On the prospects of Quantum Computing-enabled business in the midterm, somewhat overstated by a recent McKinsey report.

A just-published McKinsey Digital report confirms my last year prediction, that the most obvious applications of Quantum Computing (QC) should be in areas facilitated by quantum simulation, like developing new materials, superconductors, chemicals, and drugs.

The language of the McKinsey report, however, can lend itself to the hope of obtaining business returns more solid or quicker than they may actually be. Like when it says that (bold chars are mine) “pioneers in advanced industries, global energy and materials, finance, and (to a lesser extent) travel and logistics might start generating significant value from quantum by 2025.

The word ‘value‘ needs careful qualification in this context.

It is perfectly sensible, for leading players in those industries (except perhaps Travel and Logistics), to experiment with and learn QC. Getting ready now is a prerequisite for aspiring to be first movers in benefiting from QC, if and when it practically emerges.

However there is some risk in making these investments, because it might as well happen that QC does not scale to industry-level applicability for decades. And 2025 is too early at any rate if by ‘value’ you mean money.

QC progress is not linear

Some of the QC challenges are so deep, that a number of discontinuous breakthroughs will be necessary for it to progress to large-scale usability and leave experimental labs and boot camps.

Take for example one of the toughest problems, i.e. managing noise in a quantum computer. Unlike in classical digital computers, noise is not a collateral disturbance: it is inherent in quantum objects manipulation.

Imperfect information is caused

  1. by interactions with the environment, i.e. the outside of the near-zero Kelvin chambers where the qubits exist quasi imperturbed. This is sure to take place, for example, whenever we want to read qubits’ current values;
  2. by interactions between the qubits themselves. The special interrelation between qubits’ states, called entanglement, is a core and desirable feature; however, it also causes cross-talk and decoherence, that is the loss of qubits’ quantum properties.

Today, the ratio of actual physical (noisy) qubits to the desired logical (error-free) ones grows exponentially with the required error rate. It easily gets to 5,000 to 1 already when a mere 0.1% error rate is aimed for.

And in addition to needing so many redundant physical qubits, Quantum Error Correction algorithms – running partly on the quantum computer and partly on a classical one – create a huge processing-time overhead.

Some of the most brilliant researchers in the world have been working on Quantum Error Correction for over twenty years now. There is no way to predict whether or not a breakthrough will occur by 2025.

In October 2019, the superb Google quantum team – neither the humblest nor the most pessimistic around – estimated that an error-free quantum computer will require processors with a million or more physical qubits and that it will be “about a decade” before it appears.

The hope for hybrids

One workaround, and indeed a very active field of R&D, is to have today’s Noisy Intermediate-Scale Quantum (NISQ) computers cooperate with digital ones. If successful, this strategy could benefit quantum simulation and quantum-assisted optimization, two fields that could lead to real-world applications before fault-tolerant, error-free quantum computers arrive.

In one of their frequent appearances in the constellation of Nature Research magazines and scientific journals, in 2017 Google exponents predicted commercial manifestations of such hybrid NISQs in five years, aimed at solving the above application problems.

That’s two or three years from today. (But read some of the challenges, outlined here and here one-to-two years after the Google’s forecast).

Accordingly, McKinsey expects that businesses will experiment with the hybrid approaches in 2022-2026, in fields like hard optimization problems and chemistry. Maybe.

However I believe that, if it indeed happens in that time frame, it will be just R&D departments at work. No applications.

The chemistry of it

For example, McKinsey writes that in that time frame, “quantum computers are likely to become powerful enough to start handling meaningful simulations of molecular structures for chemical, materials, and pharmaceutical companies.

The timing of this is actually a much-debated issue in the scientific literature and a continuous chase between enthusiasm and disappointment.

To give you a rough idea, quantum simulation of molecules of medium complexity (certainly not pharma’s) is projected to require circa 1,000 qubits and 100 millions of two-qubit gates (QC’s ‘computing primitives’): about twenty times the qubits and 100,000 times the gates of the Google Sycamore computer that proved quantum supremacy.

So yes, it is possible that chemicals and materials industries R&D departments will be “handling meaningful simulations” of molecular structures: but that is unlikely to translate into product developments by 2025-2026.

Hence, the business value, assuming that it arrives, will not take the form of revenues or profits in that time frame. It will be know-how build-up. Which of course, if QC then materializes, will prove invaluable.

Optimization and AI

As to quantum-assisted optimization (thus indirectly quantum-assisted machine learning), this is even further ahead in time. And, unlike molecule quantum simulation, it is a segment that over the next five years will also be addressed by other types of digital computing accelerators.

Just in 2019, for example, we have witnessed the demonstration of the first scalable Ising Machine, i.e. an analog optical computer tailored to tackle optimization problems.

We also have seen emerging in-memory analog computing circuitry that could be implemented in CMOS and employed to solve the same class of problems: circuits that can multiply matrices in a single step, with no need to decompose the problem in a routine of binary operations.

Similar approaches are mushrooming and some are bound to have an impact on the industry of digital or analog, but non-quantum, accelerators.

So what?

The future belongs to the brave, and it makes perfect sense to invest in a powerful emerging technology. Quantum computing is one of these, for enterprises in selected industries and with robust R&D departments.

Although legitimate, it would be extremely risky to expect that profits or savings can be derived from Quantum computing by 2025. Pioneers in some industries may only generate intangible value (i.e. non-financial) from Quantum computing in that time frame.

2020 PAOLO MAGRASSI CC BY 4.0