Impact of Quantum Computing on the Global Energy System

Introduction

I recently ran across an article by the CEO of D-Wave Vern Brownell, in which he was explaining his views on the impact of quantum computing on the way world uses energy. The article was shared by the esteemed teacher Jonathan Reichental, who is also working on some education material for quantum computing. I personally found the viewpoint of the article a bit narrow.

The same topic of the energy efficiency of quantum computing came up also during my visit to the Kyberykset podcast, but we didn’t really delve into it. So, I thought I would now write an article explaining the whole question as I see it. Let’s study the question of how will quantum computing change the global energy consumption by a series of ever broadening viewpoints and levels of abstraction.

Basics: Thermodynamics, Superconductivity and Power

The first law of thermodynamics states that energy never disappears. Thus, when we talk about something using energy, we are actually talking about the second law, which states that the entropy or the disorder in any closed system never decreases. When we translate this back to our energy viewpoint, it means that energy turns into a new form that is less usable. Other kinds of energies turn into heat and heat differences blend until we can’t extract any more work from the energy of the system. So, when something like a computer uses energy, we put in highly usable electricity and we get to do computations and create heat. If we continue to watch what happens to every output, they all end up turning towards heat. For example, the light from your screen hits the wall of your room and heats it.

There is a certain irony to this, since all kinds of computing devices perform badly if they overheat. Thus, we end up spending even more energy in trying to keep the relevant parts cool, but create more heat in total. For example, we spend energy to move waste heat away a processor in a computer by running an electrical fan to cool it.

A quantum computer has many potential hardware implementations, but currently the most popular implementations are based on superconductivity and so is D-Wave’s computer. Unfortunately, we only know how to achieve superconductivity in very cold temperature, so these machines are in many ways the finest refrigerators in the world that use quite a lot of energy. For example, D-Wave’s 2015 version used 25kW of power, with the bulk used in cooling.

That usage is about 25 to 100 times more than a desktop PC. Running the D-Wave machine for an hour, would allow you to run your mobile device for months nonstop. Sounds like an energy hoard, but then again, Tesla S motors can output 581 kW of power. And the measly Mazda 6, I used to drive, had an engine that could output around 90 kW. Looking back at computing, the fastest supercomputers in the world require power in the range of 10.000 kW or four hundred times more power than the D-Wave machine.

It appears it is hard to say by looking at the power only, how energy efficient a quantum computer is.

What If We Didn’t Need Cooling?

Extremely good to note about the superconducting implementation is that the chip really is superconducting, thus the actual calculation uses very little energy. This seems to be the key tenet in mr. Brownell’s argument. In essence, the current quantum machines would be very energy efficient, if we could produce superconductivity in room temperature and remove the need for cooling.

As noted, this viewpoint is very narrow for several reasons. First, if we would be able to create room temperature superconductivity, we could solve many of the world’s energy problems with that technology alone. For example, 8 to 15% of electricity is lost between the power plant and the consumer mostly due to electrical resistance and the story is similar in all kinds of electronic devices including those classical computers. Second, the problem is extremely hard. Like quantum computing, room temperature superconductivity has credible academic opponents that claim it is simply impossible.

I wouldn’t bet on any estimate of energy impact for a technology that assumes room temperature superconductivity.

But in Theory: Let’s Say We Can Superconduct

Say we do tame the superconductivity, but continue to exponentially scale up our demand for compute. Does superconductivity really make quantum and classical computers equally effective from an energy perspective? Wouldn’t then both devices essentially have zero electrical resistance and waste no energy into heat?

Turns out a law of nature that has sometimes been dubbed the minus first law of thermodynamics prohibits information from ever disappearing. Therefore, there is thermal cost to just “losing” information. For example, summing two and three to give out five, only stores the five, it needs to “lose” the information of these input components. This “lost” information has to be output as heat.

In contrast, a quantum computer performs reversible calculations, where no information is ever lost and you can in theory always run the results backwards through the machine to produce the inputs. There is a theoretical energy improvement in moving to quantum computing that cannot be overcome by any engineering effort. This limit is still very far from what we actually use in computers these days, so it is not a constraint on current designs.

However, as the quickest readers have discovered, it is quite easy to envision classical calculation devices that simply don’t lose the input and are also thus reversible without being quantum at all. So, there is a way for us to breach this energy boundary without turning to quantum computing.

Anyway, maybe we won’t even end up using the superconducting qubits. After all, it is just one of the competing implementations.

From Device Power Input to Energy Demand

Now that we’ve gotten all the engineers juiced up, let’s come back to the practical world.

What is the quantum computer for?

Why do we want to have quantum computers in the first place? To solve problems that are exponentially hard for a classical computer and thus practically impossible.

Then what is the point of comparing the energy efficiency per time unit of a machine that would use billions of years to solve a problem, to a machine that would use forty-two minutes?

If both are actual physical machines, there is practically no way that the faster machine wouldn’t also be more energy efficient in solving that specific problem. After all it is using its power input for just a blink. And for any given problem, it is almost certain we will be able to deduce beforehand whether a quantum computer would be exponentially faster. Thus, we would only use one, when we get a huge speed and subsequently almost invariably a huge energy improvement.

With the risk of losing all the engineers altogether, consider still this. You only have a classical computer. The billions of years would mean that you are never going to get the results. Thus, why would you calculate at all?

Is there an answer you'd wait to get for billions of years?

Practically all the problems the quantum computers would be solving, would not have been solved without them. In essence all of the energy could have been left unused and every piece of electricity put into a quantum computer is a new drain on the global resources. Quantum computing represents a potentially massive increase in the overall use of compute. While they may be massively more efficient, they are an additional use of energy.

Why We Compute?

And finally, we come to the real crux of the matter. We have seen that most ways of comparing classical and quantum computers’ energy efficiency are meaningless, since we are actually creating a new use of energy for things we weren’t even trying before. However, the increased energy use should not be seen as a comparison, but as an investment. Investment towards what? Why do we want quantum computers?

Many of the problems currently impossible and best suited for a quantum computer are directly linked to the global energy system. The most famous examples of quantum computer use are related to the travelling salesman problem and simulating chemistry.

Solving the first efficiently is analogous to an extremely wide range of optimization problems in manufacturing, industry, telecommunications and logistics. Even actually optimizing the routes all of the real sales persons in the world take, might yield quite significant benefits. Globally more than a quarter of all energy is used in transportation, whereas electricity in total was less than a fifth. And since computing is at maximum ten percent of all electricity use, we can calculate that transportation uses something in the range of thirteen times as much energy as computing. Turn it around and we find out that a solution that could optimize energy use in transportation by 8% would offset all of the computing energy use in the world. And if you think about climate change rather than energy, transportation is almost exclusively fossil fuel based whereas electricity isn't.

Scientific computation has relatively good electricity demand flexibility

The second example of chemistry simulations nicely always zooms in on a molecule called nitrogenase. This molecule is used by nature to fix nitrogen. A thing one needs to do to create fertilizer. Without fertilizers there is no way to have seven billion humans fed. We humans use a different process for fixating nitrogen that happens to represent up to 3% of all the energy use in the world. With quantum computers we are expecting to understand the nature’s way well enough to copy it, opening the door to cutting world’s energy consumption by close to that 3% or roughly 1.5 times the amount we use for computation in total.

We are not running out of energy, but out of climate.

Some of the other optimization uses of quantum computing are a bit less well known and developed, but the devices could be of enormous help in optimizing the national electricity systems and even broader energy systems, modelling individual electronics or the whole climate, which is interesting, because we are not running out of energy, but out of climate, and to top it all with some recursive irony – quantum computers are also likely to be extremely useful in solving how to create superconductivity at room temperatures.

Summary

In conclusion, comparing the energy use of a quantum computer to a classical is hardly meaningful. Quantum devices will open up completely new uses for computing that are extremely likely to help us solve problems related to energy on a scale that is transformational to us as a species and to the planet we inhabit.

Since the current annual energy spent on the Bitcoin network, could run that mentioned D-Wave machine for 2.8 billion hours, even if the only thing we achieve is to break Bitcoin, we’ll save up enough energy to do planet scale magic with quantum computers.