Hyperscalers on Track to Spend $700 Billion This Year as Chip Scaling Laws Stall, MIT Technology Review Says

The article describes that spending as performance bought by brute force: pouring concrete, copper and silicon into new data-center sites faster than the electrical grid can supply them, and adding more of what the industry already knows how to manufacture. The other kind of progress, the kind that made computer chips better for decades, is quietly running out, it says.

MIT Technology Review lists five trend lines that tracked the 20th-century information age and are now bending or breaking. Moore's Law, which plotted how densely transistors can be packed onto silicon, has slowed. Dennard's Law, under which power consumption for a given patch of chip stayed the same as transistors shrank, has ended. Koomey's Law, which traced how much computation can be wrung from each joule of energy, has begun to flag. Kryder's Law, describing how tightly bits can be packed in magnetic storage, has slowed to a crawl, and Keck's Law, charting the rise in the information a single conventional optical fiber can carry, is flattening.

The explainer traces the argument to a lecture Richard Feynman gave on the evening of Dec. 29, 1959, at an American Physical Society meeting at Caltech. Titled "There's Plenty of Room at the Bottom," it argued that no physical law prevented writing the Encyclopædia Britannica on the head of a pin, storing every book ever written in something the size of a dust speck, or ultimately arranging matter atom by atom; biology already keeps a living cell's instructions in a wisp of DNA. The tools simply had not been built yet.

Feynman offered two $1,000 prizes from his own pocket. The first went to William McLellan, a Caltech-trained engineer who turned up about a year later with a working electric motor smaller than a 0.4-millimeter cube, a quarter-milligram machine roughly the size of a grain of sand assembled by hand with a toothpick and a watchmaker's lathe under a microscope. By his own account, the article says, Feynman was only slightly disappointed: the challenge had proved his point in the least interesting way imaginable, without requiring any new technology, just an unusually patient man with steady hands.

The second prize, for shrinking a page of text by a factor of 25,000 in each direction so the whole Britannica would fit on a pinhead, went unclaimed for a quarter of a century. Tom Newman, a Stanford graduate student, won it in 1985 by using an electron-beam lithography machine, an ancestor of the patterning tools used in chipmaking today, to transcribe the opening page of A Tale of Two Cities at the required scale.

The article argues that the scaling laws are observations about systems engineers have built rather than fundamental physics. The limits set by quantum mechanics, general relativity and information theory lie far below anything in a data center today. Some may be impractical to reach, but the article says the work pays twice: in technology development that forces stalled paradigms to give way, and in experiments that deepen understanding of the physical laws themselves.