I've sat through three vendor pitches in the last year that used the word "quantum" before they used the word "problem." That's the tell. When the technology leads and the use case trails, you're being sold a future, not a product.
So let me be blunt about quantum AI. The research is real and some of it is beautiful. The business case for your company, this decade, is almost nil. And the genuinely urgent quantum-related task on your plate isn't AI at all — it's your encryption. More on that below.
What the machine actually does
A classical bit is a 0 or a 1. A qubit, courtesy of superposition, can be a weighted blend of both until you measure it. String enough qubits together with entanglement and certain calculations collapse from "heat death of the universe" to "tractable."
The keyword is certain. Factoring large integers — yes. Simulating quantum systems — yes, obviously, you're using physics to model physics. Some optimisation — maybe. Your CRM, your data pipeline, your fraud model? No speedup whatsoever. A quantum computer is not a faster classical computer. It's a different machine that happens to be spectacularly good at a narrow class of problems and useless or worse at everything else. Anyone who blurs that distinction is either confused or selling.
Where quantum and AI actually touch
There are a handful of real intersections, with very different time horizons.
Molecular and materials simulation is the strong one. Drug discovery and battery chemistry involve simulating quantum systems, and that's the home turf. If quantum computing pays rent anywhere first, it's here. Pharma research groups are already running hybrid experiments, and that's sensible.
Optimisation is the speculative one. Plenty of AI systems have an optimisation problem buried in them — routing, scheduling, parameter search. Some of these might run faster on quantum hardware eventually. "Eventually" and "might" are carrying that whole sentence.
Quantum machine learning, where the model itself runs on quantum hardware, is a lively research field with very thin practical results. Watch it. Don't budget for it.
The state of the hardware, without the press release
IBM, Google, IonQ and the rest have systems with hundreds to low-thousands of qubits. That number is doing a lot of marketing work, because raw qubits aren't usable qubits. Qubits are appallingly fragile — they decohere if you look at them funny — so most of the machine's effort goes into error correction. The logical, error-corrected qubits you can actually compute with are a tiny fraction of the headline count.
Real fault-tolerant quantum computing is, by the sober estimate of researchers I trust, somewhere between five and fifteen years out. In thirty years of watching technology timelines, "five to fifteen years" has reliably meant "longer than the people saying it hope."
The one thing you should act on now
Here's the part that matters and almost nobody pitches it, because there's no shiny product attached.
A sufficiently large quantum computer will break the public-key cryptography — RSA, elliptic curve — that protects most of what you send over a wire today. That machine doesn't exist yet. But "harvest now, decrypt later" is a real adversary strategy: capture encrypted traffic today, sit on it, decrypt it when the hardware arrives.
So if you hold data that must stay confidential past, say, 2035 — medical records, legal files, state secrets, long-lived IP — you have a problem with a clock on it. NIST has already standardised post-quantum algorithms. The migration is non-trivial and your security team should be scoping it this year. This is the actual quantum priority. It's boring, it's plumbing, and it's urgent.
My advice, in one paragraph
Don't put quantum on your AI roadmap. If a consultant tells you to, ask them for one concrete problem you have that a quantum computer solves today; they won't have one. Do start the post-quantum cryptography conversation if you hold long-lived secrets. Keep half an eye on hybrid classical-quantum approaches, where a classical machine runs the workflow and hands the quantum chip the one slice it's good at — that's where the first real wins will land. And if you're in pharma or materials, find the serious groups and pay attention.
The pattern repeats with every frontier technology: hyped five years early, written off three years too soon, then it delivers on the thing nobody was excited about. We're in the written-off phase on quantum's hype and the build phase on its physics. Both can be true. Just don't let anyone rush you onto a roadmap that doesn't exist yet.
