For decades, computing improved almost automatically: transistors shrank, chips doubled in density, and machines that once filled rooms ended up on wrists. That easy descent is now stalling. The trends that tracked those gains—Moore's Law, Dennard's Law, Koomey's Law, Kryder's Law, and Keck's Law—are bending or breaking. In response, the biggest hyperscalers are pouring more than $700 billion this year into data centers, buying performance by brute force with concrete, copper, and silicon faster than the electrical grid can feed them.
The slowdown is real, but it is not a wall. These scaling laws are observations about engineered systems, not fundamental physical laws. The true limits set by quantum mechanics, general relativity, and information theory lie far below anything in today's data centers. Reaching toward those limits can pay off twice: once in new technology that forces stalled paradigms to give way, and once in deeper understanding of the physics themselves.
Richard Feynman's 1959 lecture, "There's Plenty of Room at the Bottom," made the same point. His first prize, for a tiny electric motor, was won within a year by hand assembly under a microscope—no new technology, just patience. The second, for shrinking a page of text enough to fit the Encyclopædia Britannica on a pinhead, took 25 years and a new tool: electron-beam lithography. Both outcomes show that the room at the bottom is still there, even when the easy path runs out.