Key Highlights
- Berkshire Hathaway CEO Greg Abel said he sees significant opportunity for the conglomerate's energy Business from the buildout of AI data centers, citing the large amounts of electricity required to run them.
- Berkshire added $17 billion of Alphabet shares in the second quarter, making the Google parent its third-largest Equity holding at roughly $36.6 billion.
- Abel said about 8% of Berkshire Hathaway Energy's load in Iowa came from data centers last year, underscoring energy as a potential constraint on AI growth.
- Global hyperscaler capital expenditures for 2026 are estimated at around $1 trillion, according to Goldman Sachs, though the bank says commonly cited figures may understate total global AI spending.
Abel Links Berkshire's AI Bet to Its Energy Business
Berkshire Hathaway (NYSE:BRK.B) Chief Executive Greg Abel said Wednesday he sees significant opportunities for the conglomerate's energy business stemming from the ongoing buildout of AI data centers, shortly after Berkshire made Google parent Alphabet (NASDAQ:GOOGL) its third-largest common stock holding. Abel described Alphabet as a “significant player” in artificial intelligence, a Factor that led him and Chairman Warren Buffett to authorize an additional $10 billion Investment in the company roughly three months ago.
“We are all seeing and feeling the impact” of AI, Abel said, adding that Berkshire has considerable visibility into how AI is being used and monetized across its portfolio of companies, which helped drive the firm's growing interest in Alphabet specifically.
A Rapidly Growing Position in Alphabet
Buffett originally initiated Berkshire's investment in Alphabet last year, though Abel has taken Credit for structuring the most recent addition, negotiating a 6.5% discount to Alphabet's stock price for the new investment. Abel said Alphabet had recommended Berkshire consider a $10 billion investment without initially specifying terms, after which he and Buffett settled on the size of the position and the discount before completing the transaction.
Berkshire added a total of $17 billion of Alphabet shares during the second quarter, according to the company's Q2 portfolio snapshot filed with the SEC in August, making Alphabet the third-largest holding in Berkshire's equity portfolio. As of the filing, Berkshire owned approximately 106 million shares of Alphabet's Class A and Class C stock, worth roughly $36.6 billion, positioning the holding just ahead of Coca-Cola but still behind Apple and American Express in Berkshire's portfolio. The Alphabet purchase was by far the largest addition to Berkshire's portfolio during the quarter; the conglomerate also increased its stake in Delta Air Lines (NYSE:DAL) by 44%, or roughly $1.6 billion.
Energy Seen as the Real Constraint on AI Growth
Abel said Berkshire's energy business stands to benefit from AI's continued expansion, given the substantial amount of electricity required to power data centers. He estimated that in Iowa, where Berkshire Hathaway Energy is based, data centers accounted for about 8% of the utility's total load last year. “I've sort of always had the strong view that energy would be the constraint,” Abel said, adding that Berkshire continues to see the dynamic as a significant opportunity for both the Parent Company and its energy subsidiary.
A Broader Debate Over the Scale of AI Spending
Alphabet is one of five major hyperscalers, alongside Microsoft (NASDAQ:MSFT), Meta Platforms (NASDAQ:META), Amazon (NASDAQ:AMZN), and Oracle (NYSE:ORCL), currently making enormous capital expenditures to expand computing capacity for AI workloads. Global hyperscaler capital expenditures for 2026 are estimated at around $1 trillion, according to investment bank Goldman Sachs, though the bank has noted significant uncertainty within that broader estimate. In an August research note, Goldman Sachs researchers wrote that the commonly cited hyperscaler capex forecast of $794 billion likely understates total global AI Capital Expenditure by roughly $200 billion, even as that same figure likely overstates the specific amount of AI investment occurring within the US by a similar margin.
The central question facing investors, according to Goldman Sachs, is the extent to which these substantial investments will ultimately generate returns, and over what time horizon. AI computing workloads are generally divided between algorithmic Training on large datasets and inference, or responding to real-time queries. In a separate May research note, Goldman Sachs researchers wrote that a prolonged, training-dominant phase of the AI buildout extends the overall return-on-investment timeline, since capital expenditure and research spending continue to be deployed well in advance of broad-based monetization.






Please wait processing your request...