Christian Weedbrook, founder of the quantum technologies company, Xanadu, right, at the company's Toronto office on June 20, 2019.Tijana Martin/The Globe and Mail
If quantum computers are destined to play a significant role in the next decade, they will have to scale up well beyond the machines that represent the field’s current state of the art and potentially occupy vast data centres.
Now, Toronto-based Xanadu Quantum Technologies Inc. has demonstrated its plan for achieving a commercial quantum computer, starting by successfully networking together four small, independent systems and sharing computations across the resulting ensemble.
“You can think of this as a very, very early baby data centre,” said Xanadu founder and chief executive officer Christian Weedbrook. “It only has four server racks, but what we could do today is actually scale this up to thousands.”
Xanadu’s development team also made significant strides reducing signal loss between the four components, Mr. Weedbrook said. Loss is a key challenge for the type of quantum computer the company is developing, which encodes information into packets of light waves that move through optical fibres.
The result, published Wednesday in the research journal Nature, comes at a time of duelling enthusiasm and doubt about the near-term utility of quantum computing among investors and industry watchers.
Last month, Google announced Willow, its latest quantum chip, which some experts have hailed as a breakthrough because it demonstrates “quantum advantage” – the ability to perform a specific type of mathematical task faster than any conventional digital system while also showing that a quantum computer can manage the errors that contaminate calculations as systems increase in scale.
But skeptics note that quantum computers such as those under development by Google and others are far from operating at a commercially relevant scale. After surging in late 2024, quantum computer stocks took a hit earlier this month when Nvidia Corp. Chief executive officer Jensen Huang said he thinks the technology is still 15 to 20 years away from being “very useful.” Many companies developing the technology, including Burnaby, B.C.-founded D-Wave Quantum Inc., dispute that it will take that long.
Xanadu's latest quantum machine, Aurora, sits on four server racks with its components networked together using optical fibres.Supplied
Xanadu captured the industry’s attention in 2022 when it achieved quantum advantage with a machine dubbed Borealis. In doing so, the company demonstrated other benefits that come with a light-based quantum computer, including the fact that most of its components can operate at room temperature, unlike systems developed by Google and International Business Machines Corp., among others, which employ superconducting circuits that must be kept at temperatures near absolute zero.
The Toronto-based company has raised more than US$250-million to date from investors and is hoping to raise another US$100-million this year to further its development.
But Xanadu must still grapple with the more general challenge of error correction that all quantum systems face. With this week’s announcement, Mr. Weedbrook said his company has shown how it plans to do that.
All quantum systems derived their power from the fact that they use qubits – physical systems that can simultaneously occupy a mixture of physical states. This fundamentally different from the electronic bits of conventional computers that can only be on or off, representing either 1 or 0 in a standard digital calculation. By harnessing the ambiguity in its qubits, a quantum computer can effectively perform multiple operations at once, creating a shortcut that enables certain kinds of calculations to be performed in a fraction of the time needed by even the most powerful conventional supercomputers.
Areas in which quantum computers are expected to eventually have in impact include cybersecurity, materials science, drug discovery and a range of optimization and risk management problems.
The Achilles heel that stands in the way of this is the problem that quantum systems are easily disturbed and so become error prone as they grow larger with more qubits. To combat this, each qubit requires many additional qubits – 1,000 or more – for added redundancy to tamp down errors. The holy grail of the industry is a “fault tolerant” system, which embeds all those extra qubits into its architecture and allows the device to provide coherent results.
For Xanadu, the answer is networking between devices, which allows larger numbers of qubits to be connected. Its new device, dubbed Aurora, features a total of 12 qubits linked across 35 photonic chips using a combined 13 kilometres of optical fibres.
While the number of qubits is relatively small – in comparison, Willow features more than 100 qubits – the key distinction is the modular way they are connected, Mr. Weedbrook said. The company now plans to pursue this approach on the way to a fault-tolerant machine.
“They have put together an integrated system in all its complexity,” said Christoph Simon, a quantum researcher at the University of Calgary who is not involved with Xanadu. “It’s a commendable contribution.”
Dr. Simon said the company was forthright about the substantial challenge it must still overcome to reduce signal loss as light travels through an increasingly larger number of components.
“Loss rates will still have to come down by orders of magnitude,” he said, adding that the resulting system will need to be the size of a conventional data centre to operate.
Mr. Weedbrook said loss remains Xandu’s overriding preoccupation, but he is optimistic that the company will continue to make progress by working closely with manufacturers of its components and optimizing how its modular system is designed. He added that the company may ultimately be able to assist networking other forms of quantum computers that pursue a modular approach, though he added that Xanadu still favours a system that is entirely light-based.
“All the pieces are there now for error correction and fault tolerance and scalability,” he said. “We have this imaginary dial in our minds where we’ve just got to keep cranking down the loss, and then suddenly you’re there.”