At the end of last year, Warren Buffett ended his remarkable career as the CEO of Berkshire Hathaway at the age of 95. Q4’s buying and selling decisions, therefore, are the last ones at Berkshire in which Buffett can be assumed to have had a direct hand. Among the more interesting transactions at Berkshire Hathaway last quarter was the decision to sell more than three-quarters of the conglomerate’s stake in Amazon. Why did Buffett sell Amazon, and does Berkshire’s decision to pare back its AMZN holdings raise red flags for other investors
The Possible Role of Amazon’s AI CapEx
One possible reason for the decision to sell Amazon could be its massive allocation of cash to building infrastructure for AI. While Amazon hadn’t publicly announced its plans to invest around $200 billion on AI in 2026 by Q4, rising CapEx numbers could have started to worry the team at Berkshire Hathaway even before Amazon revealed its 2026 spending plans. The rapidly rising spending was readily apparent when Q3’s earnings report came out, revealing trailing 12-month spending of over $120 billion on purchases of property and equipment.
Buffett has always prioritized sensible, disciplined capital allocation at the businesses he owns, as long-term success depends on earning returns by deploying the cash that businesses produce. In the case of Amazon’s rampant AI spending, however, Buffett and his investment team may have seen a deployment of massive amounts of useful cash to produce uncertain and possibly distant future returns.
Buffett may also have locked onto fundamental challenges in the economics of AI data centers, namely that they depreciate quickly and have a very limited window of time in which to generate profits. Right now, companies like Amazon are putting huge amounts of capital to work building data centers that may or may not generate positive returns and which will create ongoing expenses as the need to periodically update and replace equipment arises.
It’s worth noting, however, that Berkshire did invest in another business with high AI CapEx last year, namely Alphabet. Buffett’s decision to buy Alphabet, however, appears to have been driven more by its massive moat in online search and its almost unparalleled cash flows than by conviction on its AI technology. Because AI integrates directly with its search business, Alphabet may also be in a better position to profit from the technology than Amazon is.
Has Berkshire Put AI in the “Too Hard Pile”?
Despite being arguably the most successful investor of all time, Buffett has tended to avoid investments he sees as too complex or requiring too much specialized knowledge, putting them in what he and his late partner Charlie Munger often called the “too hard pile.” This simple approach keeps Berkshire from making mistakes associated with getting into overly complex investments, and it’s possible that Buffett has put investing in AI businesses into that pile.
Right now, both the promises and pitfalls of AI are incredibly uncertain. On the one hand, the technology has shown remarkable potential in areas ranging from marketing to medicine, and the tech giants of the world increasingly seem to believe that it will fundamentally disrupt nearly every aspect of the global economy. Recently, fundamental assumptions about some of the largest SaaS businesses have been challenged by the spread of AI tools, leading to the evaporation of over $1 trillion in market value among firms that previously were assumed to have impenetrable moats.
Coming back down to earth, however, a recent survey of 6,000 CEOs from the US, UK, Australia and Germany found that an astonishing 90 percent saw essentially no impact on productivity from AI. This finding mirrors previous research that has called into question how much return is really being generated by AI. Stacked on top of this seeming paradox between potential and real-world results are the economic challenges of data centers mentioned above and a vast set of risky intertwined agreements between top AI companies, including Amazon.
Against this backdrop, it would be far from uncharacteristic for Buffett to decide that predicting winners in the AI race is simply too difficult and too risky. Buffett has typically avoided tech investments, though he has put money into Apple and Alphabet due to their massive moats and widespread use among consumers. With Amazon increasingly turning toward AI as a growth driver, the stock may simply have become too complex for Buffett’s liking.
Why Did Buffett Sell Amazon?
Although it’s impossible to say with certainty why Warren Buffett and his team decided to sell Amazon, the facts above may offer some clues. With Amazon diverting gigantic amounts of its cash to AI infrastructure, the business is increasingly looking like more of a play on AI than on eCommerce. The decision to allocate so much capital to AI may have been seen by Buffett as both a questionable use of cash and a fundamental shift in Amazon’s direction as a business.
While Warren Buffett has always been reluctant to sell stocks, he has also been fairly transparent about the criteria he uses when making selling decisions. One of the core drivers of these decisions tends to be re-evaluation of the business’s long-term competitive advantages and economic characteristics. In Amazon’s case, Buffett and his team may have initially seen it in a positive light due to its incredible moat in consumer eCommerce and strong position in cloud computing. More recent events, however, have transformed owning Amazon stock into a decision about AI. With this and Buffett’s general aversion to overly complex investments in mind, it’s not hard to see Berkshire’s selling of Amazon as a rethinking of previous assumptions about the business.
It’s important to note, however, that Buffett’s decision to sell Amazon doesn’t inherently mean that the stock is a bad one to own. Buffett himself has observed that many of the stocks he has sold over his career have continued to rise and were still shares in strong businesses. While Amazon may not meet Berkshire Hathaway’s investment criteria at the moment, the company itself could still have a long growth runway ahead of it.
The author has no position in any of the stocks mentioned. Financhill has a disclosure policy. This post may contain affiliate links or links from our sponsors.