Revisiting the Computation Problem

In 2013, Engelhardt (2013) calculated that the combined power of the top five hundred supercomputers would take approximately 10.5 quintillion years to compute the distribution of eighty thousand heterogeneous goods among six billion consumers, posing a serious practical challenge to the implementation of computerized central planning. Allin Cottrell (2021) calls into question Engelhardt’s assertion, noting that the algorithm used by Engelhardt not only scales poorly but is not even valid for the problem Engelhardt posed. Cottrell offers an iterative algorithm that a one-petaflop machine could use to solve the distribution problem in about five minutes. We correct errors in both Engelhardt and Cottrell, and we offer a way to incorporate production into the problem. The result: modern supercomputers are still not powerful enough to solve central planning’s computation problem.