If you want to know where a technology really stands, skip the keynotes and read the procurement documents. Keynotes tell you what an industry hopes. Procurement tells you what it has decided.

This month we published the State of Quantum 2026, the fourth edition of the report we produce each year, this time with The Quantum Insider. It is eighty pages of the least glamorous evidence in technology: contract records, deployment counts, survey data from the people who actually sign purchase orders. The analysts tracked 121 quantum computing contracts since 2021, worth a combined $2.6 billion, with annual volume up more than sixfold. More than fifty quantum systems are now coupled to supercomputers across fourteen countries. Capital followed the same curve: quantum computing companies raised $8.3 billion in 2025, nearly five times the year before.

Those are good numbers. But the most important finding in the report is a change in the questions buyers ask. And I think it marks the moment this industry grew up.

For most of quantum computing’s commercial life, the public conversation ran on one question: how many qubits? It was a simple scoreboard, easy to report and easy to game, and it told you almost nothing about whether anyone could do useful work. Underneath it sat a harder problem, a genuine catch-22. Buyers wanted proof before they deployed. But the proof could only come from deployment: from real machines, run by real teams, on real problems. Somebody had to go first without the comfort of a finished case study.

Somebody did. While the headlines counted qubits, a quieter group of institutions started buying machines. National laboratories. Supercomputing centres. Universities. Even Enterprise Companies. The boundary spanners of every technology generation, the ones whose mandate is to learn in public. The report shows what they have built: a procurement record that no longer looks like experimentation. It looks like infrastructure.

You can hear the difference. This year, the buyers interviewed for the report barely mention qubit counts. They ask whether they can see into the machine. Whether they can access calibration data. Whether the system connects to the supercomputer they already run. Whether their own people will learn to operate it. One supercomputing director put it simply: nobody wants a black box in the middle of their ecosystem.

I notice the same shift when I walk through machine halls. At the Leibniz Supercomputing Centre near Munich, where four of our systems run alongside one of Europe’s great supercomputers, the conversations are about calibration windows, job schedulers, and queue integration. In Bologna, in Galicia, in Espoo, the same unglamorous vocabulary. To my eye, that is the most encouraging sign in this entire industry. Technologies become real when their conversations become boring.

So the question has changed. It is no longer whether you can get access to a quantum computer. Almost anyone can, through the cloud, in an afternoon. The question is whether you can build something useful around one: the trained team, the algorithms written for your own problems, the integration with the computing you already run, the operational experience that only comes from running the machine again and again. Access is available to everyone, instantly. Capability belongs to whoever builds it, slowly. Those are very different things, and the gap between them is where the next decade of this industry gets decided.

I want to be straightforward about where IQM sits in this story, because we did not arrive at it by accident.

When we founded the company in 2018, in Espoo, Finland, the fashionable model was access: quantum as a remote service, rented by the minute, with the machine kept far away behind the vendor’s curtain. We made a different bet. We believed institutions would eventually want quantum the way they want every other layer of serious computing: on their own floor, under their own control, connected to their own infrastructure, understood by their own people. So we built full machines that customers could own, operate, and build on, and we committed to delivering them.

I will be honest, it was the harder road. Delivery is a discipline with no shortcuts. A machine has to be manufactured, shipped, installed, calibrated, integrated with the host’s computing environment, and supported for years. The customer’s team has to be trained. The system has to keep working when the installation crew flies home. None of that fits in a press release, and for several years it was fair to ask whether we had chosen the slow way. Cloud demos scale in minutes. Loading docks do not.

But something happens when you deliver machines instead of demos. Every installation teaches you. The supercomputing centres that bought our systems did not behave like customers; they behaved like co-developers, finding the bugs, demanding openness, pushing the integration deeper, telling us what a quantum computer has to become before it can live inside national infrastructure. The calibration tools, the HPC integration that now lets our systems run as ordinary nodes inside a supercomputing environment, the training programs: all of it was built with the institutions running the machines, not imagined for them. We were not just selling into the ecosystem. The ecosystem was building us back.

Inside the company we came to call the result production quantum. The report’s findings suggests the market has now voted on the model, whatever name anyone gives it. Across the buyers surveyed, roughly 46 percent expect on-premises systems to form part of their access model over the next three years, against 24 percent planning on public cloud alone. Openness and calibration access have become procurement criteria. Sovereignty has become a formal requirement in tenders from Europe to the Gulf. And the transaction data shows where the demand went: IQM accounts for 19 percent of all tracked quantum computing contracts since 2021, the largest share of any vendor, and nine of the quantum computers installed at national supercomputing centres since 2025, three times the next vendor.

I read those tables with more relief than pride. For years this was a thesis. Now it is a procurement record. The model we believed in is simply how serious institutions buy.

Look at what they are buying, because the pattern matters more than the totals. Finland’s VTT is climbing a ladder from a 5-qubit training system toward a 300-qubit machine in 2027. Munich is adding a 150-qubit system to the ones already in production. Spain is pairing a research-grade system with a training machine so its people grow with its infrastructure. These are not pilots. They are staged, multi-year capability programs, planned the way nations once planned their first supercomputing centres. The institutions are not waiting for the technology to be finished. They are building the capacity to use it the moment it is.

Which brings me to the part of the report I would ask every CEO to sit with.

Among the industry respondents, 89 percent are already hands-on with quantum. Only 3 percent have reached scaled deployment. The Quantum Readiness Index at the heart of the report scores the global buyer base at 58 out of 100: engaged, budgeted, and not yet converted. And the most cited constraint, by a wide margin, is people. Training a quantum team takes two to five years. The hardware roadmaps now converge on fault tolerance between 2029 and 2031, though honestly nobody knows the exact date and I would be careful of anyone who claims to. Put those two timelines side by side and the arithmetic does the arguing for me. If the people take five years and the hardware takes four, the hiring decision is already late. The technology keeps its own calendar. Readiness keeps yours.

We watched this movie recently. The organizations that built AI capability early did not just get a head start; they shaped the tools, the standards, the architectures, and the talent market that everyone else now pays a premium to enter. The technology optimized for whoever showed up. Quantum is following the same script, with one difference that I find genuinely hopeful: this time the pattern is visible in advance, documented in transaction data and buyer interviews, published for anyone willing to read it.

I said at the Wall Street Journal CEO Council this month that most boards are managing the wrong risk. Acting early is a line item in a budget: a small team, one use case run on your own data, a machine’s worth of learning. Acting late is years of catching up that no budget can compress, in a market where the early movers wrote the standards you now conform to. The report puts evidence under that sentence. None of the early steps bet the company. All of them build what cannot be bought later.

IQM becomes a public company this summer. From that day, the market will form its own view of our work, as it should; that is what public markets are for, and we will make our case with the same evidence I have used here. What I can say today is what the data already shows. The machines are delivered. They are running inside the scientific infrastructure of fourteen countries. And the institutions operating them are quietly building something the rest of the market will eventually need to buy, or to become.

Eight years ago we bet that the quantum era would not begin when the technology works perfectly in a lab. It begins when institutions own it, operate it, and build on it. By that definition, the era has already started. The only open question is the one I keep asking the leaders I meet, and it has nothing to do with physics.

Will you lead, or follow?