Peter Thiel's Fund Reported Zero Stocks for 2 Straight Quarters. Its $419 Million Comeback Put 72% Into Energy and Power.

Thiel Macro (Peter Thiel’s hedge fund) reported no U.S. long stock holdings for the quarters ended Dec 2025 and Mar 2026, then disclosed 8 positions totaling $418.7M as of Jun 30. The portfolio is heavily tilted to energy/power (about 72% or ~$301M) with Amazon (~$118M, 28%) as the lone tech holding, plus Vista Energy (~$76M), Vistra (~$59M), and four regulated utilities (AEP, DTE, FirstEnergy, CMS) and X-Energy (~$3.7M). The news mainly signals a positioning theme around AI power supply bottlenecks rather than immediate market-moving fundamentals.

Analysis

The market implication is not a generic “buy energy” call; it is a relative-value rotation from compute hardware to the bottlenecks that monetize compute: generation, transmission, and load-serving capacity. That favors merchant power first, regulated utilities second, and upstream fuel third, because the earnings reset in power tightness can show up within one to two quarters, while rate-base compounding takes longer but is stickier.

The more interesting second-order risk is that AI capex can remain strong while equity returns migrate away from semis. If power availability becomes the gating item, cloud and hyperscale operators may keep spending but see slower time-to-revenue, which can compress software and GPU multiples even if unit demand stays healthy. That makes the market’s current “all AI winners together” framing too simplistic over 1-3 months.

The contrarian angle is that the filing is a lagging, opaque signal: the fund can hide hedges, and a small 13F snapshot can overstate conviction. Still, if we do get evidence over the next 1-2 earnings cycles that data-center load is lifting utility demand or that forward power curves stay firm, the utility basket should work as a low-beta way to own the theme. If regulators push back on rate recovery or if power prices normalize, the trade loses its edge quickly.

Key Decisions for Investors

• Long VST vs short NVDA for 1-3 months: express the view that the next AI bottleneck is electrons, not chips. Best entry on any post-earnings NVDA strength / VST pullback. Falsifier: sustained upward revisions to GPU demand without corresponding tightening in power markets.
• Build a small basket long AEP, CMS, DTE, and FE on 3-6 month weakness: these names are the cleaner way to own data-center load growth with lower commodity risk. Target is modest multiple expansion plus rate-base growth; stop if regulators delay cost recovery or utility load forecasts disappoint.
• Treat XE as a venture-style option, not a core position: only buy if you want 12-18 month nuclear optionality and can tolerate binary execution risk. The thesis needs real utility contracting, not just AI enthusiasm, to de-risk.
• Avoid chasing AMZN as a direct “power beneficiary” trade; it is more likely to absorb higher infrastructure costs than to monetize them quickly. Use it only as a barometer for whether hyperscaler capex is being slowed by grid constraints.
• Set an alert for 1-2 quarter utility commentary on load growth and interconnection queues. If those data points do not inflect, the entire power-bottleneck thesis should be cut.