Powering quantum computers – The ultimate challenge
Thu 22 Jul 2021 | Martin Keenan
Quantum computing in a practical sense might still be some way off – several decades by some estimates – but various quantum computers are in operation today, raising considerable and complex power demands.
Supercomputer = superpower demand
The world of exascale computing and beyond is one of enormous pressure – not just on scientific innovation and bleeding-edge endeavours, but also on the raw materials required. Consider the standard supercomputers that make up the top 500 list of the most powerful in the world – the top performer (as of November 2020) was the Japan-based Fugaku, a 7.6 million core monster delivering a new world record 442 petaflops result on the High-Performance Linpack (HPL) benchmark. That impressive feat requires nearly 30mW (29,899kW). Summit, an IBM-built system and the fastest system in the US at 148.8 petaflops requires just over 10mW, while China’s contender, Sunway TaihuLight achieves 93 petaflops at just over 15mW.
Beyond supercomputers: The challenges
Those are all substantial power requirements, but the next generation of quantum powered computers raise even more demanding issues, including the not-insignificant question of cooling. Quantum computers rely on ‘qubits’, rather than the traditional ‘bits’ of a standard machine, a qubit being able to reflect both 0 or 1 states simultaneously thanks to superposition. However, the qubits must be as-near-totally isolated as possible to maintain superposition, as well as very cool – around 0.015 Kelvin or -273C. The result is that quantum computers not only need to begin operating with qubits only very slightly above absolute zero but need to be able to maintain that temperature consistently in use – a weighty challenge.
Creative cooling solutions
Clearly, this demands some unique and creative approaches to creating and maintaining such low temperatures, and there have been plenty of contenders. One research team from Aalto University in Finland has built a standalone cooling device for quantum computers dubbed the ‘nanofridge’, which uses an energy gap dividing two channels to produce a cooling effect. The two channels conduct electrons at different speeds – one a superconducting fast lane, the other a resistive slow lane. Electrons with enough energy can jump the gap, removing energy from the system much like the way a standard fridge works. However, the team believes many years might pass before the concept would be commercially useful.
Diamonds are – surprisingly – cold
Another potential method of minimising power requirements for cooling is to use industrial diamonds in so-called ‘ diamond-based spin qubits’. Although this has been discussed as a concept for many years, a recent announcement from Fujitsu specifically details the technology as being under active research for a ‘large scale system’. The research team, based in the Netherlands, will investigate the practical application of the diamond qubits, noting: “The size of refrigeration units for cooling qubits is expected to be small because the quantum states of diamond-based spin qubits can be stable at higher temperatures than most competing platforms”.
Quantum debate continues on every level
Interestingly, one of the key reasons that the best way of cooling – and powering – quantum computers is still under debate is that the technology as a whole is still in the research stages, with different measurement metrics and completely different hardware approaches very much in evidence. For example, some of the most recent news in quantum computing was a series of announcements concerning researchers achieving ‘quantum supremacy’, including claims from Google and IBM in the US and a combined discipline research team in China.
These announcements as a whole indicate the vast amount of research being undertaken into quantum computing in general, but also expose some of the key tensions. Even the term ‘quantum supremacy’ – originally taken to mean a successful calculation using a quantum computer that could not practically be done with a traditional machine – has been a point of debate. Google’s initial announcement of success is debated by IBM, which claims that the random sampling calculation could have been done with supercomputers – admittedly in 2.5 days as opposed to Google Sycamore’s quantum-powered 3 minutes and 20 seconds.
Meanwhile, China’s ‘quantum supremacy’ entry – the Jiuzhang quantum computer – uses a completely different optically-based system, measuring photons travelling through optical circuits. The final readout is conceptually similar to a qubit on a processor like Google’s or IBM’s, but making direct comparisons in terms of processing power is still a matter of scientific debate.
Commercial applications already in the mix
There are several clear certainties emerging, however, the clearest being the desire to commercialise quantum technology as soon as possible. D-Wave Systems Inc (used by Google’s Quantum AI Lab) launched the Leap quantum cloud service, which businesses such as Volkswagen and Accenture are already using to deploy ‘hybrid quantum applications’. D-Wave’s most recent launch is the Advantage quantum system, with more than 5000 qubits and 15-way qubit connectivity. Other cloud contenders include Amazon’s Braket, the IBM Q network, and most recently, Microsoft’s Azure Quantum service.
It is also clear that use cases for quantum computing among large corporates are beginning to gather pace, as cloud offerings open up the market. As those applications crystallise, it will become clearer whether on-premise quantum computing is more desirable than cloud-based quantum, and where the power bottleneck will appear. For now, the research – and debate – continues..