The tech world may have to make room for a new acronym, perhaps qubits-as-a-service, QaaS, or some such, as Honeywell has introduced what appears to be the first subscription-based plan for quantum computing usage.
With the introduction Thursday of the company's Model H1 quantum computer, with 10 qubits and a logical quantum volume of 128, the company detailed a plan to charge in a subscription fashion based on monthly access to the machines.
The subscriber license gives a company access over the course of a month to blocks of "dedicated time," in two different flavors, standard and premium, with eight hours per month of dedicated time or sixteen hours, respectively.
Also: Honeywell claims to have world's highest performing quantum computer according to IBM's benchmark
The idea of blocks of time is to accommodate the need to provide something like quality time in which to do quantum experiments.
"We're talking about significant time," said Tony Uttley, president of Honeywell Quantum Solutions, in an interview with ZDNet. "That's what we are finding is the most useful right now."
"It's not going to be a five-minute interaction with a quantum computer, it's not even going to be an hour-long interaction," said Uttley. "These are interactions that you want to have tens of hours on before you can demonstrate the kind of utility coming from these systems."
The H1 follows a Model H0 released in June and tested with customers since the fourth quarter of last year. Both models are based on Honeywell's unique hardware approach that makes use of "trapped ions." An ion, of course, is an atom that has a net positive or negative electrical charge. The trap in this case is a fabricated device, like a computer chip, that can be used to manipulate those ions, similar to moving electrons through the gates made up of silicon transistors.
Also: Microsoft's Azure Quantum employs Honeywell quantum hardware
The H1 has been up and running for several months internally at Honeywell, but has been in use by customers for about three weeks, said Uttley. Honeywell has been working with eight enterprise customers, including DHL, Merck, and JP Morgan Chase. Some of those customers had been working on the H0 system and were able to easily "port over" work to the new machine, said Uttley.
One reason for the subscription is that there is still substantial hand-holding that happens. Those windows of time include participation with the customer by Honeywell quantum theorists, and Honeywell operations teams, who work "hand in hand" with customers.
The hands-on approach of Honeywell to customer subscriptions makes sense given that much of the work that customers will be doing initially is to gain a sense of trust, said Uttley. They will be seeing what results they get from the quantum computer and matching those to the same work on a classical computer, to validate that the quantum system produces correct output.
On top of the blocks of dedicated time, each subscriber can get queueing time, said Uttley, where jobs are processed as capacity is available.
"As we have availability of the system, we would just run it from the queue, and so both of those would happen over the course of the month of their subscription."
The initial use cases seem surprisingly conventional. For example, a drug developer like Merck might be running simulations on the distribution network for a new drug.
"Think about COVID-19, and coming up with a vaccine, and how you distribute that, these are systems where you have to understand how can things be refrigerated," offered Uttley.
That may not sound like a quantum problem, but the virtue of the quantum machine is to explore optimization where the solution might be more complex than a search among known variables.
Also: Is the world ready for cross-platform quantum programming?
"What classical computers do really well in machine learning is non-generative work, what should the next outcome be, with lots of factual data," said Uttley. "What quantum computers lend themselves to are more generative models, where I don't have a lot of information."
"And so think about the first use cases as being, what happens to a system of movements when something has been disrupted, as opposed to what do you use to create a more steady-state, balanced system."
Another reason why blocks of time would be useful is that working on a quantum computer may involve a repetitive process of extracting results with each pass and re-submitting those results.
That iterative process, known as mid-circuit measurement, is a special feature of Honeywell's configuration, and it is what allows a system of only 10 qubits at the moment to simulate the gate configuration of a much larger quantum machine, as Dr. Patty Lee, the chief scientist of Honeywell Quantum Solutions, explained.
"We can program in our quantum simulation to the quantum computer, and then we can measure part of it, and get some information out of that, and then put that back into the system, and simulate a much larger system," said Lee.
"Even though we have only 10 qubits, we can simulate systems that have a lot more than 10 atoms or 10 electrons."
As the system gains more stable qubits, larger and larger machines can be simulated, said Lee.
Blocks of time are an arrangement that may be well suited to the fact that customers are on a learning curve with a brand new technology. Although there is an industry taking shape for programming quantum, it's not as simple as spinning up virtual machines or a Kubernetes container.
A software platform called Orquestra from Honeywell's partner Zapata, based in Boston, provides a whole set of middleware that will abstract details of quantum hardware, and let the programmer stage things right from inside of a Jupyter notebook.
Nevertheless, submitting jobs via Orquestra is only one piece of the puzzle. One still has to know how to write quantum algorithms, and how to make use of the unique capabilities of Honeywell's particular hardware.
"It's as much tied into companies building up their own internal capabilities," said Uttley. "Are they going to build their own expertise to write their own algorithms, or are they going to leverage partnerships with a software company like Cambridge Quantum Computing and Honeywell together?" he offered.
"Those come in very different maturity curves, and some of them take a lot longer." Building internal capabilities may take companies many quarters or years to achieve, whereas leveraging partnerships, with existing algorithms, may be a process that comes together within days or weeks.
Subscriptions also allow Honeywell to update its offerings on a regular basis, Uttley emphasized. There is added utility to a usage plan where customers receive new quantum computing hardware in the cloud regularly. Honeywell has a roadmap for a Model H2, a Model H3, and on and on.
"Our customers love the fact that we're doing this as a subscription," said Uttley. "They get access to our system, and then they get to sit back and have it become more and more powerful with each time that they get a chance to use it."
"These systems are designed to be released at a point where our customers can have very high utility from them right now, and then we can continuously upgrade them over the courses of their life to be more and more capable," said Uttley.
"Imagine having a streaming service like Netflix, and a month from now, it's twice as good."
Perhaps streaming qubits is the new rubric to focus on.