When You Don’t Want to Use the Quantum Cloud

By Yuval Boger

The quantum cloud is the preferred way of accessing quantum hardware. According to a December 2022 survey from Hyperion Research, 50% of surveyed users access quantum through the cloud, and another 17% use hybrid on-prem/cloud access. Still, 21% want on-prem quantum hardware. Significant multi-year partnership deals have been inked, sometimes including installing a quantum computer on site.

Since today’s quantum hardware will be largely obsolete in two years and quantum computers are expensive, the 21% slice is perhaps more surprising than the 50% segment. Those that use quantum computers on the cloud often enjoy “pay as you go” pricing, the ability to choose the computer that best fits their particular application, the ability to experiment with multiple vendors, and the peace of mind that they are likely always to be able to access the newest models.

Given these attractive advantages of cloud usage, what is the motivation for on-prem usage?

Some – especially government, defense, and pharmaceutical organizations – take great pains safeguarding their quantum algorithms. Sending their algorithms outside the organization is deemed an unacceptable risk despite available highly-secure communication links. This risk is highlighted by the fact that several cloud providers that offer quantum computing services don’t host the quantum computer themselves but rather rely on hosting it on the hardware vendor’s premises.

Large research centers sometimes opt for on-premise solutions to guarantee how quickly their quantum jobs get executed. Only a couple dozen quantum computers are accessible on the cloud, and access is limited to the cloud provider’s availability hours and subject to a quantum jobs queue. By hosting their own computer, organizations control the operating hours, don’t need to share this capability with other users, get more quantum bandwidth, and can prioritize jobs based on their internal priorities.

Other organizations worry about data residency requirements and must ensure that their data remains in-country. The organization’s geography also impacts latency. For instance, since the distance between Paris and New York is approximately 5,800 kilometers, a Paris customer that accesses a quantum computer in New York would experience about a 40-millisecond round-trip delay. This is particularly important for hybrid classical/quantum algorithms such as VQE where part of the algorithm executes on the classical computer and the other on the quantum computer.

Last, national and regional financial incentives sometimes make it very attractive to purchase a quantum computer instead of “leasing” one over the cloud.

Whatever the reason may be, what options does an organization have if it does not want to access quantum computers on a public cloud?

One alternative is partnering with an in-country provider that physically hosts classical and quantum computers in their data center. This approach can address the latency requirements – since the classical and quantum computers are co-located. It also provides an option to negotiate guaranteed availability and operating hours. Organizations might find it convenient to shift the responsibility for ongoing calibrations of the quantum computer to the hosting center and ask the hosting center to also deal with operating condition requirements such as cryogenic cooling for those quantum modalities that require it. The downside of this approach is that the data is still outside the customer's walls, so significant trust and verifiable safeguards must be negotiated.

If this compromise is not acceptable, there are several points that organizations can emphasize when negotiating the purchase of a computer from a quantum vendor:

1. Upgradability. Quantum computers continuously improve dramatically. Unlike our mobile phones, these improvements are not just somewhat better cameras and new emojis. Quantum improvements include the number of qubits, improved fidelity, and new features such as dynamic circuits – all taking computers one step closer to quantum advantage. Organizations that buy quantum computers should emphasize architectures that can be continuously upgraded and vendors that are willing to keep the customer at the leading edge.

2. Comprehensive support and training services. Such services should not be limited to just calibration and maintenance. Instead, they can include workforce development or even integration and customization services. Purchasing a quantum computer should be an opportunity to turn the region into a thriving quantum hub.

3. A staged approach. Rarely is a quantum computer available “in stock” for delivery. Instead, building a high-end quantum computer might take a couple of years. What can the customer do in the meantime? Vendors have come up with innovative approaches to bridge this schedule gap.

Quantum computers have the potential to revolutionize industries. Though most organizations prefer the convenience of cloud access, there are arguments in favor of an on-prem machine. In this case, organizations would do well to consider upgradability, support, and the delivery timeline when choosing their quantum partner.

Yuval Boger is an executive working at the intersection of quantum technology and business. His “Superposition Guy’s Podcast” can be accessed on the Quantum Computing Report here and on most audio platforms. He can be reached on LinkedIn or at this email.

January 29, 2023