Esoteric hardware 

Speaking at the forum, Alexander Keesling, chief technology officer at QuEra, discussed the limits of lasers, which are being optimised for completely different application areas. “You can find very high-powered lasers, but these lasers do not have any kind of coherence,” he said, which makes them unsuitable for use in quantum computers. 

Another challenge the industry needs to resolve, according to Keesling, is scalability. Realistically, a system with twice the number of qubits should not require a machine that is twice the size, he said, adding that the infrastructure required by a quantum computer must be kept in check as the number of qubits increases.

“If we just build another system next to a quantum computer and then figure out how to connect it, I would say that it is extensible. It’s not scalable,” he said. 

Owen Arnold, vice-president of product development at Oxford Quantum Circuits, discussed the challenges in working with suppliers of dilution refrigerators, which are needed to achieve a superconductive environment, as required by some quantum computer systems. “These are very impressive, but they have largely been built for an academic market,” he said.

In Arnold’s experience, the manufacturers of dilution refrigerators need to consider how such specialised equipment can be scaled to operate in a datacentre environment. “If you need a dilution fridge, then you will need the expertise to run that dilution fridge,” he added. 

Another consideration is that quantum computers tend to be built for experimentation. “The experiment goes down; it fails, which is largely acceptable. A small number of people are affected by that,” he said. 

But when you consider the ambitions of the quantum computing industry to deploy such machines in datacentres, it is no longer just an experiment that can fail with limited disruption. Instead, five-nines (99.999%) uptime is required. “We want to have much better diagnostics for these units. We want to control the maintenance cycle, and we want to make sure the redundant power is there,” Arnold added. 

Even if the science behind quantum computing can progress to the point where it can perform with a high degree of reliability, and the industry can make datacentre-ready quantum computing systems, ease of programmability will need to be addressed to achieve datacentre-scale adoption. 

Ease of programming and hybrid complexity 

Josh Savory, director of offering management at Quantinuum, discussed the company’s 2029 roadmap at the forum, and the challenges it will need to overcome to deliver something commercially viable by then. He believes there needs to be industry alignment around standards.

Savory said QIR, the community-wide effort to design and implement the necessary compiler technology to accelerate advances in quantum programming language design, and CQASM, a programming language for quantum computing that provides a way of expressing quantum circuits as code, are needed to provide a uniform programming interface for application developers. 

The panel of experts also discussed how quantum computers would fit alongside classical high-performance computing (HPC) as a kind of coprocessor. While this approach could be the direction of travel for many quantum computing companies in deploying their machines in datacentres, such an architecture is hugely complex. 

“How the types of problems that we are trying to solve today can be accelerated with a quantum computer is still unknown,” said QuEra’s Keesling.

Factors like considering the application code that can be sent to a quantum computer with code that is best run on HPC architecture, and orchestrating between these different classes of compute resources, are questions the industry will need to answer if a hybrid approach is adopted.

There has been plenty of research into quantum computing that suggests the goals of the industry are achievable, at least from a technology perspective. But the message from the panel of experts at the Quantum Datacenter Alliance Forum was that a heck of a lot more work needs to be done before these machines can truly scale in a commercial datacentre setup.