Greg Abel Just Quantified Berkshire's AI Power Opportunity

Berkshire Hathaway CEO Greg Abel said the company’s AI upside is increasingly clear, with Berkshire Hathaway Energy (BHE) positioned to supply data-center power. Abel cited that in Iowa, data center customers accounted for ~8% of total demand last year, while BHE plans ~$33.5B of capex from 2026 to expand generation, storage and transmission. Berkshire also increased its stake in Alphabet this year, reinforcing a dual AI exposure through both cloud/AI platforms and power infrastructure.

Analysis

The market is still framing AI as a semis/software trade, but the cleaner second-order beneficiary is regulated power and transmission. Berkshire’s utility platform can turn incremental data-center load into longer-duration rate base growth, which is more durable than a one-time equipment cycle; the catch is that the economics only work if regulators allow timely cost recovery, so the near-term risk is capex dilution before earnings catch up.

That makes BRK.B a better AI proxy than most investors think: it has embedded option value on grid scarcity without the multiple risk of a pure-play infrastructure name. The flip side is that not every utility wins equally; regions with constrained transmission, slow interconnection, or hostile rate cases may see demand growth that benefits customers more than shareholders, while hyperscalers increasingly have incentives to self-generate or sign behind-the-meter deals that bypass traditional wires returns.

Consensus is probably underestimating how much this shifts bargaining power toward vertically integrated utility owners with balance-sheet flexibility. The main falsifiers are a slowdown in data-center announcements, regulators blocking above-trend returns on incremental capex, or hyperscalers choosing onsite generation at scale; any of those would compress the “AI power” premium over the next 3-12 months even if the secular story remains intact.

Key Decisions for Investors

• Go long BRK.B on weakness for a 6-12 month hold: this is a lower-volatility way to own the AI power bottleneck, with upside driven by utility rate-base expansion and optionality from Alphabet; cut the view if BHE-related capex fails to translate into authorized ROE expansion.
• Pair trade: long BRK.B / short XLU for 3-6 months. BRK has more embedded AI infrastructure optionality and balance-sheet firepower; the ETF is crowded and more exposed to rate-case disappointment if load growth does not convert into allowed returns.
• Watchlist alert on GOOGL into the next 1-2 quarters: any evidence that AI capex is starting to normalize while power spend is rising would favor a long GOOGL / short NVDA expression as the market rotates from infrastructure scarcity to monetization.
• Avoid chasing NVDA on this headline alone; the incremental signal is demand for electricity, not a new semiconductor cycle. If data-center power constraints persist into the next earnings season, that is more supportive of utilities and power-transmission names than of semis at current valuations.