IBM, the Quantum Food Court

Brian N. Siegelwax
5 min readSep 1

But, standalone quantum restaurants exist.

I recently described IBM Quantum as a food court. With the hardware, the simulators, the portal, and the Qiskit framework, there is nowhere else you can go and find that much variety. A good rule of thumb is that if you have a craving for something, you can probably find it there. The same holds true for New Jersey diners, but the concept of a food court is probably more familiar to most readers.

Way-Too-Long Lines

A growing problem with IBM Quantum, which they are very well aware of, is that this food court has become very popular. And if you go through the main entrance, there is a fair-share queueing system. In principle, if you’ve never been there before, you get to move up in line. In practice, though, I know someone whose first-ever circuit took 2–3 weeks to run. In 2023, I’ve cancelled most of my jobs, because once I stand in line the line keeps getting longer and longer. Over the course of a few hours, my projected wait usually gets longer, not shorter.

Cutting In Line

Admittedly, I could cheat. I don’t have to use the “open” program, but I’m not inclined to use projects for a boost if I’m not actually working on a project. That would be like using a walking cane to cut in line if you’re otherwise healthy. Not only would that be abusive of the privileges that are granted, it’s not always helpful anyway. Projects also have some priority system, such that I might be further ahead in line than I would otherwise be, but my jobs still aren’t running.

Smoke and Mirrors

Also, I never want to use projects for demonstrations. I don’t know if you have any special access, but I do know that you have “open” access. I explain how fair-share queueing works, but I show you what I know you can use, instead of potentially teasing you with queue positions you’re not likely to see. Besides, during a 90-minute talk I can’t get an IBM Quantum job to complete no matter what I do. In fact, I always have two identical copies of a Jupyter notebook: one to submit at the start of the talk in hopes of it completing, and one that ran a long time ago so I can show what the results should look like if I could actually do a live demonstration in…