Energy Ceiling: S&P Global Estimates Big Tech AI CAPEX at $635 Billion

Physics continues to dictate the terms of the AI industry. On March 31, 2026, analytical agency S&P Global published a macroeconomic report that brings together the infrastructure issues of recent weeks. Big Tech's capital expenditures (CAPEX) will reach a colossal $635 billion, but this growth is hitting a severe "energy shock."

According to analysts, spending solely on data center construction will jump to $280 billion in 2026 and $330 billion in 2027. The problem is that the national power grids of developed countries are technically unprepared for such a load. This explains the market turbulence: from US bills blocking new AI clusters to investments in highly efficient chips (Google's TurboQuant, new NPUs from Rebellions). The era of mindless extensive model scaling is over—the battle for every kilowatt and lowering PUE (Power Usage Effectiveness) has begun.

Source: S&P Global / Reuters