Greg Abel's first annual meeting as Berkshire Hathaway's CEO produced a sentence the market wrote down. "We're not going to do AI for the sake of AI," Abel told shareholders in Omaha on May 2. [1] [2] The line was paired with another: any AI deployment must be "additive to our businesses," with focus on use cases that produce value, while acknowledging potential risks for "humanity." [1] What the line said operationally is what Berkshire's 14th consecutive quarter of net stock selling is doing: the conglomerate sold $24.1 billion of equities in Q1 against $15.9 billion bought, leaving cash near $397 billion. [3] The discipline is not narrative; it is a pattern.
Apple authorized an additional $100 billion in share repurchases on April 30, alongside Q2 fiscal results that beat across the board. [4] [5] CFO Kevan Parekh said the company is "no longer providing net cash neutral as a formal target" — a signal that incoming CEO John Ternus, taking over September 1 from Tim Cook, will run the cash differently than the prior cap. [6] The dividend was raised 4 percent to 27 cents per share. [4] The buyback record is May 11; pay is May 14. The combined Apple capital return — $100 billion authorized buyback plus the dividend raise — into the same week Cerebras's IPO got cut from $40 billion to $26.6 billion is the disciplined cohort's most direct receipt yet. Cerebras Day Three was the documentary instance of the deployment cohort recalibrating; the disciplined cohort wrote checks the same week.
Tuesday's Berkshire $397B cash and the only anti-AI-stock frame read the post-Q&A repricing as the cohort split's confirmation. Tuesday's Apple executes the $100B buyback into a tape that just halved Cerebras named the same-week alignment. Wednesday's frame is the cohort's structural fact: 5:2 deployment-vs-discipline (AMD, Cerebras, Meta, Microsoft, Amazon on capex; Apple, Berkshire on capital return) is operating, not theoretical.
Berkshire's portfolio composition supports the discipline narrative. The top five holdings — American Express, Apple, Bank of America, Coca-Cola, Chevron — represented 61 percent of the equity portfolio's aggregate fair value at the end of March. [1] The Apple holding alone made Buffett's most successful single bet of the post-2010 cycle. Buffett's praise for Tim Cook — "How would you like to step into the shoes of Steve and come through his record? It's one of the miracles of American business management" — was the moment he framed the Apple position as the exception that proved Berkshire's tech-skepticism rule. [1] Abel said Berkshire is exploring AI tools at BNSF Railway; the framing was operational, not investment.
The growth story Abel disclosed Saturday was unrelated to AI capex but adjacent to it. Berkshire Hathaway Energy expects data-center-driven power demand to rise by up to 50 percent over the next five years, with current data-center load already approximately 8 percent of peak in territories like Iowa. [7] "If I look at our peak load, it's at 8 percent. People in the industry want to get to 5 to 10 percent, and we're already at 8, so we see opportunities to grow that by 50 percent over the next five years or potentially more," Abel said. [7] The Berkshire framing: hyperscalers must bear the full cost of the new load. The utility company sells to the deployment cohort while the parent doesn't buy the deployment cohort's stocks.
Apple's Q2 results — revenue $111.18 billion (+17 percent YoY), EPS $2.01 vs. $1.95 estimate, services revenue $30.98 billion (+16.3 percent), MacBook Neo demand exceeding supply — are the kind of operating leverage that makes a $100 billion buyback authorization look like a cash-flow problem rather than a capital-allocation problem. [4] [8] Operating cash flow exceeded $82 billion in the first half of fiscal 2026; capex was $4.3 billion. [9] The remainder is discretionary capital return. The math forces the choice rather than allows it.
What the disciplined cohort is rejecting is the regime, not the technology. Abel articulated that distinction Saturday in plain language: AI must be "additive," and Berkshire's investment exposure to AI is principally through the Apple position and a smaller Google holding. [1] The $397 billion cash pile is the structural alternative the deployment cohort cannot match. The choice is not whether to participate; it is whether to deploy at the prevailing valuations the deployment cohort's IPO calendar implies. Cerebras's $40 billion-to-$26.6 billion cut is what the deployment cohort's pricing test produced this week. Apple's $100 billion is what the disciplined cohort's pricing test produced this week. The split is now operating fact.
-- THEO KAPLAN, San Francisco