Marin Ivezic
A board is asked to approve seven figures for a machine that, today, loses to a well-configured laptop on every commercially relevant problem. Finance wants to know what the money buys. Quantum advantage, meaning the point at which a quantum computer beats the best classical method on a problem someone actually has, is not the answer. No credible vendor claims it for a 2026 system.
So the business case has to be built out of things that are not compute. Some of those things hold up under review. Several don’t, and the executives we teach are usually the ones asked to tell them apart.
Start with the calendar
Suppose a regulated buyer, a national laboratory or a large bank, wants an operational quantum system in the year its vendor roadmap says the machines get interesting. Work backwards from that date. Vendor evaluation and competitive bidding come first, then contract negotiation, then facilities work, because a superconducting system needs cryogenic plant, power, and floor loading that most data centers were not built for. Security accreditation follows, and so does integration with existing high-performance computing (HPC) infrastructure, which is custom engineering every time. Hiring has to finish before any of the rest produces value.
Put modest figures against each step, a year to select, a year to build out, eighteen months to accredit, and the total passes three years before anyone runs a production circuit, because several of those steps run in sequence rather than in parallel. In a classified environment it runs longer.
The calendar is where we start with every group that asks the question, because it reframes what the decision is. The purchase decision comes years before the capability question is settled, so a buyer who waits for proof of advantage has already chosen to arrive late. Enterprises that started cloud programs around 2006 met the same problem from the other side. The platforms of that period were genuinely unfit for mission-critical workloads, and the organizations that said so were right on the technology. They were also the ones spending 2012 building the identity integration, the security models, and the operating skills that their competitors had finished years earlier.
Three buyers, three ledgers
Almost every purchase we look at falls into one of three categories, and each one is buying something different. Confusing them is how a business case ends up defended with an argument that was never available to the buyer making it.
Governments buying an industry
Much of the state spending is anchor-customer procurement, an instrument borrowed from defense and aerospace. The government buys early units not because it needs the capability, but to give a domestic supplier revenue during the years between working prototype and sellable product, and to tell private investors that demand exists. The structure is consistent across programs: early-phase contracts pay a supplier to build and validate an integrated system, and the larger public procurement follows once the supplier has something to sell.
Running alongside it is a sovereignty argument. Classified, defense-related, and heavily regulated workloads cannot run on a foreign cloud platform, whatever the machine can currently do. Buyers who have lived through supply dependency in other technologies treat an access agreement as weaker than ownership, and they price that difference into the decision.
We regularly see commercial decks borrow this language, and it does not transfer. A bank has no domestic supplier base to seed and no sovereignty mandate to satisfy. Any purchase justified on national-capability grounds by a private buyer is justified on someone else’s grounds.
Teams that need real noise
Classical simulation of quantum circuits gets harder as the circuits get wider and deeper, and long before it becomes impractical it is modeling idealized noise rather than the behavior of an actual device. Algorithms that run cleanly in simulation fail on hardware because of calibration drift, crosstalk between qubits, and connectivity constraints that force extra operations into the circuit.
Hardware in the loop means developing against a physical device rather than a model of one, so that resource estimates, error budgets, and circuit designs are calibrated against real behavior. Groups building chemistry, materials, or portfolio workloads use it to learn what a future machine will cost them in runtime and error correction overhead. Vendors sell early access on exactly that basis, and the industrials and banks who buy it are not getting advantage from it. What they take away is a set of resource estimates calibrated against a real device instead of a model.
What we tell teams is that most of this is available through cloud access, which is still where the large majority of quantum spending goes. Owning the machine adds queue priority, data control, and the freedom to run experiments a vendor would not schedule. Those are worth something. They are worth less than a first-time buyer usually assumes.
Organizations measuring the threat
Harvest now, decrypt later (HNDL) describes an adversary recording encrypted traffic today and storing it against the day a cryptographically relevant quantum computer (CRQC) can break it. For data with a long confidentiality life, the governing question is how long that data must stay secret, not what year the machine arrives.
A small number of defense and intelligence buyers use hardware to measure the timeline directly, running small instances of Shor’s algorithm and testing the error correction codes a future CRQC would depend on. This is the point we make most often to security leaders: for everyone else, the argument does not reach a purchase. The response to HNDL is migration to the NIST standards, ML-KEM (formerly Kyber) for key establishment and ML-DSA (formerly Dilithium) for signatures, and that work needs a cryptographic inventory and a sequenced migration plan rather than a cryostat. NIST’s National Cybersecurity Center of Excellence runs a public migration project covering discovery, inventory, and sequencing: Migration to Post-Quantum Cryptography.
Where the business cases get thin
Three arguments appear in almost every deck and carry less weight than they are given.
Competitive extinction. The claim is that at a few thousand logical qubits – logical qubits being error-corrected units assembled from many noisy physical ones – firms without quantum access will face problems their competitors can solve and they cannot. The first half is defensible for a narrow set of chemistry and materials problems. The second half assumes algorithmic expertise and machine access cannot be bought later, and both can. We hear this one in nearly every deck, usually with the hardware as the proposed remedy. The capability that genuinely cannot be bought in a hurry is the internal ability to recognize which problems are candidates at all, and that is a training question.
Reputation. Being first in a sector or a country generates coverage, board credibility, and recruiting pull, and several buyers say plainly that the press release was part of the case. Those returns are real. We would put them on the marketing line, measured against other marketing spend, and size the program accordingly.
Standards influence. Owning hardware opens doors to advisory boards, national research centers, and consortia. Most of those doors also open for a serious cloud commitment or a joint research agreement at a small fraction of the cost, so influence is rarely a reason to buy a machine.
The line item that doesn’t need a cryostat
Hardware ownership functions partly as a hiring instrument, since researchers stay where the machines are, and specialist hiring is competitive enough that some buyers count it as part of the return. But the shortage the executives in our programs describe is not usually a shortage of physicists.
They are short of people who can state a business problem in terms a quantum algorithm could address, read a vendor roadmap and separate an announced result from a demonstrated one, tell physical from logical qubit counts in a specification, and produce the resource estimate that decides whether a purchase is three years early or ten. None of that requires a device. All of it is required before a device is worth buying, and it holds its value whichever way the procurement decision goes.
That’s our reading of the market. The strongest argument for engaging early is workforce and institutional knowledge, and it happens to be the one argument a buyer can act on immediately, at low cost, without committing capital to a system that may be superseded before accreditation finishes.
If your organization is working through a quantum procurement decision, or wants the people who will evaluate it to be able to read the specifications rather than the press release, our programs cover the technical and commercial fundamentals those teams need. Current offerings and enrollment details are at quantumacademy.com/.