Quantum computing is in many senses the holy grail of high-performance computing. The technology promises to dramatically accelerate simulations, machine learning, and perhaps most terrifyingly, render modern encryption schemes mute. At least that’s the hope – or fear – depending on how you look at it.

And like the holy grail, making good on these promises remains elusive. While progress is being made in the field, many unsolved challenges remain, according to Peter Shadbolt, chief strategy officer at PsiQuantum.

“We’re at a really exciting time for quantum computing where lots of people have been successful building small systems on handfuls of qubits,” he said, adding that the problem is that while these projects have been successful, for them to be commercially viable at scale, they still need error correction. And that will require millions of qubits.

For those who aren’t familiar, qubits are the fundamental building blocks on which quantum computers run. They’re the quantum equivalent of the bits used by conventional computers, but in addition to being represented as a one or a zero, they can also be in a superposition of both. They're like the Schrodinger's cat of the computing world.

They’re also really difficult to work with. “The qubits all have an error rate on the order of a fraction of a percent. That’s much, much worse than the transistors in your cell phone,” Shadbolt said. And that’s before you take into account all of the exotic materials and cryogenic temperatures necessary to make them work, he added.

The biggest problem in quantum is manufacturing

Unlike other quantum computing startups focused on commercializing small-scale systems, PsiQuantum aims to build the first million-plus qubit quantum computer.

“We haven't gone and built a small system of 100 qubits and hooked it up to the internet in the way that many of our competitors have done,” Shadbolt said. “We've invested all of our time and money into an architecture and into the manufacturing processes, packaging, assembly challenges that we think are actually going to carry us to that valuable endpoint."

He explained that increasingly, the biggest challenges facing quantum computing don’t have anything to do with qubits. They aren’t “quantum physics problems, they're not science problems that belong in a university lab, they are manufacturing problems.”

Solving this problem isn’t a small task, Shadbolt admitted. These systems often require exotic materials, extremely low temperatures, strong vacuums, and atomic-scale fabrication capabilities. But PsiQuantum has a plan to address these issues.

Earlier this year, PsiQuantum entered into a formal partnership with semiconductor manufacturer GlobalFoundries to tackle this challenge head-on.

“You really can’t get more mature than building things in a 300-millimeter [wafer] line with someone like GlobalFoundries,” Shadbolt said. “There’s nowhere to go beyond that.”

Anthony Yu, VP of computing and wired infrastructure at GlobalFoundries, views the collaboration as a win-win for both companies.

“From my perspective, what makes this kind of a unique arrangement is Pete is using our commercial process, which is based on data [communications] plugables and transceivers, but we’re adding unique materials to the processing for things like single photon detectors,” he said. “He's actually challenging us to be better … because their requirements are more stringent than the data centers'.”

Additionally, developing these technologies could give GlobalFoundries a leg up once technologies like co-packaged optical switches start hitting the market.

“Quantum computing is in the path of photonics,” Yu said, adding that many of the same technologies used in PsiQuantum’s quantum computers are necessary to bring co-packaged optics to market.

Building a million-qubit quantum computer

PsiQuantum is now producing the first of several key components, a single-photon detector, at volume in GlobalFoundries’ facilities. The detector takes signals transmitted as photons – in other words, light – and converts them into an electrical signal.

“Instead of controlling the passes of electrons through field-effect transceivers … we're talking about controlling photons, bringing light on and off chip, controlling the light through waveguides, making sure you're not losing light, and then detecting light and detecting photons and converting them to electrons for processing,” Yu explained.

While a significant milestone for the company, Shadbolt said that an actual quantum computer based on this technology remains a ways out. Over the next few years, PsiQuantum and GlobalFoundries aim to develop new low-loss waveguides and high-performance switch modules using new materials designed specifically for quantum computers.

“We’re just now starting to integrate the blocks together,” Yu said.

Once these pieces integrate onto a single package, the challenge becomes scaling that up, Shadbolt said.

“When you recognize you need a million qubits, immediately you're talking about many, many silicon chips that are hooked together to build that system,” Shadbolt said. “You're talking about a machine that is going to be like the size of a building. It's going to look like a regular supercomputer in many respects.”

Today, the largest part of most commercial quantum computers is the cryogenic plant it’s attached to, he explained. They’re limited to a single chip, running at a couple of qubits.

As to when PsiQuantum’s first quantum supercomputer will come online, the company remains optimistic it’ll be sooner rather than later. “We believe that by the middle of this decade, we should be able to stand up all the manufacturing processes that we need to be able to build that million-qubit quantum computer,” Shadbolt said.