US Needs to Invest More in AI Buildout, Says Cato’s Frazier

Washington is weighing a national framework for artificial intelligence as the U.S. seeks to cement AI leadership. Policymakers and experts warn that energy capacity bottlenecks are constraining data center buildouts, creating a practical delivery risk to those goals (source: Bloomberg, Kevin Frazier interview).

Analysis

The near-term bottleneck to scaling large compute footprints plays out as an energy-infrastructure arbitrage: firms that can guarantee firm, dispatchable power (behind-the-meter generation, long-term PPAs, or on-site storage) will capture outsized pricing power versus pure-play real-estate landlords. Expect a two-tier market over the next 12–36 months where sites with immediate interconnect capacity trade at 20–40% revenue multiple premiums versus assets that require new transmission or transformer installs with 18–24 month lead times.

This dynamic creates durable demand for grid-capex and power-electronics supply chains: transformers, switchgear, microgrids, and BESS integration services have order books and lead times that translate into multi-year revenue visibility for incumbents. A credible federal push to de-risk permitting or subsidize interconnects would concentrate incremental returns into vendors and vertically integrated utilities in the first 6–18 months rather than into hyperscalers’ incremental datacenter build programs.

Downside scenarios are clear and quantifiable: faster-than-expected deployment of modular containerized compute (15–30% cheaper upfront and deployable within 3–6 months) or a rapid fall in battery costs could erase the premium for firm-power incumbents and re-price real-estate. Monitor three catalysts closely — congressional budget language for grid upgrades (3–9 months), major transformer manufacturing capacity announcements (6–18 months), and near-term blackout or pricing events that would accelerate procurement decisions by one quarter or more.

Key Decisions for Investors

• Pair trade (6–12 months): Long AES (AES) 30% position / Short Digital Realty (DLR) 30% — rationale: AES benefits from storage + PPAs and captures price for firming capacity; DLR faces the most exposure to delayed interconnects. Target return 25–40% pair payoff if grid service contracts accelerate; stop-loss at 12% pair drawdown.
• Directional long (12–24 months): Buy Fluence Energy (FLNC) equity or Jan-2028 call spread (buy FLNC 15, sell FLNC 25) — thesis: BESS integrators shorten lead-time to usable capacity. Risk: execution/conversion; reward: 2x+ if annual deployments grow 50%+; position size 2–4% NAV.
• Supply-chain long (12–18 months): Buy ABB (ABB) or Eaton (ETN) with a 12-month horizon — these vendors win transformer/switchgear orders with long lead times and limited new capacity. Target 20–30% total return; hedge with 6–12 month put protection if legislative funding stalls.
• Tactical short (3–9 months): Short or buy puts on a data-center REIT with weak hinterland power access (e.g., Digital Realty or an under-diversified peer) sized at 1–2% NAV — catalyst: any major interconnect denial or a disclosed >90 day transformer lead-time. Risk is limited to 10–15% premium spikes in leasing; reward is capitalizing on a 15–30% re-rating if leasing slows.