We were promised that intelligence would become weightless and nearly free. The opposite is happening. Artificial intelligence is turning out to be the most physical industry in a generation: an enterprise of power plants, substations, cooling water, and land, governed by the same hard laws as any factory ever built.The Machine That Ate Power follows the wire. It begins with a single query that feels like nothing and traces it back through the model, the chip, and the server hall to where it actually ends, at a power line. Along the way it surfaces the fact that reframes the entire AI economy: there is a thermodynamic floor beneath computation, and we are still a billion times above it. That one number explains both the runway ahead and the wall waiting at the end of it.From there the argument builds, chapter by chapter. Why intelligence carries a metered bill that never stops. Why the scarcest input has quietly shifted from chips to megawatts, so that time-to-power has become the whole game. Why a single AI rack now draws more electricity than a shopping mall, why heat is the second wall of physics, and why nations are starting to treat intelligence as infrastructure to be owned rather than software to be regulated. It ends where the money does: electrons turning into tokens, and tokens into equity.Written by a Ph.D. engineer who has spent his career at the unglamorous end of technology, in power systems, electronics, and heavy infrastructure, this is not another book about algorithms. It is about the ground the algorithms stand on. Every chapter opens somewhere you can stand: a windswept Texas campus, a substation yard, a server hall at 2 a.m., and reasons from that place to the idea.For investors, operators, policymakers, and anyone trying to see past the hype, the book offers a durable lens. Understand the physics and the economics, and you will move through the coming decade far more surely than those dazzled by the surface.