Amazon and Microsoft race on $200B AI power plays

Amazon and Microsoft are racing to solve a fundamental problem in AI: power. Both companies are spending close to $200 billion（約32兆円） each to address the electricity demands of artificial intelligence. The challenge is not merely technical but physical — data centers running AI models require enormous amounts of electricity, and that demand has become a critical bottleneck limiting how fast AI infrastructure can expand. What makes this competition noteworthy is not the scale of spending alone, but the fact that the two companies are pursuing fundamentally different strategies to tackle the same crisis. The article poses the core question: which approach survives when it collides with the actual constraints and limitations of power grids in the real world.

The AI industry has hit a constraint that cannot be engineered away: power consumption. Amazon and Microsoft, the two largest cloud operators, have each committed close to $200 billion（約32兆円） to solve this problem — a signal that the electricity bottleneck is not a marginal issue but an existential one for AI scaling. However, the article indicates their strategies differ materially. This divergence suggests the industry has not yet converged on a single solution, and both companies are placing large bets on competing approaches. The ultimate test will be whether either strategy can deliver on the grid realities companies face — not just laboratory projections of demand.