The 1947 invention of the transistor eliminated the vacuum tube, but created a new problem: building a computer required thousands of transistors, resistors and capacitors to be soldered together one by one. This "tyranny of numbers" made more complex circuits economically unbuildable.

In the summer of 1958 Jack Kilby, a new engineer at Texas Instruments, was the only one in the lab during the company holiday weeks — too junior to have earned holiday himself, he stayed and worked on his ideas. On 12 September 1958 he ran a circuit of transistors, resistors and capacitors mounted on a small germanium bar — all the parts on a single piece of material, connected without external solder. The first "integrated circuit" was working.

Kilby's design was not practical for mass production: the inter-part connections were hand-drawn thin gold wires. Months later, at Fairchild Semiconductor, Robert Noyce (one of Shockley's "traitorous eight") proposed printing the connections directly on a silicon surface using photolithography. This "planar process" — Jean Hoerni's discovery — combined with Noyce's patent became the foundation of modern integrated-circuit manufacturing.

In 1965 Gordon Moore, then Fairchild's research director, made an observation about the coming decade: the number of transistors on a cost-effective integrated circuit was doubling every two years. "Moore's Law," as it was named, would set the trajectory of the industry for 60 years. The path from Kilby's few-transistor chip of 1958 to today's processors holding hundreds of billions of transistors at 5-nanometre features is the fastest engineering scale-up in human history. The infrastructure of the smartphone, cloud computing and deep learning was laid down in a single September week in Dallas; Kilby received the Nobel Prize in Physics in 2000, while Noyce — who had co-founded Intel in 1968 — had died a decade earlier.