Greg Abel does not expect Berkshire Hathaway to train a large language model. In an interview this week, Berkshire’s chief executive described a more mundane route into artificial intelligence, one that begins with a power plant and ends with a stake in the company that operates the world’s largest search engine.
Mr. Abel laid out the strategy in two parts. Through Berkshire Hathaway Energy, the conglomerate’s utility arm, the company wants to supply electricity to the data centers that technology firms are building to run AI workloads. The business is welcome, he said, on one condition: it should not push up electricity prices for Berkshire’s other customers.
The second leg is financial. Berkshire’s stake in Alphabet, the parent of Google, has grown to nearly $36 billion, Mr. Abel said, putting it among the conglomerate’s largest stock positions. The holding was built by Warren Buffett beginning last year, and Mr. Abel said he and Mr. Buffett reached their conclusion from the inside, watching how their own businesses were adopting the technology. Both men decided AI would reshape American commerce, he said, and that Google would be an important player in that reshaping.
The Alphabet stake fits a pattern Berkshire has followed since it first bought Apple shares in 2016. Mr. Buffett spent decades arguing that he did not understand technology companies, then built Apple into Berkshire’s largest holding before paring it in 2024 and 2025. Alphabet extends that logic to a second giant: a business with durable pricing power, a strong balance sheet and cash flows that arrive whether the economy is expanding or contracting.
The electricity side is the one Mr. Abel knows best. He built his reputation inside Berkshire running its utilities, and the energy arm he once led operates regulated power companies across the western United States, including PacifiCorp, NV Energy and MidAmerican Energy, serving millions of customers from the Pacific Northwest to the desert Southwest.
For years, the puzzle in American electricity was the opposite of today’s. Demand barely grew as efficiency improved, and utilities planned around stagnation, retiring old plants and hesitating to build new ones. The arrival of data centers reversed the arithmetic. Power demand is now the fastest-growing constraint on the AI build-out, grid operators warn of shortages, and technology companies are signing long-term contracts to secure electricity years in advance.
For a regulated utility, that demand is a double-edged opportunity. Growth without higher rates for existing customers requires the new load to pay for the infrastructure it consumes, which is exactly the condition Mr. Abel attached. Whether Berkshire can capture the business at scale will depend on regulators who must approve new plants and transmission lines, a process that runs for years and often turns into public fights over who pays for the build-out. Mr. Abel’s caveat is also the standard those regulators will apply, and it will determine how much of the data-center market Berkshire can serve.
Analysts who follow the company said the two legs of the strategy share a common shape: long-duration assets whose economics are decided by contracts and regulation rather than by the shifting tastes of consumers. Power agreements stretch across decades. Alphabet’s search business has grown through every economic cycle of the past twenty years, and its cloud unit is spending heavily to compete in the same AI infrastructure build-out that Berkshire’s utilities hope to serve.
The Alphabet bet carries complications of its own for a firm built around insurance and railroads. Alphabet faces antitrust pressure on its search business, which the U.S. government has spent years in court trying to restrict, and its AI investments are expensive. Mr. Abel did not dwell on those risks in the interview, beyond expressing confidence that Google would remain central to the AI economy.
What Berkshire is not doing may matter as much as what it is. The company has shown no interest in building its own AI products, financing its own data centers on the scale of the technology giants, or placing speculative bets on the startups of the moment. Its approach is to sell the power that makes the industry run and to own a share of the businesses that convert AI into revenue, a division of labor that fits a conglomerate whose returns have always come from patience rather than speed.
Where the arithmetic lands will show up in two places over the next few years: in the rate filings of Berkshire’s utilities, and in the quarterly disclosures of its portfolio. The data-center contracts will become public record when they are filed with regulators, and the Alphabet stake will appear in Berkshire’s reports to shareholders. Between the power lines and the search ads, Mr. Abel said, is where Berkshire expects to find its return on AI.
