During my 3.7 years at the company, we had dozens of sales of full-stack quantum computers, chips + cryostat + control electronics + software, ranging from 5 to ~100 physical qubits. Naturally, the buyers are mostly research institutions and univercities who need a real quantum computer to do research on quantum computation and simulation (to a lesser degree with the superconducting technology, though).

(Often the research is done purely within clasically-simulated quantum computers, i.e. virtual QCs, but to verify and make the research publishable they need to run at least partial sub-problems on a real chip.)

Another, smaller, market is HPC centers. They buy and install quantum computers into existing HPC infrastructure because a) they have a few customers who need it (sometimes those same research institutions/universities), and b) they need to solve the integration problem for the future when QCs are actually used for real-world problems and big customers come to HPCs to run both classical and quantum high-performance jobs.

Here is an excerpt from my book I linked above, just to give a bit more context:

---

Since quantum computers are essentially analog devices that allow you to control, in a limited fashion, a set of quantum objects, you can do some research in foundational quantum physics. [...] Still, given the current state of the industry, classical computers outperform most quantum systems. But the research applied to smaller QCs can be scaled once the hardware scales.

Of course, the main area is quantum computing itself. From abstract, mathematical notions of algorithms to very low-level questions of calibration, many universities and research organizations are eager to have a quantum computer available to prove their theories and discover new properties. Commercial companies that deal with material science, battery technology, agriculture, and chemistry are buying quantum computers (or at least buying access to one) because they want to be ready if and when truly large-scale QCs become available. [...]

And finally, integration research. This is the least known and least discussed topic in the industry but is very important. Its significance is one of the motivations for writing this book. Quantum computers, being research tools, are not normal products. They are driven by software, like anything else, but this software changes rapidly and is rarely written with long-term evolution in mind. If you buy a quantum computer today, chances are your code will not work on any other quantum computer, or even on the next iteration of the same machine. At the same time, researchers often need to work with multiple types of machines simultaneously, and HPC (high-performance computing) centers, i.e. supercomputing data centers, want to integrate quantum computers into their existing infrastructure and provide a "quantum compute" service to their users.