Every Single Indicator

The map of where the US can build data centers is shrinking, but not gone yet. everysingleindicator.com/blo…

The amazing thing about Google getting hammered today is not that its core Search business is under stress. It’s how long it took investors to figure that out. AI is changing the regime under which the hyperscalers became some of the highest-return-on-capital businesses in human history. Google & Meta, and to a lesser extent Amazon, are at the centre of this disruption. They became the modern-day landlords of the digital economy. Google auctioned search intent. Meta auctioned attention. Amazon auctioned product discovery. They controlled access to consumers and charged advertisers increasingly expensive rents for the privilege. Beautiful low capital intensive businesses. Until the rules change. AI is doing precisely that,faster than the internet revolution that created these empires. I barely use a web browser for search anymore. It has degenerated into a window for opening websites. Why scroll through sponsored links & SEO garbage for dog food when an AI assistant can find what I need? I doubt I am alone. Their response? Join the party and build staggering amounts of compute. Google alone is heading towards US$200bn of capex in 2026, while its Search growth is already slowing. Clayton Christensen explained the problem in Innovator’s Dilemma. Incumbents struggle when technological disruption destroys the economics of the very businesses that made them successful. But throwing trillions at the disruptive technology won’t make another high ROIC units. Instead it renders the entire shop average. Worse, AI is on track to become commoditised, with economics eventually resembling airline seats: enormous fixed costs, relentless capacity additions, falling unit prices and fierce competition. Models become cheaper, more interchangeable and increasingly tailored to individual needs. Much of that intelligence will eventually run locally on personal devices. Who pays the landlord when the tenants no longer need his building? In 2025, Google generated US$225bn from Search, Meta US$196bn from advertising. How much of that can they defend, let alone replace, while committing hundreds of billions to a data centre arms race? It’s just ridiculous. In 2 years time they can buy these zero-moat compute factories for 20 cent on the $. The problem is the moat. I’d argue that 98% of businesses have no durable moat. Schumpeter’s creative destruction eventually erodes any competitive advantages in a free enterprise system, growth fades and returns on capital gravitate towards 5–7% as competition brings supply into balance with demand. Bruce Greenwald explained in Competition Demystified just how hard it is to build a economic moat. My favourite is what Buffett calls share of mind. Try displacing Coca-Cola, Pampers or WD-40 with a 5% discount. Good luck. Where is AI’s moat? Switching costs are near zero. I switch bw 4-6 models and already use 2 specialised AIs for commodities, each offering proprietary data and knowledge depth. If at all, I see opportunity for high ROICs in owning scarce, valuable info & domain expertise rather than another tech layer. But that is hardly the 10 trn-dollar TAM that spoiled Google & Meta executives want to present on their next earnings calls. They think unrealistically big, with precious little capital discipline. Instead of accepting that the regime change, rest assured that they will throw endless good money after bad. Like 100% sure! History offers zero comfort for that process. Look at the former kings of the S&P 500: AT&T, GM, DuPont, IBM, Exxon, Kodak, Sears, Xerox, GE, Philip Morris, Pfizer, Citigroup, Intel or PanAm. Some went bust, others lost dominance; few retained their extraordinary economics. Expect Google to join the decay list.