Relay computer
The computer built from the telephone network's basic part: thousands of electromechanical switches clacking their way to an answer.
In the 1930s the relay was the only proven digital component going. Telephone exchanges had used it for decades, it switched reliably between two states, and it could be ordered by the crate. Konrad Zuse's Z3 (1941), George Stibitz's model series at Bell Labs and the Harvard Mark I (1944) all rested on it, and Sweden's BARK followed in 1950.
The price was speed. A relay throws in five to ten milliseconds, a vacuum tube in about a microsecond, so multiplication took seconds rather than fractions of one. Against that, relays rarely failed unexpectedly, while ENIAC's 17,468 tubes meant a permanent hunt for whichever one had just burnt out. Relay machines were therefore slow but finished sooner — which in practice decided which problems actually got computed. By the mid-1950s they were gone.