Katie McGinty: The energy economy’s biggest waste problem is already inside the system

The article argues energy supply expansion should shift toward faster-deployable efficiency and thermal recovery rather than only building new generation. It cites untapped potential of over 3,000 TWh/year of usable waste heat, data-center cooling efficiency improvements (PUE moving from ~1.5–1.6 to <1.3), and claims technologies like absorption chillers can cut chiller electricity by up to 90% and heat pumps can slash energy bills by 32% while cutting emissions by 60%. Net message: unlocking existing capacity could add “new supply” quickly, though it does not eliminate the need for large-scale grid and generation investment.

Analysis

The market will probably read this as a capex-shift story, not a demand-shock story. If AI operators can harvest efficiency faster than they can add generation, the near-term winners are the picks-and-shovels around thermal management, controls, and waterless cooling rather than the obvious power-supply beneficiaries. That favors names like CARR, JCI, TT, and ETN on 1-3 quarter order acceleration, while reducing the urgency premium embedded in merchant-power and scarcity trades such as VST, CEG, and parts of the gas-turbine complex.

The key second-order effect is that lower effective load growth can delay the point at which grids, peakers, and new transmission become binding constraints. That matters because a lot of AI-linked power optimism assumes linear growth in net electricity demand; if PUE improvements and heat-reuse systems take even a modest bite out of the curve, the multiple expansion for pure power beneficiaries becomes more fragile over 6-18 months. The flip side is that data center operators and industrials that can document lower cooling intensity may win on both cost and siting flexibility, so the value capture may sit closer to equipment vendors than to utilities.

Contrarian take: this is likely directionally right but probably early. Most hyperscaler capex is still about getting connected and powered, so the first-order load story is not broken; what changes is the slope of the curve and who gets paid along the way. The thesis is falsified if quarterly data-center commentary keeps showing PUE stuck near legacy levels or if utility load forecasts continue to surprise materially to the upside despite efficiency retrofits.

Key Decisions for Investors

• Over the next 1-3 months, accumulate CARR/JCI/TT on weakness as the cleanest expression of faster-deployable energy productivity; use a basket to reduce single-name execution risk.
• Pair trade for 3-6 months: long CARR, short VST or CEG. The thesis is that efficiency adoption monetizes sooner than new generation scarcity, compressing the premium on pure power-tightness names.
• If you want lower theta exposure, use 6-12 month call spreads in CARR or JCI rather than outright equity longs; the catalyst is quarterly backlog/order commentary, not an immediate macro print.
• Stay underweight or hedge exposure to power-scarcity beneficiaries until hyperscaler disclosures confirm that higher AI workloads are outpacing efficiency gains; watch for any guidance that PUE improvement stalls above ~1.4 as a thesis break.