Ex-Google CEO Eric Schmidt says AI may hit a money wall before it hits a power wall. The real limit is cash, not energy. At $50 billion per gigawatt, 10 gigawatts costs half a trillion dollars. Very few companies or countries can hand the industry $1 trillion in capital — China could certainly do it, and America's capital markets allow borrowing at that scale. Europe cannot, which they are sore about.
The deeper implication: AI capex is inflationary before it becomes disinflationary. It demands power, copper, transformers, gas turbines, land, construction labor, water, chips, debt, and grid upgrades now. That pushes costs up immediately. The productivity benefit arrives later. Hyperscalers are spending hundreds of billions because intelligence production becomes one of the dominant economic control points of the century. The payback comes from making cognition cheaper, faster, more scalable, and less dependent on human headcount. Higher rates do not stop the AI buildout — they concentrate it. The user sees magic. The owner controls the factory. America can financialize the future faster than anyone. China can force mobilization without waiting for private-market proof. Europe lacks both unified capital aggression and state mobilization speed, risking becoming a governed market for other people's intelligence infrastructure.
The deeper implication: AI capex is inflationary before it becomes disinflationary. It demands power, copper, transformers, gas turbines, land, construction labor, water, chips, debt, and grid upgrades now. That pushes costs up immediately. The productivity benefit arrives later.
This is exactly the same with any new industry. Be it rail roads, shipping, Oil, whatever. But we dont need to invest half a trillion dollars for 10 gigawatts in one go with no returns. One would expect this would happen incrementally as the industry is already primed and set to consume whatever new productivity hits the market.
Ex-Google CEO Eric Schmidt says AI may hit a money wall before it hits a power wall. The real limit is cash, not energy. At $50 billion per gigawatt, 10 gigawatts costs half a trillion dollars. Very few companies or countries can hand the industry $1 trillion in capital — China could certainly do it, and America's capital markets allow borrowing at that scale. Europe cannot, which they are sore about.
The deeper implication: AI capex is inflationary before it becomes disinflationary. It demands power, copper, transformers, gas turbines, land, construction labor, water, chips, debt, and grid upgrades now. That pushes costs up immediately. The productivity benefit arrives later. Hyperscalers are spending hundreds of billions because intelligence production becomes one of the dominant economic control points of the century. The payback comes from making cognition cheaper, faster, more scalable, and less dependent on human headcount. Higher rates do not stop the AI buildout — they concentrate it. The user sees magic. The owner controls the factory. America can financialize the future faster than anyone. China can force mobilization without waiting for private-market proof. Europe lacks both unified capital aggression and state mobilization speed, risking becoming a governed market for other people's intelligence infrastructure.
SOURCE: https://x.com/The_Prophet_/status/2055154486373658880
This is exactly the same with any new industry. Be it rail roads, shipping, Oil, whatever. But we dont need to invest half a trillion dollars for 10 gigawatts in one go with no returns. One would expect this would happen incrementally as the industry is already primed and set to consume whatever new productivity hits the market.