In April 1965, Gordon Moore, then director of research at Fairchild Semiconductor, published a short article titled “Cramming more components onto integrated circuits” in the trade magazine Electronics (Volume 38, Number 8, April 19, 1965). The Computer History Museum, which holds the original scanned pages, describes it as “the original magazine article that led to the hypothesis that today is called Moore’s Law.” Moore would go on to co-found Intel three years later.
The core observation was about economics as much as physics. Moore noticed that the number of transistors and other components that could be packed onto a single chip at the lowest cost per component had been doubling at a steady pace, and he projected that this trend would continue for at least a decade. The doubling cadence was later popularly settled at roughly every two years. The practical effect was that computing power per dollar fell relentlessly, year after year, for decades.
Moore’s Law matters to artificial intelligence because it is the substrate the whole field grew on. Every advance that depended on more calculation, from training larger neural networks to running them at scale, rode on the exponentially cheaper compute that Moore’s Law delivered. As later analyses of AI compute would show, the demand for calculation in AI eventually outpaced even Moore’s Law, but the baseline of ever-cheaper transistors is what made that race thinkable in the first place.
Why business readers should care: Moore’s Law is the reason computing got cheap enough for machine learning to become practical at all. Understanding that AI capability is tightly coupled to the cost and availability of compute helps explain why chip supply, data center buildout, and hardware efficiency are now front-page business concerns rather than back-office engineering details.