Quantum Computing in Healthcare: A Field Guide for Health System Executives

In March 2023, IBM switched on a Quantum System One on the campus of Cleveland Clinic in Ohio, the first quantum computer installed anywhere for biomedical research. It anchors a ten-year partnership between the two organizations. It is also a machine of a few dozen qubits, which means that for almost every problem a clinician would recognize, an ordinary laptop still wins.

Both of those sentences are true at once, and holding them together is the whole executive skill here. The press coverage tends to pick one. Vendors pick the first, sceptics pick the second, and the health system executive who has to sign a budget gets no useful signal from either.

Three Clocks Running at Different Speeds

A quantum computer stores information in qubits rather than bits. A bit is either 0 or 1; a qubit holds a controlled combination of both, and a machine with enough well-behaved qubits can solve certain narrowly structured problems in far fewer steps than any classical computer. Everything depends on enough and certain. Qubits are fragile, they lose their state within fractions of a second, and errors accumulate faster than useful work gets done.

The fix is error correction, which bundles many unreliable physical qubits into one reliable logical qubit. Current designs need hundreds to thousands of physical qubits for each logical one. Google reported in December 2024 that its Willow processor had crossed the error-correction threshold, meaning that adding more physical qubits made the logical qubit better rather than worse. That was a genuine milestone, and it’s still several engineering generations away from a machine that could simulate a drug binding to a protein.

So there are three clocks in healthcare, and they run at different speeds.

The first is already running and has regulators attached to it. Quantum computers of the future will break the encryption protecting patient records, medical devices, and clinical networks today. That is a migration project, not a research project, and it competes for the same budget as any other security programme.

The second starts when your organization chooses to start it. Pilot work on scheduling, logistics, imaging analysis, and molecular screening is available now through cloud access, at modest cost, with modest expected return. The value is organizational learning more than throughput.

The third is a decade-scale research bet. Simulating molecules accurately enough to change how drugs are designed is the application most likely to justify the whole field, and it is the one furthest from delivery.

Confusing the three is the expensive mistake. We see health systems fund a research bet as though it were a compliance obligation, and postpone the compliance obligation as though it were a research bet.

What Is Actually Installed

The honest inventory of quantum computing in healthcare is short, and it is worth knowing because it sets the scale.

Cleveland Clinic’s Discovery Accelerator with IBM is the flagship. It puts a machine on a medical campus, gives researchers standing access, and trains people. IBM and Moderna announced a joint programme in 2023 applying quantum methods alongside generative AI to mRNA research. In the United Kingdom, a quantum biomedical sensing hub was funded in 2024 as one of five national quantum research hubs, aimed at early diagnostics rather than computation.

They are research partnerships with ten-year horizons and named academic leads. None of them is a product. None has changed a care pathway. What they buy the participating institutions is a decade of accumulated familiarity, which is a defensible reason to join one and a poor reason to expect near-term clinical output.

Everything currently deployed runs on what the field calls NISQ hardware, for noisy intermediate-scale quantum: machines large enough to be interesting and too error-prone to be trusted with a result you can’t verify some other way. That constraint governs every near-term use case below.

The Clock That Is Already Running

If you read one section, read this one. It is the only part of quantum computing that belongs in a healthcare capital plan today, and it has nothing to do with running quantum algorithms.

Why Health Data Ages Badly

Encryption protects data for as long as the underlying mathematics holds. RSA and elliptic-curve cryptography, the two schemes protecting most clinical traffic and stored records, are both broken by a sufficiently large fault-tolerant quantum computer. Nobody has built one. That doesn’t settle the question, because of a pattern the security community calls harvest now, decrypt later: an adversary copies encrypted traffic or archives today, stores them cheaply, and decrypts them once the hardware exists.

For most industries this is a manageable risk, because the stolen data goes stale. A stolen credit card number is worthless in eight years. A genome is not. Neither is a psychiatric history, an HIV status, a fertility record, or a sealed adolescent diagnosis. Health data has the longest useful life of almost any category of personal information, which puts healthcare near the front of the queue for this specific threat regardless of when the machine arrives.

The practical question for an executive is not “when is Q-Day.” It is arithmetic you can do in a meeting. Take the number of years a given record must stay confidential, add the number of years your organization will need to replace the cryptography protecting it, and compare the total against your own estimate of when capable hardware appears. If the sum exceeds the estimate, you’re already late for that record type.

The Standards Are Finished

The replacement algorithms exist and are no longer draft. Post-quantum cryptography means encryption designed to run on ordinary computers while resisting attack by quantum ones. On August 13, 2024, NIST published three final standards:

• FIPS 203, ML-KEM (formerly CRYSTALS-Kyber), for establishing shared keys, which is the workhorse
• FIPS 204, ML-DSA (formerly CRYSTALS-Dilithium), for digital signatures
• FIPS 205, SLH-DSA (formerly SPHINCS+), a signature scheme built on different mathematics as a hedge

A fourth signature standard, FN-DSA, was still in draft when these three were finalized. For a health system, ML-KEM and ML-DSA are the two names that will appear in vendor contracts and procurement language. ML-KEM is already deployed in mainstream browsers and libraries for hybrid TLS key agreement. ML-DSA support is emerging in cryptographic libraries, but post-quantum signatures aren’t yet accepted in the Web PKI, so adoption remains uneven.

This changes the character of the conversation. Until 2024, waiting was defensible because the target was moving. It has stopped moving. A vendor who cannot describe a migration plan in these terms is behind the standards, not ahead of the risk.

Medical Devices Are the Hard Part

Enterprise IT will handle its own migration reasonably well, because operating systems, browsers, and cloud services are updating underneath it. Clinical technology will not.

MRI scanners and linear accelerators routinely stay in service for 10 to 15 years. Infusion pumps, monitors, and imaging workstations run embedded software that is patched slowly if at all, and often carry cryptographic keys burned in at manufacture. A device bought in 2027 with a fifteen-year service life is a device you are committing to operate deep into the 2040s. Whether it can accept new cryptography is a question to ask before purchase, because after purchase the honest answer is usually no.

Two regulatory currents make this tractable rather than hopeless. Since 2023, United States law has required manufacturers of internet-connected medical devices to submit a software bill of materials, an itemized list of the software components inside the product, as part of premarket submissions. And in January 2025, the Department of Health and Human Services published a proposed update to the HIPAA Security Rule that would make asset inventory, network mapping, and encryption explicit requirements rather than addressable ones. The proposal was not final at the time of writing, and the direction of travel is unmistakable.

Both point at the same artefact: a cryptographic bill of materials, or CBOM. It extends the software inventory idea to cryptography, recording which algorithms, key lengths, certificates, and protocols each system depends on and where they live. Without one, “migrate to post-quantum cryptography” is an instruction with no addressee. With one, it becomes a scheduled programme of work that can be sequenced by data sensitivity and device replacement cycle.

Practical note on scope. In a mid-sized health system, the inventory will surface cryptography in places nobody expected to own: pharmacy automation, building access, cold-chain monitoring, badge printers, laboratory instruments talking to a middleware layer nobody has touched in six years. Budget for the surprise. Every organization we work with finds one.

Where the Near-Term Pilots Live

The second clock. These are worth funding at pilot scale, with clear eyes about the return.

Operations and Scheduling

Staff rostering, theatre scheduling, bed allocation, and fleet routing are combinatorial optimization problems, and optimization is a plausible early target for quantum hardware. It is also a mature classical discipline, which is the part most vendor decks omit.

The phrase quantum-inspired describes classical algorithms that borrow mathematical structure from quantum physics and run on ordinary servers. Quantum-inspired methods are real, they occasionally beat conventional solvers, and they involve no quantum computer at any point. That’s not a criticism. If a quantum-inspired scheduler cuts your theatre turnover, take the win. Just don’t record it as evidence that quantum computing works, and don’t pay quantum prices for it.

The credible near-term posture on operations research is to improve your classical optimization first. It builds the data pipelines, the objective functions, and the institutional trust that any future quantum approach would need anyway. Organizations without a working classical scheduler have no path to a quantum one.

Imaging and Pattern Recognition

Quantum machine learning is an active research area with no clinical results. Published demonstrations run on datasets small enough to fit current hardware, which means far smaller than any radiology study. There’s no evidence today that a quantum model detects anything in a scan that a well-trained classical model misses.

Meanwhile, quantum sensing is a genuinely separate technology that shares only the adjective. Magnetometers based on quantum effects are being investigated for magnetoencephalography and other functional measurements, and they may reach clinical use before quantum computers do anything for medicine. If a proposal crosses your desk with “quantum” in the title, establish first whether it concerns computing or sensing. The maturity, the vendors, and the regulatory pathway are all different.

Molecular Simulation

This is the application that makes the field interesting to life sciences. Chemistry is quantum mechanical at its base, so simulating how a candidate molecule behaves is a natural fit for a quantum computer in a way that scheduling never will be.

It’s also the furthest away. Published resource estimates for industrially relevant molecules run into the millions of physical qubits, against machines that today count in the hundreds. Pharmaceutical partnerships in this area are buying algorithm development and talent, not compute. That is a rational purchase for an organization whose R&D pipeline is measured in decades. It’s not a rational purchase for a regional hospital network.

What to Discount in a Vendor Pitch

A short list of claims that appear regularly and do not survive contact with the literature.

“Grover’s algorithm will search patient records exponentially faster.” Grover’s algorithm is a real quantum search method, and its advantage is quadratic, not exponential: it turns a million steps into a thousand, not into ten. Realizing even that requires a fault-tolerant machine and loading the whole dataset into quantum memory, which is itself an unsolved problem. Database search is not a near-term quantum application.

“Quantum AI will personalize every treatment plan.” There is no published quantum system generating a clinical recommendation validated against a classical baseline. Personalized medicine is limited today by data quality, interoperability, and evidence generation. None of those constraints is computational.

“Quantum key distribution will secure your network.” QKD uses the physics of photons to distribute encryption keys so that eavesdropping is detectable. It requires dedicated fibre or line-of-sight optics, it doesn’t solve authentication, and it protects only the key exchange. The United States National Security Agency has published guidance declining to recommend it for national security systems, citing exactly these limits. For healthcare, post-quantum cryptography in software is the answer, and QKD is a science project with a large capital cost.

“Quantum will crack encryption next year.” Fear-based timelines sell training and appliances. The migration case doesn’t need them. It rests on record lifetimes and device replacement cycles, both of which you already track.

A Twelve-Month Agenda

For a health system executive, the next year of quantum work is unglamorous and entirely within reach.

1. Name an owner. One accountable executive, usually the CISO, with a standing line to the board’s risk committee. Quantum readiness fails most often through diffusion of responsibility.
2. Build the cryptographic inventory. Start with the systems holding data that must stay confidential past 2035: genomics, behavioural health, oncology, research cohorts, and long-term archives. Extend to clinical devices and third-party connections.
3. Change procurement language now. Every contract signed this year for a device or platform with a ten-year life should require support for NIST post-quantum standards and the ability to update cryptography in the field. This costs nothing today and saves a replacement cycle later.
4. Ask your major vendors for dates. Electronic health record, imaging, laboratory information systems, medical device manufacturers, cloud providers. A written roadmap with dates, not a statement of commitment.
5. Design for crypto-agility. Crypto-agility means building systems so that the algorithm can be swapped without redesigning the application around it. Post-quantum is not the last migration anyone will run, and agility is the durable asset.
6. Fund one pilot, deliberately small. Cloud access to quantum hardware is inexpensive. Pick a bounded problem with a classical baseline you already measure, and treat the result as workforce development.
7. Brief the board once, with the arithmetic. Record retention plus migration duration against hardware timelines. No countdown clocks.

Who Needs to Understand This

The constraint here is rarely money and almost never hardware. It is the number of people in a health system who can tell a research bet from a compliance obligation, read a vendor claim about post-quantum support and know what question to ask next, and explain to a board why a scanner purchase in 2027 is a cryptography decision.

That capability is not physics. It sits with the security architects who will own the migration, the procurement and clinical engineering teams who write the specifications, and the executives who approve the sequence. Each needs a different depth, and none needs to write quantum code.

Quantum Academy builds that layer. Our programs separate the executive briefing from the practitioner track, teach the NIST standards as they will actually appear in contracts and configurations, and work through inventory and migration sequencing on realistic estates rather than clean diagrams. Healthcare participants tend to arrive with the device problem and leave with a plan for it.

You can review the full program portfolio and enrollment options at quantumacademy.com/. For migration methodology in more depth, the open framework at pqcframework.org is a useful companion, and PostQuantum.com carries the deeper technical analysis behind several of the points above.

The first clock is running. The other two can wait for you. That one cannot.