Berkshire Hathaway reported a record $397.4 billion cash position at Saturday's annual meeting, up from $381.6 billion at year-end. First-quarter net income more than doubled to $10.1 billion. Buybacks for the quarter ran $234 million — all of it executed in March, the first repurchases since May 2024. Greg Abel, in his first Q&A as chief executive, told the floor he was "not going to do AI for the sake of AI." [1][2]

The numbers are the position. The line is the policy.

The May 2 paper framed Abel's first Saturday as the public catechesis the conglomerate had been holding in private, and carried the print as a continuity event. The Saturday session resolved the catechesis question in one direction. Abel did not pivot. He hoarded.

The Cash Number

A $397.4 billion cash and Treasury position is the largest dry powder in S&P 500 history. The increase from $381.6 billion at year-end, against a more-than-doubling of operating earnings, is not the math of a buyer. It is the math of a seller waiting for prices to come to him. Berkshire was a net seller of equities in the quarter, offloading $24.1 billion in stock against $16 billion in purchases. Insurance underwriting profit rose 28.5% to $1.72 billion. Operating earnings climbed 18% to $11.35 billion. [3][4]

The $234 million in March-only buybacks is the corollary. Abel views the stock at roughly 1.5 times book as either fair or cheap — he bought a small slug and stopped. The decision-rule reads: the alternatives are worse. Equities at current multiples are worse. Private deals at current multiples are worse. Treasury bills at 4.0% are not worse. So Abel parks.

The AI Refusal

The Q&A line came in response to a shareholder asking whether Berkshire would deploy capital into the AI build-out. Abel's answer ran past the literal question. He said Berkshire is "not going to do AI for the sake of AI" — and then added that the firm "doesn't mean you need to deploy all your capital." The first clause is a sector refusal. The second is a methodology refusal. Together they reject the cohort's working assumption that a hundred-billion-dollar capital base requires a hundred-billion-dollar deployment. [1][2]

Set the line against the cohort. Apple authorized $100 billion in fresh buybacks last week, with first-half FY26 capex at roughly $4.3 billion and full-year guide near $13 billion. [5] Microsoft's FY26 capex track is $190 billion. Google's is $185 billion. Meta priced $25 billion in bonds at four-times oversubscription against a raised $125-145 billion capex envelope. Amazon printed $44.2 billion in Q1 capex, with FY26 guide near $200 billion and free cash flow collapsed to $1.2 billion from $26 billion. The Magnificent Four AI hyperscalers project roughly $715 billion in 2026 capex, against $410 billion in 2025. [6]

Two firms refused. Berkshire chose cash. Apple chose return. The cohort splits 5-to-2 in favor of capex. The 2-to-5 minority controls more equity-yield optionality than the majority. Abel's no-AI-for-the-sake-of-AI is the explicit version of the implicit Apple line — the Cupertino buyback was the action, the Omaha sentence was the principle.

What the Cohort Is Actually Pricing

The bond market and the equity market are pricing the same balance sheets incompatibly. Meta's $25 billion bond issue cleared 4x oversubscribed at the same time the equity dropped 9.5% on the print. [7] The bond market funds the capex. The equity market punishes the suspended buyback. Inside that disagreement, Berkshire's $397.4 billion cash and Apple's $100 billion buyback are the two cohort votes that say: the equity-yield trade beats the capex trade.

The Microsoft FQ2 10-Q already disclosed that 45% of its $625 billion commercial RPO — roughly $281 billion — is tied to OpenAI. Cerebras opens its $40 billion roadshow on Monday with OpenAI named as customer ($10 billion-plus multi-year), lender ($1 billion), and warrant-shareholder, all on one S-1 page. Three counterparty disclosures inside one editorial week. The capex regime is now also a counterparty-concentration regime. The Berkshire cash hoard is the counterparty-free hedge against it.

The Saturday Was the Sentence

Buffett, 95, sat as chairman emeritus. Abel ran the floor. Annual-meeting attendance fell sharply without Buffett as CEO — Fortune called it the lowest in years. [4] The continuity ritual lost some of its tourist economy, but the substantive transition held: Abel did not soften the cash position, did not signal a major buyback expansion, did not name a target acquisition. He named the principle. He named the refusal.

The bank-war-economy thread now reads: Apple buyback at $100 billion, Berkshire cash at $397.4 billion, Meta bonds at $25 billion, Amazon capex at $44.2 billion in one quarter, Microsoft RPO at $625 billion with OpenAI inside half of it. Five reference points, one cohort, one editorial week. The cohort split is no longer a forecast. The Sunday newspaper has it as a print.

The Saudi PIF rebalance — LIV wind-down on one side, Paramount-WBD entry on the other, with 38.5% Middle East non-voting equity in the post-merger entity — is the same balance-sheet move from the sovereign-wealth side. Capital is reallocating away from the AI capex regime and toward equity-yield assets, IP catalogs, and dry powder. Abel's Saturday was the cleanest version of that reallocation. The Cash Fortress reached new heights because Abel decided that nothing on offer was worth a brick from it.

The Q&A is over. The number stays.

-- THEO KAPLAN, San Francisco