Peter Thiel's Fund Rebuilds Portfolio Betting on AI Power Industry Chain
Peter Thiel's Thiel Macro has rebuilt a public stock portfolio of approximately $419 million in the second quarter of 2026 after two consecutive quarters of zero holdings.
The largest position is in Amazon at about $118 million, followed by Argentine shale oil producer Vista Energy at approximately $76 million.
The portfolio adds or increases holdings in energy and utility stocks such as Vistra, American Electric Power, DTE Energy, FirstEnergy, CMS Energy, and small modular reactor developer X-Energy.
Energy and power-related holdings account for more than 70% of the portfolio.
This layout focuses on the electricity, grid, and long-term energy demand driven by the expansion of AI data centers.
Market mechanisms show that event-driven capital is concentrating on the bottleneck segments of AI power: funds are shifting from pure computing power and chip themes to generation, transmission, and nuclear power assets, benefiting utilities and energy producers from expected load growth, while traditional tech stocks are under pressure.
Supplementary information shows that the related holdings are based on 13F disclosures, reflecting the situation as of June 30.
Source: Public Information
ABAB AI Insight
Top venture capital is repositioning after being in cash, shifting focus from direct AI targets to the physical energy infrastructure that supports computing power.
In terms of capital pathways, resource allocation is concentrated on regulated utilities, nuclear power, and shale oil and gas producers, motivated by capturing the certainty of demand and long-term contract value brought by the surge in data center loads.
Similar cases can be seen in the previous cloud computing expansion period with indirect bets on power and cooling infrastructure; currently, the transition is from chip bottlenecks to power and grid bottlenecks in AI scaling.
Structural judgment belongs to the reconstruction of the industry chain: the value capture point is shifting from models and hardware to upstream energy entry, with the mechanism being that the elasticity of power supply is far lower than the speed of computing power expansion, thus transferring the pricing power of scarcity.
ABAB News · Cognitive Law
The end of computing power is electricity
Bottlenecks determine the allocation of excess returns
Physical constraints outweigh software imagination