IBM is intending on offering a commercial quantum computing service over the cloud that they are calling “IBM Q”. This service will be a logical extension of the IBM Quantum Computing Experience that they started to provide last year for learning and academic usage. This move was not totally unexpected as we expect several other organizations will also offer such services in the next few years including Google, Rigetti, D-Wave, Microsoft, and potentially a few others. Details on the actual timing and milestones were not announced but they have indicated that the first steps will start later this year.
Here are some key takeaways that I see in the announcement:
1. IBM is initially targeting a 50 qubit machine for this service, but they have only indicated that such a machine will be available “in a few years”. A 50 qubit machine would indeed be useful and could potentially start solving commercially interesting problems that cannot be3 solved on a classical computer. For comparison, the current IBM Quantum Experience offers a 5 qubit capability. This is good for teaching purposes, but not commercially interesting since 5 qubits can be easily simulated on a standard laptop computer.
2. The IBM Quantum Experience also has a simulation capability which is currently limited to 5 qubits, same as the real machine. IBM has indicated that they will upgrade the simulation capability in the near term to 20 qubits which will be a good stepping stone to get users started on developing more powerful quantum programs before the 50 qubit machine is on-line.
3. Perhaps the most interesting part of the IBM announcement has received the least amount of external press, is the announcements of the QISKIT API, SDK, and OPENQASM (assembly language) toolkit. These Python based software tools, available on Github, allow a user to develop a program on their laptop and then upload, run, and display the results on their own computer. No longer will a user be required to log into the Quantum Experience web site, use their graphical interface to develop a score, and see the results on their screen. Although that arrangement is great as a learning tool for a 5 qubit system, it would rapidly become too cumbersome for development of real programs with larger numbers of Qubits.