So I've been digging into what Buffett's actually doing with his AI bets, and it's pretty interesting when you look at the actual numbers. Everyone assumes the Oracle of Omaha avoids trendy tech stuff, but here's the thing - roughly 23% of his $317 billion stock portfolio is tied up in three companies that are absolutely crushing it with AI. These aren't random picks either.

Let me start with the elephant in the room: Apple. This is massive. Berkshire's got about 20.6% of its entire portfolio in Apple stock - we're talking hundreds of billions here. Now Buffett did trim the position significantly, selling over 70% of what they used to hold, but they're clearly still believers. And honestly, when you think about it, Apple's positioned better than almost anyone for consumer-level AI. They've got 2.35 billion active devices worldwide, all running their own chips that are increasingly optimized for AI workloads. The battery efficiency angle is something people sleep on. Every iPhone, iPad, and Mac is getting smarter without turning into a power drain. That's a real competitive advantage. The new iPhone 17 lineup is apparently driving upgrade cycles way harder than Wall Street expected, so there's probably more runway here.

Then there's Alphabet. Only 1.7% of the portfolio, but the timing is telling - Berkshire just loaded up in Q3 2025. Google Search looked vulnerable when ChatGPT started blowing up, right? Everyone thought AI chatbots would cannibalize search traffic. But Alphabet moved fast. They integrated AI directly into Google Search with features like AI Overviews, basically making the search experience smarter without losing the monetization. Google Cloud is another angle - they're competing head-to-head with AWS, selling data center access and their own Tensor chips as alternatives to Nvidia. That $155 billion order backlog isn't a joke. The stock's up 62% this year alone, and if you're looking for best AI shares to buy with real infrastructure behind them, Alphabet's doing the work.

Amazon rounds out the trio at 0.7% - smaller weighting but strategically important. AWS is the real play here. They've got this $200 billion backlog of customers waiting for data center capacity. Amazon's spending $125 billion this year just to build infrastructure to meet demand. Here's what's clever though: they're not just buying Nvidia chips. They designed their own - Trainium2 - which is 40% cheaper for training AI models. Anthropic is already using 500,000 of these chips for Claude. That vertical integration matters. Buffett's admitted he wishes he'd spotted this opportunity earlier, but the stock's more than doubled since they got in, and AI infrastructure demand is nowhere near peaked.

The broader pattern here is that Buffett isn't chasing AI hype for hype's sake. He's backing companies that already had real businesses - e-commerce, search, consumer devices - and are now using AI to unlock new efficiency and new revenue streams. These aren't speculative plays. They're established companies with moats, using AI as a competitive multiplier. If you're thinking about best AI shares to buy right now, this is basically the Buffett framework: look for companies with existing scale, real infrastructure, and a clear path to monetizing AI. Not the flashy startups, but the companies that already print money and are about to print more. That's where the real returns usually hide.