Google's parent company, Alphabet, announced on Monday it will raise $80 billion by selling its own stock to investors. The stated purpose is straightforward: build more data centers, buy more computing hardware, and expand the capacity needed to run its AI products and services. The company said demand from businesses and consumers is already outpacing what it can supply.
The raise is structured in three parts. About $30 billion will come from a public stock offering. Another $40 billion will come from a slower, ongoing sale of shares tied to how Alphabet pays its employees. And $10 billion is a direct private investment from Berkshire Hathaway, the conglomerate now led by Greg Abel after Warren Buffett stepped down as CEO at the end of 2025.
The Berkshire piece deserves attention on its own. Abel's firm had already tripled its Alphabet stake last quarter, making it one of Berkshire's five largest equity positions. This latest $10 billion tops that up further, at a slight discount to the market price. For decades, Berkshire stayed well away from complex technology companies. That is changing. Abel is signaling, with real money, that he sees Alphabet's AI business as something with durable, long-term returns.
The scale of Google's spending plan is worth sitting with. Alphabet spent around $52 billion on infrastructure in 2024. In 2025, that jumped to over $91 billion. The current plan for 2026 is between $180 billion and $190 billion, essentially doubling year over year. This is not a project with a clear finish line.
Alphabet is not alone. The five largest cloud and tech companies, including Amazon, Microsoft, and Meta, plan to spend somewhere between $660 billion and $690 billion on AI infrastructure in 2026 combined. This level of spending, year after year, has to be paid for somehow. Traditionally these companies used their own cash or took on debt. Selling equity is a different move: it dilutes existing shareholders, but it avoids loading the balance sheet with debt at a time when interest rates remain elevated.
For businesses that use AI services as customers, the picture is worth monitoring closely. These companies are building this infrastructure because demand exceeds supply right now. That is good news in one sense: the tools are being used. But the cost of running AI-heavy workloads is also rising unpredictably. One recent case saw an anonymous enterprise rack up roughly $500 million in AI tool charges in a single month after no one set a spending cap. Uber burned through its entire 2026 AI budget in four months.
The pattern here is not that AI tools are a bad investment. The pattern is that the costs are hard to forecast, and most finance teams do not yet have a system for tracking them properly. As the infrastructure arms race continues, the companies building the pipes will get their money back from the businesses using them. For any operator who has AI tools deployed at scale, building a clear view of what you are spending and what you are getting back is no longer optional.