But automation and communication technology have evolved to the point that a large number of small units may be better, cheaper, or more efficient than a small number of large units. Consider supercomputers: until the 1990s, the supercomputer industry focused on increasing computing speed and capacity by building bigger, more specialized machines with greater processing power. “But by the mid-1990s it had become cheaper to network the capacity of CPUs and memory from large numbers of personal computers and computer workstations rather than relying on a single microprocessor,” van Ryzin notes. This shift from large to massively modular computing led to an abrupt collapse of the traditional supercomputer industry in the 1990s. Are we now on the cusp of a similar radical shift in other industries?
The researchers, who also worked with doctoral students Eric Dahlgren and Caner Göçmen, conducted a plausibility analysis to show why thinking smaller might be viable or even preferred. They conclude that many industries are approaching tipping points that will soon make transitioning to small modular infrastructure practical and cost-effective.