Technology, along with increasing access to cheap energy, is the lifeblood of a growing, modern economy. As we discussed in our December 2, 2012 article (“The Global Productivity Riddle and the Supercomputing Race“), fully 85% of productivity growth in the 20th century could be attributed to technological progress, as well as increasing accessibility/sharing of cheap energy sources due to innovations in oil and natural gas hydraulic fracturing, ultra-deep water drilling, solar panel productivity, and the commercialization of Generation III+ nuclear power plants and deployment of smart power grids.
Perhaps the most cited example where the combined effects of technological and human capital investments have had the most economic impact is the extreme decline in computing and communication costs. Moore’s Law, the ability of computer engineers to double the amount of computing power in any given space every 2 years, has been in effect since the invention of the transistor in the late 1940s. Parallel to this has been the rise of the supercomputing industry. Started by Seymour Cray at Control Data Corporation in the 1960s, the supercomputing industry has played a paramount role in advancing the sciences, most recently in computationally intensive fields such as weather forecasting, oil and gas exploration, human genome sequencing, molecular modeling, and physical simulations with the purpose of designing more aerodynamic aircrafts or better conducting materials. No doubt, breakthroughs in more efficient supercomputing technologies and processes is integral to the ongoing growth in our living standards in the 21st century.
Unfortunately, advances in both the U.S. and global supercomputing industry has lagged in the last several years. Every six months, a list of the world’s top 500 most powerful supercomputers is published. The latest list was compiled in June 2015; aside from providing the most up-to-date supercomputing statistics, the semi-annual list also publishes the historical progress of global supercomputing power, each country’s share of global supercomputing power, as well as a reasonable accurate projection of what lies ahead. Figure 1 below is a log chart summarizing the progression of the top 500 list from its inception in 1993.
Figure 1: Historical Performance of the World’s Top 500 Supercomputers
As shown in Figure 1 above, both the sum of the world’s top 500 computing power, as well as the #1 ranked supercomputer, has remained relatively stagnant over the last several years. Just three years ago, there was serious discussion of the commercialization of an “exaflop” supercomputer (i.e. a supercomputer capable of 1 x 10^18 calculations per second) by the 2018-2019 time frame. Today, the world’s top computer scientists are targeting a more distant time frame of 2023.
From the U.S. perspective, the slowdown in the advent of the supercomputing industry is even more worrying. Not only has innovation slowed down at the global level, but the U.S. share of global supercomputing power has been declining as well. Three years ago, the U.S. housed 55% of the world’s top 500 supercomputing power; Japan was second, with 12% of the world’s supercomputing power. Rounding out the top five were China (8%), Germany (6%), and France (5%). Today, the U.S. houses only 46% of the world’s supercomputing power, with countries such as the UK, India, Korea, and Russia gaining ground.
Figure 2: Supercomputing Power Distributed by Country
Bottom line: Since the invention of the transistor in the late 1940s and the advent of the supercomputing industry in the 1960s, the U.S. has always led the supercomputing industry in terms of innovation and sheer computing power. With countries such as China and India further industrializing and developing their computer science/engineering expertise (mostly with government funding), U.S. policymakers must encourage and provide more resources to stay ahead of the supercomputing race. To that end, President Obama’s most recent executive order calling for the creation of a National Strategic Computing Initiative–with the goal of building an “exascale” supercomputer–is a step in the right direction. At this point, however, whether the industry can deploy an energy-efficient exascale supercomputer by the less ambitious 2023 time frame is still an open question.