure: Public Market Proxies

Within the public markets, alpha will be found in businesses that own "permissioned" assets. These are entities that benefit from massive capital expenditure requirements and a regulatory environment that makes new competition nearly impossible.

• Pathways and Pipelines: Physical infrastructure like railways and energy pipelines is virtually impossible to replicate today. The combined hurdles of environmental permitting and land acquisition create a permanent moat. These businesses move the physical world, and their "terminal value" is anchored in the earth.

• The Permitting Bottleneck: Look for companies that transform raw inputs into essential end-products - such as refineries or specialized chemical manufacturers. The "moat" here isn't just the factory; it’s the legal right to operate a facility that would take a decade to permit from scratch today.

• Upper-Middle-Class Leisure: This category relies on "lock-in" to scarce natural resources and massive upfront capex. Cruise lines own finite port access and years of shipyard capacity. Mountain resorts own the geography. While valuations in this space can be volatile - as seen in the recent re-rating of mountain resort operators - the underlying scarcity of the experience remains a powerful long-term tailwind.

The Final Word

The next stage of value capture is not in who builds AI, but rather in what AI still cannot build very fast.

The first stage of this market was the infrastructure to create intelligence. The second stage, still playing out, is the destruction of vulnerable software moats and of businesses chiefly selling intelligence. The third stage - now beginning to unfold - is the re-rating of scarce physical assets in a world of abundant digital labor. Focus on what is hard to permit, positionally advantaged, and impossible to download.

ArcVest is a fee-only fiduciary registered investment adviser. This article is for educational purposes and does not constitute personalized investment, tax, or legal advice. Investing involves risk, including the potential loss of principal. Past performance does not guarantee future results.