I'm biased because I work at a full-stack quantum computing company (we produce chips and software), but I agree.
Quantum software-only companies are sometimes... strange. Most QC hardware purchases today are made by either purely research institutions (universities, labs, etc.) or research departments of commercial entities. Either way, there are usually public tender descriptions or private lists of requirements.
When you look at public tenders, you will often find seemingly arbitrary software integration requirements like "the QC must support quantum circuit compilation with ABC", where 'ABC' is some product of some quantum software company you never heard of. Then you go research that library, and if you're lucky it's open source, so you dive into their code and realize it's another attempt at building a "universal" standard for some layer of the QC stack, based on some research paper out of some university. The developers of that library would often claim that, despite them having no QC hardware of their own, they can build better software for arbitrary QCs. It's strange and unlikely to be true simply because there is no such thing as standard or even common quantum hardware. There is rarely interoperability between architectures (and I'm only talking about superconducting quantum computers, not other kinds).
And then you have to either comply and tackle a huge integration project, or convince the buyer that the software you have written for your own hardware is actually better suited for your hardware than software written by people who never seen your hardware.
I have a bit of a strange question for you: as a physicist I sometimes really wonder if quantum computing will ever work. Any chance you could elaborate on that? I guess you’re biased but still love to hear your thoughts.
Since nothing in the known laws of physics explicitly prohibit it, I think it's safe to say that quantum computing will some day work, in the same sense that nuclear fusion or interstellar human travel may work. Unlike e.g. faster-than-light travel, which will probably not work.
But in practice? On Earth, with our human society and economics? Hell, I don't know...
One pessimistic view is that QCs, being such complex and expensive engineering projects, will always be like flying cars: useful in theory, but other solutions are "good enough" and are cheaper. Quantum computers have to really, really advance a lot before they can actually solve problems better, cheaper, and/or faster than classical computers. The danger is that funding will end before we reach that point.
Another important point is that most progress seems to be in specifically superconducting quantum computers (IBM and the company I work at both produce QCs of this type). Superconducting QCs are very hard to scale; it's an incredibly difficult task to scale from hundreds of physical qubits (best chips today) to thousands/millions (necessary for error correction). I'd say it's more difficult than the task of colonizing the solar system. But there are other approaches to building quantum computers. One promising theoretical idea is topological QC [1]. Microsoft is betting on this; it it works, it may make superconducting QCs idea obsolete, like transistors made vacuum tubes obsolete.
I keep comparing QCs with interstellar travel because it's equally hard to answer questions about the possibility of the latter when human seemingly still can't figure out how to stop killing each other.
Anyway, I think humans made impossible things possible in the past. Splitting the atom, sending probes to other planets, landing on the moon, discovering new particles, etc. So I'm optimistic in general.
P.S. I gave a talk on quantum computing hardware & software recently; the organizers uploaded it to YouTube as unlisted, and want to make it public later. If anyone's interested, send me an email to hello[at]rakhim.org, and I'll send you the link.
The way you describe it it’s a big bet wether or not we’ll see QC working within our lifetime. But who knows… I love your optimistic stance.
Re other forms of QCs: what surprised me was that Sabine Hossenfelder of all people became more bullish on photonic QCs. Maybe progress will be faster after all.
One of the biggest problems with superconducting QCs is qubit connectivity. Ideally you want to be able to have all-to-all connectivity, because you want to apply multi-qubit gates on any subset of qubits. With superconducting QCs, qubits are literally connected physically with (kind of) wires. Each connector is another qubit-like component which is driven by classical analog instrument, which again requires a wire going into the chip. It's basically impossible in practice to build a fully connected 100 qubit chip today, for example.
I don't know a lot about photonic QCs, but it at least "solves" this issue: you can apply a signal on multiple qubits at once and you don't need to build a physical connection between them. You also don't need near absolute zero temperatures.
EDIT: sorry, I was thinking of trapped-ion QCs, not photonic!
The company I work at is developing (among other things) chips with a so-called computational resonator[1]: think of a long "bus" to which multiple qubits are connected. As a result, you essentially get all-to-all connectivity with intermediate state moves.
Speaking of talks, here is an older presentation I gave about "oh no, quantum computers will break cryptography, danger danger!: https://www.youtube.com/watch?v=H6ANtrjbqN4
Quantum software-only companies are sometimes... strange. Most QC hardware purchases today are made by either purely research institutions (universities, labs, etc.) or research departments of commercial entities. Either way, there are usually public tender descriptions or private lists of requirements.
When you look at public tenders, you will often find seemingly arbitrary software integration requirements like "the QC must support quantum circuit compilation with ABC", where 'ABC' is some product of some quantum software company you never heard of. Then you go research that library, and if you're lucky it's open source, so you dive into their code and realize it's another attempt at building a "universal" standard for some layer of the QC stack, based on some research paper out of some university. The developers of that library would often claim that, despite them having no QC hardware of their own, they can build better software for arbitrary QCs. It's strange and unlikely to be true simply because there is no such thing as standard or even common quantum hardware. There is rarely interoperability between architectures (and I'm only talking about superconducting quantum computers, not other kinds).
And then you have to either comply and tackle a huge integration project, or convince the buyer that the software you have written for your own hardware is actually better suited for your hardware than software written by people who never seen your hardware.