The Real AI Trade Nobody Is Talking About

Thiel Macro LLC reported no reportable 13F holdings at year-end 2025 and again at March 31, 2026 — two consecutive quarters with no U.S.-listed equity positions disclosed. Then, in a filing submitted on August 14, 2026, the fund resurfaced with $418.7 million across eight U.S.-listed stocks. Amazon was the only technology company. The other seven positions were concentrated in oil, electricity generation, regulated utilities, and advanced nuclear power.

Not chips. Not AI software. Not memory or photonics. Power infrastructure.

Seven of the eight positions are power plays: Vista Energy at $75.9 million, Vistra at $59.1 million, American Electric Power at $42.2 million, DTE Energy at $40.3 million, FirstEnergy at $39.9 million, CMS Energy at $39.6 million, and small modular reactor developer X-energy at $3.7 million.

The mainstream AI investment narrative has been built around semiconductors: Nvidia, TSMC, memory, and photonics. Everyone is crowding into the layer that processes intelligence. Thiel bought the layer that powers it.

The portfolio points to a power-bottleneck trade. Data centers need firm capacity. Existing generators and utilities can monetize scarcity, grid investment, and long-term supply agreements as AI demand surges.

American Electric Power disclosed 69 gigawatts of new contracted load additions through 2030, backed by a $78 billion five-year capital plan. It is also participating in a public-private initiative with SoftBank and the U.S. government for a planned 10-gigawatt computing and energy campus in southern Ohio.

DTE Energy reported 2.4 gigawatts of executed data-center agreements: a 1.4 GW Oracle project under construction and a 1 GW Google agreement that was still awaiting regulatory approval at the company’s July earnings call. FirstEnergy reported total forecast data-center demand of 24.8 GW, 30% above its first-quarter forecast, with 6.4 GW contracted.

Amazon raised its 2026 capital-spending plan to $220 billion, with the increase tied heavily to AI infrastructure and capacity. T. Rowe Price technology fund manager Dom Rizzo estimated that hyperscaler capital expenditure could reach $1.5 trillion to $1.6 trillion in 2027.

“That capital has to land somewhere physical. It lands on power infrastructure.”

In my previous analysis on the five industries that will define the future, I argued that fusion energy and sustainable energy infrastructure were not merely long-term bets. They were structural consequences of AI demand, policy choices, and the hard physics of data-center power requirements. Thiel Macro’s portfolio is evidence that this thesis is entering institutional portfolios faster than many expected.

The specific move into X-energy is worth isolating. X-energy completed its Nasdaq IPO in April 2026 and reported approximately $1.1 billion in net proceeds. As of its 2026 reporting, its project pipeline comprised 144 potential Xe-100 reactors across the United States and United Kingdom — approximately 11.5 gigawatts electric if customers exercise their contingent rights in full — anchored by relationships with Dow, Amazon, and Centrica.

Amazon anchored an approximately $500 million Series C-1 financing round in 2024. The companies set a target to bring more than 5 GW of X-energy SMR capacity online in the United States by 2039.

Thiel Macro’s portfolio carries a risk that bullish coverage often underweights. Regulators, new generation, or customer-specific tariffs could weaken scarcity economics. More supply or stricter cost allocation could lower realized power prices and compress earnings even while data-center electricity demand rises.

Utility Dive’s first-quarter 2026 roundup described utilities as divided on data centers as affordability concerns loomed larger. That tension applies directly to regulated utilities including DTE, FirstEnergy, and CMS. When a utility signs a 1 GW data-center agreement, residential customers and regulators ask who will pay for the generation, transmission, and distribution upgrades required to serve it.

What Thiel Macro’s move signals is not just an investment opportunity. It is confirmation that the energy-infrastructure question — who controls the electrons, who builds the grid, and who owns the generation capacity — is now inseparable from the AI question.

I have argued in previous publications that fusion energy and sustainable fuels are not peripheral industries. They are foundational ones. This portfolio is a major institutional signal that parts of the market agree — and that the window to position before the thesis becomes consensus may be narrowing.

Everyone crowded into the chip trade. Thiel Macro bought the electrons underneath — often at lower valuation multiples and, in the regulated-utility positions, with comparatively predictable cash flows and regulated returns. That does not make the trade risk-free. It makes the trade different.

“The infrastructure play is not coming. It is already here.”

Angga Setiawan

Portfolio values are calculated from Thiel Macro LLC’s SEC Form 13F information table for the quarter ended June 30, 2026. Figures are rounded for display. A 13F is a partial disclosure of reportable long U.S. securities and is not a complete picture of a fund’s exposures, cost basis, hedges, or intent. Infrastructure pipelines and capital plans contain forward-looking statements and should not be treated as completed projects.

This publication is independent analysis, not investment advice. © 2026 Angga Setiawan | Policy & Power Review, Vol. I, Issue 003 | All views expressed are the author’s own.