The race to quantum computing

Main visual : The race to quantum computing

Computing has really come a long way, in a short time frame. In 1970 we had around 2,000 transistors per chip. According to Moore’s law, the number of transistors per chip doubles every 18-24 months – so today we have around five to ten billion in individual chips.

As a result, we all carry smartphones with quad-core processors and several billion transistors – by far exceeding even space technology from the 1960s or 70s.

However, to progress further, we must continue to make these transistors smaller to increase their performance and minimize the energy they consume. From an architectural perspective, this means getting to a sub-atomic level. That’s when we enter the realm of quantum computing, which means quite some changes in comparison to current architectures.

Leveraging quantum computing and effectively harnessing emerging technologies was very much the theme at the recent GoTech World event in Bucharest, where I was privileged to share my views on the topic.

Thanks to a series of announcements of milestones achieved by the innovation teams at some of the world’s largest companies, quantum and quantum-inspired computing has become the hot technology topics of the year.

Recent achievements include prototype systems with 57 and 72 qubits – which means that the technology has clearly moved beyond theoretical physics and is now in the experimental phase.

You might think that the billions of transistors we already have at our disposal are enough – but actually every single industry has challenges that are still difficult to address, from portfolio risk optimization in financial services to drug discovery in pharmaceuticals or material sciences.

The issue here is the extreme complexity that comes from the vast numbers of variables. To use an example from manufacturing, optimizing the path that robots take to perform seam sealing, a sequence of 64 seams requires considering 10106 possible combinations.

Determining the optimal one by working through them all in sequence – which is how classical computers fundamentally work – would simply take millions of years. That’s why quantum computing is attracting so much attention right now.

There are two forms of quantum computing. The first system to be introduced to the market was quantum annealing. This leverages tunneling to solve complex combinatorial problems. A Canadian vendor currently offers a 2,000-qubit production system that solves these challenges and further announced a 5K qubit system.

The second type is a significantly more complex type of quantum technology. Called Quantum Gate, it leverages superposition to solve much wider problems. However, both these technologies in particular have major limitations today.

For example, they run at 270 degrees centigrade below zero, and must be isolated from any disturbance such as noise or magnetic fields. They can also only currently hold their quantum state for approximately 90 nanoseconds and presently accuracy can be variable.

The size of the problem that you are trying to solve also determines how many physical qubits are needed. For example, a chemistry problem that explores combinations of only five atoms needs between 5,000 and 5 million physical qubits, depending on the quality of the qubits themselves.

With recent quantum computing qubit achievements still in double figures, we’re some way from having that number at our disposal.


How can organizations solve some of these complex problems today?

As an industry, we’ve made great strides, but these quantum technologies are not yet up to the task of solving the real-world problems that all industries face.

Consequently, Fujitsu has invested in leveraging quantum algorithms and its advancement on a unique IP architecture to deliver a quantum-inspired solution that is 10 thousand times faster than standard computing at solving complex combinatorial problems.

That’s the equivalent of leapfrogging twelve Moore’s generations. And what is more, it can be easily deployed from edge to cloud (at any temperature!). It is also cost-effective as it is a digital circuit-based architecture – and highly accurate. Our solution is the Digital Annealer.

What’s more, the Digital Annealer is already solving real-life customer optimization challenges today. For example, we recently worked with a large automotive manufacturer to optimize the seam welding route we mentioned above. This is part of the paint shop process, which represents between 30 and 50 percent of total car manufacturing costs. We also ran a comparison to currently available quantum computing.

Today’s quantum computers can optimize seven seams in the process while our first generation Digital Annealer (with 1024 bits) could optimize the best path for 22 seams. Our current generation digital annealer (8192 bits) has successfully optimized 64 seams, which represents a real competitive advantage for the manufacturer.

The next generation digital annealer (which will deliver 1 million bits) will have the ability to process more than 90 seams – a capability that could prove potentially disruptive.

The Digital Annealer comprises a quantum-inspired hardware layer and a software layer that we’ve partnered with 1QBit to deliver as a simple software development kit for customers. We work closely with customers to identify the problem they want to solve, and translate that problem into a mathematical model, known as a quadratic unconfined Ising model. Then we feed that model into the hardware annealer and extract the results.

We’ve successfully deployed this with customers in all industries – for example we enabled a large car manufacturer to help optimize tasks such as job shop scheduling, and to find the optimum position for a car’s wing mirror to minimize wind noise in new vehicles.

In Japan, we’ve worked with the Post Office to optimize transport routes, truck types and cargo loads. This enabled Japan Post to shrink its delivery fleet while accelerating delivery times and reducing cost.

We’ve also worked with UK retail bank NatWest to optimize its portfolio. Starting with 30 assets, and now handling thousands, we have helped ensure the portfolio is diversified, ensuring minimum risk and maximum return.

We’ve also optimized one of our own factories in Japan, streamlining the parts picking process and resulting in massive 45 percent time savings.

Isaac Chuang, MIT Professor of Physics and Engineering summed up the current state of quantum saying: “The thing driving the hype is the realization that quantum computing is actually real. It is no longer a physicist’s dream — it is an engineer’s nightmare”.

Quantum computing might have moved to the prototype stage, but it isn’t quite ready for prime time yet. Personally, I am very optimistic about its future evolution.

But while quantum computing comes with an uncertain timeline, the good news is that Fujitsu Digital Annealer – the world’s first quantum-inspired solution – can implement and leverage quantum algorithms to deliver performance today, that is generations ahead without the cost and complexity associated with quantum engineering.