PsiQuantum envisions a datacenter-sized quantum computer

We're promised less science fiction, more contemporary hardware

Startup PsiQuantum has a rough vision of what its one-million-qubit quantum computer could look like.

"It's going to look like a big concrete building with a whole bunch of modules," Pete Shadbolt, chief scientific officer, told The Register.

"Inside those modules, a bunch of silicon chips, half photonic, half electronic, and the whole thing is wired together with the same optical fiber that you find today."

PsiQuantum is thus betting that silicon photonics will play a central role in its data-center-sized quantum computer.

"When you think of quantum computing, you think of milli-kelvin temperatures, atoms flying around in space, atomic-scale fabrication, mad materials, science-fiction stuff. We need to repurpose something that's already manufacturable," Shadbolt said.

We need to repurpose something that's already manufacturable

The system will be based on components made using today's manufacturing techniques, and won't require massive cooling refrigerators, we're told. "If you look at it, as a casual observer, it's like a big industrial facility with some steam coming off the top," Shadbolt said.

The big question is whether the ambitious system will ever see the light of day. While the biz couldn't provide a specific date on when the system will land, it hopes the silicon photonics approach will expedite the commercialization.

PsiQuantum's approach is one of many quantum architectures being chased by companies including IBM, Google, and Microsoft, and startups that include Rigetti and IonQ. Quantum computing is highly speculative given these companies are trying to commercialize systems rooted heavily in academic concepts. Some quantum systems are available in the cloud. Most companies are focusing on building systems with error correction that are scalable and reliable.

In quantum computers, information is encoded in quantum bits, or qubits. This, it is hoped, allows quantum computers to quickly solve large problems that would be infeasible for conventional computers. Error correction is required to deal with the finicky nature of qubits.

"You build some lattice-like fabric of qubits that are entangled together, then you do measurements on those qubits," Shadbolt said. Those measurements drive the algorithms you wish to run, "and also implement the error correcting code, checking for errors and fixing it," he said.

Avoiding errors

PsiQuantum's road to an error-corrected quantum computer specifically involves silicon photonic modulators, optical networking fiber, and other components.

"The idea is we'll take the same manufacturing processes that we used to make transistors. Instead, we'll make optical waveguides. And we'll put light inside the chip. And then we can manipulate that light with a toolbox of components," Shadbolt said.

"At the physical level, it's really single photons propagating in the same way that you'd find in a data center."

It's been shown that photon beam splitters can be used to build a universal quantum computing system.

"That's kind of the starting point," Shadbolt said. "There's a whole ton of complexity on top about that turns that into a computer, and it's going to be building-scale, high-performance computer-like system, and lots of silicon, lots of optical fiber in that system."

Like PsiQuantum, IBM is looking to build a one-million qubit system, which is aimed for release by 2030. But cooling could be a limitation for such a superconducting system, given they're operating at a hundredth of the temperature of deep-space, Shadbolt said.

"Photons don't feel heat. We do use some cryogenic cooling systems, but nowhere near as much," Shadbolt claimed. "Our qubits undergo photonic loss, they fall out of the waveguide but they don't really feel heat, they don't feel electromagnetic interference."

Connectivity is also a consideration in building the data-center scale system as it won't be possible to fit a million qubits on a single chip.

"You need to network chips together. You can't just use Ethernet, you need a quantum interconnect that can send qubits from one chip to the other chip. And the only good way to do that is with light, with photons," Shadbolt said.

The startup is getting the system components, such as single-photon sources and single-photon detectors, for its quantum system made via chip manufacturer GlobalFoundries. PsiQuantum said it already had the control electronics required for qubit coherence.

Intel is, meanwhile, targeting a quantum computer based on quantum dots that can be made in its factories.

Go public or not?

PsiQuantum is still privately held even after Rigetti went public and D-Wave followed suit via deals with blank-check companies called SPACs.

PsiQuantum has so far raised $665m in funding, with the last round raking in $450m from the likes of organizations that include Microsoft-backed M12, Blackbird Ventures, and Temasek.

Venture capital firms are pouring billions into quantum-computing companies, hedging bets that the technology will pay off big time some day as conventional computing reaches its limits.

Supernova Partners Acquisition Company II, a finance house that merged with Rigetti in a $1.5bn deal, stated that Rigetti's quantum technology is scalable, practical, and manufacturable.

PsiQuantum has customers and partners in the financial, pharmaceutical, energy, and automotive worlds to help commercialize its quantum plans. The company may sell time on its quantum computer via the cloud. But the high-risk, high-reward nature of the quantum computing space hinges on getting a commercial product to market quickly, Shadbolt said.

"We're building a data-center-like system – we've got to lay concrete, we've got to put up steel beams, and that's going to take time. But a sort of safe, immediate answer is that middle of the decade, we'll have stood up all the manufacturing processes. And shortly after that, we'll have a quantum computer," Shadbolt promised. ®

Broader topics

Other stories you might like

Biting the hand that feeds IT © 1998–2022