Whoever controls the future of computing, owns the world. Tech companies know it. Countries know it. And that's why those parties that aim to own the biggest footprint on our future world are in a panicked race to build increasingly powerful supercomputers.
Who’s winning out? And how exactly will all these computing investments pay off? Before we explore these questions, let’s recap the state of the modern supercomputer.
A supercomputer perspective
Just as in the past, today’s average supercomputer is a massive machine, comparable in size to a parking lot that holds 40-50 cars, and they can calculate in a day the solution to projects what would take the average personal computer thousands of years to solve. The only difference is that just as our personal computers have matured in computing power, so too have our supercomputers.
For context, today’s supercomputers now compete at a petaflop scale:
1 Kilobyte = 1,000 bits
1 Megabit = 1,000 kilobytes
1 Gigabit = 1,000 Megabits
1 Terabit = 1,000 Gigabits
1 Petabit = 1,000 Terabits
To translate the jargon you'll read below, know that a ‘Bit' is a unit of data measurement. ‘Bytes' are a unit of measurement for digital information storage. Finally, ‘Flop' stands for floating-point operations per second and measures the speed of computation. Floating-point operations allow the computing of very long numbers, a vital ability for a variety of scientific and engineering fields, and a function that supercomputers are specifically built for. This is why, when talking about supercomputers, the industry uses the term ‘flop.'
Who controls the world’s top supercomputers?
When it comes to the battle for supercomputer supremacy, the leading countries really are who you’d expect: mainly the United States, China, Japan and select EU states.
As it stands, the top 10 supercomputers (2018) are:
(1) AI Bridging Cloud | Japan | 130 petaflops
(2) Sunway TaihuLight | China | 93 petaflops
(3) Tianhe-2 | China | 34 petaflops
(4) SuperMUC-NG | Germany | 27 petaflops
(5) Piz Daint | Switzerland | 20 petaflops
(6) Gyoukou | Japan | 19 petaflops
(7) Titan | United States | 18 petaflops
(8) Sequoia | United States | 17 petaflops
(9) Trinity | United States | 14 petaflops
(10) Cori | United States | 14 petaflops
However, as much as planting a stake in the global top 10 holds prestige, what truly matters is a country's share of the world's supercomputing resources, and here one country has pulled ahead: China.
Why countries compete for supercomputer supremacy
Based on a 2017 ranking, China is home to 202 of the world's fastest 500 supercomputers (40%), while America controls 144 (29%). But numbers mean less than the scale of computing a country can exploit, and here too China controls a commanding lead; aside from owning two of the top three supercomputers (2018), China also enjoys 35 percent of the world's supercomputing capacity, compared to the US's 30 percent.
At this point, the natural question to ask is, who cares? Why do countries compete over building ever faster supercomputers?
Well, as we’ll outlined below, supercomputers are an enabling tool. They allow a country’s scientists and engineers to continue making steady progress (and sometimes giant leaps forward) in fields such as biology, weather forecasting, astrophysics, nuclear weapons, and more.
In other words, supercomputers allow a country’s private sector to build more profitable offerings and its public sector to operate more efficiently. Over decades, these supercomputer-enabled advancements could significantly transform a country’s economic, military, and geopolitical standing.
At a more abstract level, the country that controls the biggest share of supercomputing capacity owns the future.
Breaking the exaflop barrier
Given the realities outlined above, it shouldn’t come as a surprise that the US is planning a comeback.
In 2017, President Obama launched the National Strategic Computing Initiative as a partnership between the Department of Energy, Department of Defense, and National Science Foundation. This initiative has already awarded a total of $258 million to six companies in an effort to research and develop the world’s first exaflop supercomputer called Aurora. (For some perspective, that’s 1,000 petaflops, roughly the calculation power of the world’s top 500 supercomputers combined, and a trillion times faster than your personal laptop.) This computer is set for release around 2021 and will support the research initiatives of organizations like the Department of Homeland Security, NASA, the FBI, the National Institutes of Health, and more.
Edit: In April 2018, the US government announced $600 million to fund three new exaflop computers:
* ORNL System delivered in 2021 and accepted in 2022 (ORNL system)
* LLNL System delivered in 2022 and accepted in 2023 (LLNL system)
* ANL Potential System delivered in 2022 and accepted in 2023 (ANL system)
Unfortunately for the US, China is also working on its own exaflop supercomputer. Hence, the race continues.
How supercomputers will enable future science breakthroughs
Hinted at earlier, current and future supercomputers enable breakthroughs in a range of disciplines.
Among the most immediate improvements the public will notice is that everyday gadgets will begin working a whole lot faster and better. The big data these devices share into the cloud will be processed more effectively by corporate supercomputers, so that your mobile personal assistants, like the Amazon Alexa and Google Assistant, will start to understand the context behind your speech and answer your unnecessarily complex questions perfectly. Tons of new wearables will also give us amazing powers, like smart earplugs that instantly translate languages in real time, Star Trek-style.
Likewise, by the mid-2020s, once the Internet of Things matures in developed countries, nearly every product, vehicle, building, and everything in our homes will be web connected. When this happens, your world will become more effortless.
For example, your fridge will text you a shopping list when you run out of food. You’ll then walk into a supermarket, pick out said list of food items, and walk out without ever engaging with a cashier or cash register—the items will automatically be debited from your bank account the second you exit the building. When you walk out to the parking lot, a self-driving taxi will already be waiting for you with the trunk open to store your bags and drive you home.
But the role these future supercomputers will play at the macro level will be far larger. A few examples:
Digital simulations: Supercomputers, especially at the exascale, will allow scientists to build more precise simulations of biological systems, like weather forecasts and long-term climate change models. Likewise, we’ll use them to create better traffic simulations that can aid the development of self-driving cars.
Semiconductors: Modern microchips have become far too complex for teams of humans to effectively design themselves. For this reason, advanced computer software and supercomputers are increasingly taking a leading role in architecting tomorrow's computers.
Agriculture: Future supercomputers will enable the development of new plants that are drought, heat, and salt-water resistant, as well as nutritious—essential work necessary to feed the next two billion people projected to enter the world by 2050. Read more in our Future of Human Population series.
Big pharma: Pharmaceutical drug companies will finally gain the ability to fully process a massive range of human, animal, and plant genomes that will aid new drug and treatment creation for a variety of the world’s common and not-so-common diseases. This is especially useful during new virus outbreaks, like the 2015 Ebola scare from East Africa. Future processing speeds will allow pharmaceutical companies to analyze a virus’ genome and build customized vaccines within days instead of weeks or months. Read more in our Future of Health series.
National security: This is the main reason why the government is investing so heavily into supercomputer development. More powerful supercomputers will help future generals create precise battle strategies for any combat situation; it will help design more effective weapons systems, and it will help law enforcement and spy agencies better identify potential threats long before they can harm domestic civilians.
And then we come to the controversial topic of artificial intelligence (AI). The breakthroughs we’ll see in true AI during the 2020s and 2030s depend entirely on the raw power of future supercomputers. But what if the supercomputers we hinted at throughout the entirety of this chapter could be made obsolete by an entirely new class of computer?
Welcome to quantum computers—the final chapter of this series is just a click away.
Future of Computers series