True quantum supremacy and the race to a million qubit chip
Is quantum supremacy - the moment when conventional computers become obsolete - a fiction?
In quantum computing, many people look at quantum supremacy as the real breakthrough, but is it really a mirage?
'Quantum supremacy' - even the phrase gives an impression of an almost Terminator-like self-awareness: the single moment when everything changes, when conventional computers become obsolete and quantum computers take over
The reality is more nuanced: quantum supremacy would represent the latest of several milestones in quantum computing, which have taken us from the first ideas in the 1980s that quantum mechanics could be harnessed to build a profoundly more powerful type of computer; to the realisation in the 1990s that quantum errors could be corrected; to the demonstration of quantum bits in a number of platforms in the 2000s; and to the construction of quantum integrated circuits in the 2010s.
But what quantum computing insiders know is that 'basic' quantum supremacy, achievable with fewer than 100 qubits, doesn't necessarily mean it will solve useful problems.
What is supremacy?
Quantum supremacy is reached when a quantum computer can do something that a conventional computer finds practically impossible, either because it might take millennia to perform the calculation, or because it runs out of all of its available memory. The catch is that this task need not be anything useful. Indeed, the first demonstrations of supremacy are perfectly useless in and of themselves.
The more relevant questions to ask are: once supremacy is achieved, what's next on the route to useful quantum computing and how do we get there?
The post-supremacy era
There is some optimism that quantum software will identify early-stage, non-error corrected quantum processors to be harnessed to deliver useful solutions. Quantum software start-ups are jostling to lead the development of ‘quantum apps' to unlock their computing potential.
Despite this optimism, we don't yet know for certain what really useful problems, if any, these early-stage machines will be able to solve. There are exciting possibilities in the modelling of materials and chemicals - for example using a small-scale quantum processor running in tandem with a conventional computer - but even these have been predicted to require vast numbers of quantum co-processors running in parallel.
These 'noisy intermediate-scale quantum' (NISQ) processors are limited, because they're designed to simply tolerate and work around errors, not to correct them. We can be much more confident of the power of quantum computers with built-in error correction, but these will require a thousand physical qubits or more for each pure, error-corrected ‘logical qubit'. It's likely that millions of qubits will be needed to get a sufficient number of error-corrected qubits to run general purpose quantum algorithms.
While NISQ processors have a key role to play in the exploration and development of quantum apps, and understanding the capabilities of the technology, it is also essential to look beyond this era towards quantum processor technology that can scale to over a million qubits, integrate well with classical computers and have built in density and error tolerance.
Scaling with silicon
The design of a million-qubit quantum chip seems an enormous leap from the 50-100-qubit processors currently under development. But look at silicon transistors for comparison: there are now not millions, but billions on a single chip. By 2025 it's predicted there will be more silicon transistors on planet earth than human cells.
Building qubits out of silicon transistors, or devices that closely resemble them, offers huge potential for scaling by benefiting from mature silicon fabrication processes and facilities.
The advantages of silicon qubits go beyond tapping into a mature trillion-dollar manufacturing industry. Silicon also offers some of the lowest inherent error-rates, increasing the capacity of the physical qubits we have so that more can be dedicated towards useful computation, instead of correcting errors.
Silicon qubits have already shown some early star power in this regard: their quantum state lifetimes exceed seconds, longer than any other qubit in a solid-state device. A lot of this comes down to chemistry: the primary isotope of silicon is non-magnetic, allowing a quiet environment for silicon qubits, and ‘large' 300 mm wafers of this purified silicon-28 isotope are already being produced in industrial grade facilities, ready for processing using standard CMOS production methods.
Additional key benefits of silicon qubits are their high density, leading to the potential to accommodate millions of qubits on a single 1 cm2 chip, and the ability to directly integrate conventional co-processors and control logic on the same chip as the quantum core.
Silicon quantum chips will still need to operate at super-cold temperatures, close to absolute zero, and like today's high performance computing facilities, quantum computers will probably need to be housed in datacentres, accessed using a standard cloud interface or via an API.
Looking forward
We envisage two eras for silicon-based quantum processors:
- The Multi-core NISQ Era, where large numbers of NISQ processors (e.g. with hundreds of qubits) are tiled across the same chip. Even without full quantum error correction this could support machine learning and approximate optimisation algorithms by asking the same question of thousands of identical quantum processors.
- A Million-Qubit Era, when scalable manufacture enables the quantum processor to scale to many thousands or millions of qubits. This would allow universal fault tolerant (UFT) quantum computing, and we expect enormous leaps in our ability to calculate solutions to problems such as materials modelling and quantum chemistry for drug discovery.
A global effort
The tremendous progress in the development of quantum computing we are witnessing has been the result of close collaboration between academic groups, microelectronics centres, national labs and industry, reaching all across the globe.
There are many technical challenges in the road ahead. It's a huge, complex, fascinating puzzle. The quantum computing landscape is already unrecognisable from five years ago, and we're accelerating fast. It's going to be quite a ride.
