These Billionaire Investors Keep Betting on AI Infrastructure Despite Recession Fears

David Tepper, Chase Coleman, and Bill Ackman are all increasing exposure to AI infrastructure, with Amazon, TSMC, Broadcom, and Vistra appearing across their latest 13F filings. The biggest split is Microsoft: Tepper cut his stake 82% and Coleman cut 54%, while Ackman initiated a new $2.1 billion position. The article is broadly constructive for AI sentiment, but the data are backward-looking and unlikely to move markets materially on their own.

Analysis

The most important signal here is not “AI is still hot,” but that the trade has narrowed from software enthusiasm into bottleneck ownership. When multiple large allocators converge on the same infrastructure stack, incremental upside migrates to the scarcest inputs: advanced packaging, foundry capacity, memory, networking, and power. That makes the current setup more durable than a simple momentum rotation, because capital spending is now being validated by buy-side positioning rather than just hyperscaler guidance.

The clearest second-order beneficiary is TSMC, which sits at the choke point of the AI supply chain and should continue to gain pricing power as leading-edge capacity remains tight. AMAT and AVGO also look underappreciated relative to the headline AI names because they monetize the buildout regardless of which model wins; that lowers single-product risk and makes them cleaner infrastructure expressions over a 6-18 month horizon. VST is the more interesting “hidden beta” trade: if data-center load growth stays elevated, power capacity and interconnection become as valuable as compute, which can re-rate utility-like assets with merchant exposure.

The contrarian issue is that consensus is clustering in the same crowded names while underweighting the probability of a capex digestion phase. If hyperscalers pause spending for even one quarter, the market will punish the highest-multiple beneficiaries first, especially those with the least visibility into near-term revenue conversion. GOOGL looks like the weakest relative position in this basket because its AI narrative is more platform-option value than direct monetization, and the negative ownership signal suggests fund managers are still questioning whether its capex is translating into defensible share gains.

The biggest tactical risk is not an AI demand collapse, but a sentiment unwind from any of three catalysts: weaker cloud bookings, semicap lead-time normalization, or a policy-driven move lower in power prices/energy sentiment that reduces the scarcity premium on data-center infrastructure. That said, the setup still favors a 3-9 month continuation trade in semis and power, while MSFT appears more “show me” than the others due to capex skepticism already discounted into positioning.

Key Decisions for Investors

• Go long TSM / short GOOGL for a 3-6 month relative-value trade: own the supply-side choke point against the least-convincing AI monetization narrative; target 10-15% outperformance if AI capex stays firm, with a stop if cloud spending trends decelerate.
• Buy AMAT on pullbacks and hold 6-12 months: this is a cleaner way to express AI capex than chasing the megacaps; risk/reward improves if equipment orders re-accelerate, with downside limited versus more crowded AI leaders.
• Initiate a basket long AVGO + NVDA vs short equal-weight software index over 1-2 quarters: favors hardware/content capture over application-layer names that may see slower monetization; use as a hedge against AI spending broadening rather than contracting.
• Add VST as a medium-term power-scarcity hedge: 6-18 month holding period, with upside tied to persistent data-center load growth; trim if power policy or merchant pricing turns adverse.
• Stay underweight MSFT tactically for the next 1-2 quarters unless you see evidence of capex converting into faster Azure share gains; the market is likely to demand clearer proof before re-rating the stock.