Intel 4004
The Intel 4004 was the first commercially sold single-chip microprocessor, released by Intel in November 1971, and it is the concrete object the abstract idea of "a CPU on one chip" first became. It was a 4-bit processor of roughly 2,300 transistors fabricated in Intel's silicon-gate PMOS process, running at a clock of about 740 kHz, and it was sold not as a computer but as a building-block component — the central piece of the four-chip MCS-4 family. Its historical weight comes less from its modest capability than from the fact that it existed at all and proved a general programmable processor could be manufactured and sold as a standard catalog part.
The 4004 was designed to fill a contract for the Japanese calculator company Busicom, and the pivot away from that original purpose is the whole story. Busicom commissioned Intel in 1969 to build a chipset for a line of printing calculators, initially specified as a dozen or so custom logic chips. Intel's Ted Hoff proposed replacing that custom logic with a smaller set of general-purpose programmable chips, Stan Mazor helped define the architecture and instruction set, Busicom's engineer Masatoshi Shima contributed the detailed logic requirements, and Federico Faggin led the actual silicon design that turned the proposal into a working chip using the silicon-gate technique that made the density practical. The result solved the calculator problem by making a tiny computer and then writing the calculator in software.
The 4004 was one member of the MCS-4 chip set, not a standalone system, which clarifies what the single-chip claim does and does not mean. The processor itself was the single-chip CPU, but a working machine also needed its companions: the 4001 ROM holding the program, the 4002 RAM holding data, and the 4003 shift register for expanding input and output. So "computer on a chip" is precise about the processor and misleading about the system — the 4004 collapsed the CPU onto one die while memory and I/O still lived on separate chips, which is exactly the boundary that later microcontrollers would erase by pulling those companions on-die too.
Intel initially did not own the right to sell the 4004 on the open market, and reclaiming that right is why the part became an industry rather than a single product. Under the original agreement Busicom held exclusive rights to the design. Intel negotiated to buy back the general-purpose marketing rights — reportedly in exchange for a price reduction on Busicom's parts — which freed Intel to sell the processor to any customer for any application. That commercial decision mattered as much as the engineering: a brilliant chip locked to one calculator vendor would have been a footnote, while a brilliant chip Intel could sell to everyone became the seed of the whole merchant microprocessor business.
The 4004's real significance is as an existence proof that reset the industry's sense of what was possible. Before it, the economical way to build digital products was fixed-function logic designed per product; after it, the question for a huge range of products became whether a programmable processor plus a ROM program could do the job more cheaply and flexibly. The 4004 itself was quickly superseded — Intel's 8-bit 8008 followed in 1972 and the far more influential 8080 in 1974 — but those successors were walking through a door the 4004 opened, which is why it is remembered as a beginning rather than for anything it individually computed.
Read the Intel 4004 as a proof of concept that happened to ship: a calculator contract answered by building a general-purpose computer small and cheap enough to sell as a part, which turned the microprocessor from an idea into a product category the rest of the industry then spent fifty years scaling.