Palantir Billionaire Peter Thiel Has 42% of His Portfolio in These 2 Artificial Intelligence Stocks

Peter Thiel's Thiel Macro deployed into eight equities in Q2 2026 after holding no stocks in Q1, with Amazon and power producer Vistra comprising 42% of the portfolio, including a 28% Amazon position. Amazon's AI business surpassed a $25 billion annual revenue run rate and helped drive AWS revenue growth of 37% year over year, its fastest pace in 18 quarters. Vistra reported first-half adjusted EBITDA of $3.3 billion versus $2.6 billion a year earlier and guided to $6.8 billion-$7.6 billion for 2026, above $5.9 billion in 2025, while trading at a forward P/E of about 13.

Analysis

The relevant signal is not the disclosed investor ownership but the convergence of hyperscaler capex with increasingly scarce deliverable power. VST's upside depends less on aggregate U.S. electricity demand than on whether its generation is located in constrained load zones and remains unhedged enough to reprice into higher forward power curves. That creates potentially material operating leverage, but also means a low headline P/E may reflect the market discounting the finite duration of contracted or hedged earnings rather than a simple valuation anomaly.

AMZN is a cleaner demand-side expression of AI infrastructure, but near-term equity returns hinge on the spread between AWS growth and incremental depreciation, networking, and power costs. A sustained acceleration in cloud revenue can expand the AWS multiple only if management demonstrates that AI workloads are becoming durable, higher-margin consumption rather than discounted capacity used to defend share against MSFT and Google. The next 1-3 months should therefore be driven by cloud backlog, capex guidance, and margin commentary—not investor-following flows from a delayed 13F disclosure.

The contrarian risk for merchant power is that data-center announcements are being capitalized as firm load before interconnection, transmission, and financing constraints convert them into actual demand. Over 6-18 months, grid bottlenecks may benefit incumbent generation, but they can also delay load realization while new gas, storage, demand-response, and behind-the-meter generation cap regional price spikes. Any broad utility-sector rerating is unlikely: regulated utilities capture rate-base growth, while merchant generators carry power-price, fuel, weather, and political-intervention risk.

A disclosed concentrated portfolio is weak standalone evidence because filings do not reveal cost basis, derivatives, subsequent sales, or hedging. Treat the positioning as a watch catalyst rather than confirmation; the investable question is whether forward power prices and capacity revenues are rising faster than consensus EBITDA assumptions.

Key Decisions for Investors

• Initiate a 3-6 month long VST / short XLU pair only after confirming upward revisions to 2027-28 EBITDA and improving ERCOT/PJM forward power curves. Target 15-20% relative upside if merchant-power scarcity is being under-modeled; exit if VST lowers its multi-year EBITDA outlook, materially increases hedging at weak prices, or regional forward curves fall more than 10%.
• Maintain AMZN as the preferred hyperscaler AI long versus MSFT only if the next earnings release shows AWS growth sustaining above the peer-adjusted cloud trend without a meaningful AWS operating-margin decline. Use a 6-12 month horizon; the thesis is falsified by decelerating backlog or capex rising faster than revenue such that consolidated free-cash-flow estimates are cut.
• Do not chase VST solely on insider activity or the filing. Before adding, obtain the generation-by-market mix, hedge book, nuclear outage assumptions, and exposure to large-load contracts; absent this data, set an alert for an earnings revision rather than treating the forward multiple as evidence of undervaluation.
• Monitor CEG and NRG as read-throughs: CEG strength relative to VST would indicate the market is paying for nuclear-duration scarcity, while NRG outperformance would favor gas-fired capacity and retail-load optionality. A reversal across all three alongside weaker power forwards would be an early warning that AI-power demand expectations are being repriced.