In his 1950 paper “Programming a Computer for Playing Chess,” Claude Shannon estimated that the number of possible chess games is on the order of 10 to the 120th power. This figure, sometimes called the “Shannon number,” was his way of showing why a chess program cannot simply search the whole game to the end: the tree of possibilities is vastly larger than the number of atoms in the observable universe.
The estimate motivated the rest of his design. Because exhaustive search is impossible, Shannon argued a program must cut off its look-ahead at a limited depth and use an approximate evaluation function to score positions, rather than calculating all the way to checkmate. Every practical chess engine since, including Deep Blue, has worked within that constraint.