IBM Quantum Pricing Models. Credit: IBM

IBM has added a new Pay-As-You-Go pricing model to their quantum cloud service. Previously, a user was either in the Free tier that would allow them to access the simulators and smaller machines at no charge or they were in the Premium tier and a member of the IBM Q Network to allow access to all of IBM’s available quantum processors at a hefty contract price. Now, IBM has introduced a middle tier called Pay-As-You-Go that allow access to two of their more advanced 27 qubit processors. These two processors go by their codename Falcon R5 and can be accessed through the IBM cloud at a price of $1.60 per Qiskit Runtime second. A Qiskit Runtime second includes quantum compute time as well as classical pre- and post-processing time (near-time). Any time waiting for results and in a queue for the quantum computer is excluded from the classical processing time.

Although the Pay-As-You-Go model does not currently include access to IBM’s 65 or 127 qubit processors, we would point out that the 27 qubit processors currently have a higher Quantum Volume rating, so from that standpoint they would actually perform better for many quantum programs. IBM does expect they will update the processor line-up for the Pay-As-You-Go model in the future as they make changes in their available quantum processor fleet.

The target customers to use the Pay-As-You-Go model are more sophisticated users like professors and startups who may not be ready to make an expensive yearly commitment. It’s interesting to note that the Premium pricing model is oriented towards large enterprises that have large pre-determined usage requirements and can afford to pay a minimum of $500 thousand per year with a three-year agreement. The Open Plan still remains and is more targeted for students and other people starting to learn how to program quantum computers and wishing to try it out. The Open Plan still remains and is more targeted for students and other people starting to learn how to program quantum computers and wishing to try it out.

The Amazon Braket also has a Pay-As-You-Go pricing model, but it is hard to compare with IBM’s since it based upon a Per Task and a Per Shot basis, which is different from IBM’s approach.

IBM’s second announcement was an expansion of the Qiskit Runtime system to add the first two new primitives that help simplify the use of the system and allow for managed performance. Qiskit Runtime is a containerized execution environment for classical-quantum programs that can greatly improve the overall execution time of a program by more tightly integrating the quantum and classical resources together.

The first new primitive is called the Sampler which can estimate the entire quasi-probability distribution at the output of a quantum circuit by sampling from its output. It could potentially be used for such things as measuring error mitigation or circuit cutting. The Estimator calculates expectation values of observables at the output of a circuit. It can potentially be used for things like entanglement forging, zero-noise extrapolation, and probabilistic error cancellation.

IBM has posted a blog article that provides some additional details about both of these announcements that can be accessed here.

April 13, 2022