Quantum Riches and Hardware Diversity Are Discouraging Collaboration

Quantum computing is viewed as a technology for generations, and the spoils for the winners are huge, but the diversity of technology is discouraging collaboration, an Intel executive said last week.

There are close to five to six quantum computing technologies being developed, and they all require different levels of expertise or optimization on the top of the system, Jim Clarke, director of quantum hardware at Intel Labs, told HPCwire.

“That makes it hard to have common projects across such disparate programs,” Clarke said.

Intel’s version of quantum bits, which it called spin qubits, differs from other technologies that includes superconducting qubits, which is being chased by the likes of Google and IBM. IonQ, which made news last week with a deal to provide quantum computing services via Dell, is working on trapped-ion qubits.

“Companies view quantum as the technology for the next 100 years and they are interested in sharing the spoils. Companies just are not interested in working [together]. I think this could be one of the flaws. There is not any collaboration across the industry,” Clarke said.

That is unlike how the conventional chip market evolved, spurred by SEMATECH, a DARPA-backed incubator where chip technologies were developed and tested, with the winner being widely adopted.

“We brought together all of these companies to work on pretty competitive ideas. Once you had an idea through this pre-competition, the one that took the idea, adopted it and quickly optimized it was the one that won,” Clarke said.

Intel is working to create its universal quantum computer, and last month produced stable quantum dots in its existing factories. By next year, the company hopes to make its 12-qubit hardware system accessible so developers can start writing applications. Intel has a mass-manufacturing advantage as it has shown it can make quantum chips in its existing factories.

“Our philosophy is to make our qubits exactly like transistors. The closer that we can make these two transistors, both in terms of fabrication and design… and layout, the easier it is going to be for us,” Clarke said.

Beyond the quantum processing unit, Intel’s quantum portfolio includes the controller called Horse Ridge and a software-development kit that includes the compiler, runtime, mapper and scheduler. It has a quantum simulator to sample the computing environment. Intel’s next step is to upgrade the quantum simulator to represent the spin qubit hardware that the company plans to ship.

Quantum computer hardware developers have largely been civil with each other and focused on their own technologies. But it was not always like that. About nine years ago, IBM and D-Wave, which were the earliest quantum hardware developers, had a public battle that questioned the very concept of quantum computing.

IBM, in a 2014 paper argued that D-Wave’s quantum annealing system should not be scientifically classified as a quantum computer, which was counterargued by D-Wave. IBM, which was developing a superconducting qubit, ultimately had to track back and acknowledge that D-Wave’s system was a quantum computer.

D-Wave was a pioneer with its quantum annealing system, which is specifically used of optimization. IBM, on the other hand, was working toward building a universal quantum computer with error correction, which researchers now agree will require a minimum of one million qubits. But technological developments and research has evolved since, with a recognition that quantum hardware and topologies will be diverse, with different types of systems solving different types of problems.

Governments are now trying to bring some order to the chaotic quantum computing industry to expand the markets and preserve national security. The U.S., China and Europe have put quantum computing on their priority lists of technologies to keep inside the country, much like with AI and high-performance computing chips.

The National Quantum Initiative Act of 2018 boosted quantum research and collaboration across 70 organizations, which include academia and private sector companies. The initiative also prioritizes the develop of algorithms to protect the U.S. infrastructure against cyberattacks from quantum systems, which could break current encryption schemes within a matter of seconds. The National Institute of Standards and Technology is testing quantum-resistant algorithms developed by companies that include Intel, Microsoft and IBM.

Argonne National Laboratory is testing quantum hardware and software technologies from many companies through its Q-NEXT program, which lists Intel, IBM, Microsoft, ColdQuanta and others as members. Argonne will test quantum hardware from Intel.

The EU-funded European High-Performance Computing Joint Undertaking (EuroHPC JU) last month announced it will deploy quantum computers in Czechia, Germany, Spain, France, Italy, and Poland. The systems will be plugged into the continent’s wider supercomputing network. The deployment will be completed by the second half of 2023 and cost roughly €100 million.

Last month, Chinese president Xi Jinping listed the development of native quantum computing technologies as one of his top priorities, according to an article in Nature.

Notable companies in the financial, pharmaceutical and transportation sectors are testing quantum chips or simulators to find the best types of qubits for applications. But the confusion around quantum computing resources has given birth to companies like Classiq, which has a software development platform so customers can write quantum applications without knowing the hardware.

“The industry programs at the assembly language, and anybody who’s ever programmed at assembly level knows that that’s not a tenable thing,” said Erik Garcell, technical marketing manager at Classiq, during the SC22 conference in Dallas earlier this month.

Software from Classiq, which is funded by HPE, also allows users to benchmark quantum systems, either against classical hardware or other quantum hardware, so they can find the best qubits. The customer can then deploy the application and get results without learning new languages or data sets.

“We want to shield our users from having to decide ahead of time which hardware I’m going to run on or which assembly language program do I learn. We want them to build efficient quantum circuits and then decide whether to run on AWS or IBM,” Garcell said.