Berkshire Hathaway tripled the size of its Google parent stake in Greg Abel's first quarter as CEO, building a $15.6 billion position in an AI hyperscaler that is now one of the company's five largest holdings - then Abel went to the 2026 annual meeting and told shareholders that AI "in no way can replace humans."
That is not a contradiction I would miss. The portfolio says one thing. The philosophy says another. And the $397 billion cash pile on the balance sheet says something else entirely.
The competitor headline on this story is: "This isn't Warren Buffett's Berkshire anymore." It isn't. But the question that actually matters for capital allocation is not whether the brand feels different. It is whether the money has moved to the right side of the AI infrastructure cycle - or whether Berkshire is still sitting on the bench while the market has already entered the second half.
The Portfolio Told You What the Speech Didn't
Abel took the CEO title on January 1, 2026. By the end of March, the public stock portfolio had been reshaped from 40 positions down to 26. Sixteen stocks were sold entirely. The portfolio shrank from roughly $274 billion to $263 billion. He was a net seller.
But the money he deployed tells the story the AGM speechbook couldn't.
Alphabet - Google's parent - was the single biggest addition. Berkshire initiated the position in Q3 2025 under Buffett, buying about 17.85 million shares worth roughly $4.93 billion. Abel then tripled that stake in one quarter, adding roughly 36 million more shares to bring the total to approximately 54 million shares worth $15.6 billion, or 5.9% of the portfolio. That is not a diversification sprinkle. That is a conviction position.
And the conviction is in a company that just reported 63% cloud growth and raised its annual AI capex guidance to between $180 billion and $190 billion. Alphabet is an AI infrastructure player first and an advertising company second - at least in the way the market is beginning to price it.
Put plainly: Abel bet $11 billion of new capital in one quarter on one of the companies building the backbone of the next AI infrastructure cycle. Then he went to Omaha and said Berkshire would adopt AI "judiciously" and only where it adds clear value.
I am not asking for consistency between operating philosophy and investment philosophy. Those can and should be different. But when a $1.05 trillion company makes an AI hyperscaler a top-five holding while its CEO tells shareholders not to expect much from the AI revolution, I want to understand which signal carries more weight.
What Changed From Buffett to Abel - and What Didn't
The headline story about the transition is nostalgia dressed as analysis. Yes, Buffett is gone from the CEO role. Yes, 16 stocks were dumped. Yes, the AGM was quieter, less theatrical, less Oracle-of-Omaha.
But let's look at what actually changed in the substance of capital deployment.
What changed: Abel is not afraid to add concentrated positions in companies Buffett wouldn't have touched. The Alphabet bet is the cleanest example. Buffett was a lifelong skeptic of companies he couldn't fully understand - and by his own admission, he couldn't fully understand tech architecture. Abel, who came up running Berkshire Energy, has always operated more like a CEO who understands infrastructure and capital intensity. A company spending $180 billion on AI infrastructure is, in his language, an infrastructure story. That matters.
What changed: the Apple sell-off accelerated under Abel's watch, but the thesis didn't flip. Apple remains the largest holding at roughly $49 billion - about 24% of the portfolio - down from roughly 50% at its peak. The trimming is continuation, not reversal. Buffett sold Apple for 13 consecutive quarters before he handed over. Abel is continuing the exit, not initiating one.
What didn't change: the cash pile. It hit $397 billion at the end of Q1. That is not a Buffett holdover - that is an Abel decision. He resumed the share buyback for $234 million in March, the first in 21 months, while the cash pile grew. The signal is clear: he sees few deals at the right price. The market is expensive, he has nothing to do, and he is not going to force it.
The AI Infrastructure Question Nobody at Omaha is Asking
Here is where the story gets interesting for someone who thinks about where AI capital is flowing and where it isn't.
Alphabet's AI capex is projected at $180–190 billion this year. That is money flowing into Nvidia data center GPUs, network switches, memory, power infrastructure, and the entire supply chain around building compute. Berkshire just made a $15.6 billion bet on a company that is one of the primary channels for that capital deployment.
But Berkshire itself is not building AI infrastructure. It is not buying chips, it is not deploying GPU clusters, it is not making the kind of capex commitment that defines who is "on" the AI transition. It is holding Alphabet shares while holding $397 billion in Treasuries.
This is what separates Berkshire from the AI trade. The company is not missing the AI cycle because it doesn't understand it. It is positioned one layer removed - as an owner of a major AI capex player, not as a participant in the AI infrastructure build. That is a deliberate choice. It carries less upside but also carries far less execution risk.
Abel said at the AGM that AI offers productivity gains but "in no way can replace humans". That is the language of a company that sees AI as an operational tool, not a structural transformation. I can respect that. It's a defensible operating stance for a conglomerate of Berkshire's age and employee base. But it is not the language of an investor who believes AI will reshape the global economy the way I believe it will.
The question is whether that gap - between an operating philosophy that treats AI as incremental and an investment thesis that should treat it as structural - is a feature or a bug.
I believe it is a feature. Berkshire does not need to be an AI company. It needs to own companies that benefit from AI capital expenditure without bearing the binary risk of being right about which architecture wins. Alphabet, as a hyperscaler with the balance sheet to sustain a multi-year capex cycle, fits that role. It is a leveraged bet on AI spending growth without a bet on any single chip vendor or software stack.
What This Means for the Investor
Berkshire trades at roughly $483 per share for Class B, with a market capitalization around $1.05 trillion. The trailing P/E is roughly 14.5x - that is not cheap for a company sitting on $397 billion in cash earning 4–5% in Treasuries, but it is not expensive for a diversified conglomerate whose operating profit engine generated $11.35 billion in Q1 alone, backed by insurance float - unpaid claims that serve as cheap capital - of roughly $177 billion.
The real question for me is not whether Berkshire is still Berkshire. It is whether the new configuration - heavier in AI hyperscaler exposure, lighter in concentrated consumer tech, sitting on a mountain of cash with a disciplined approach to deployment - is the right positioning for the next three years.
I believe Abel has made the right structural move by increasing exposure to AI capex players without pretending Berkshire can compete inside the AI build. The Alphabet position is a smart bet on demand that doesn't require calling which accelerator vendor wins. It is exposure to the infrastructure cycle through the buyers, not the builders.
However, the $397 billion cash pile is a warning signal of a different kind. It tells me Abel sees very few deployment opportunities at current valuations. When the largest private capital allocator in the world has nearly $400 billion sitting in short-term Treasuries, the implicit signal is that most of the market is overpriced - or that the kind of transformative deals Berkshire needs are simply not available.
Demand for AI compute is not the issue. The issue is whether there is enough capital deployment bandwidth inside Berkshire's existing framework to capture returns that justify the allocation.
The debate is not about whether Berkshire stays important. It is about whether the Abel-era capital allocation framework - more concentrated, more infrastructure-adjacent, more cautious on AI as a standalone theme - delivers returns that compete with what's available elsewhere in the AI trade. I believe it will, but gradually