Quantum Computing in Business (2026): What Actually Works, What’s Marketing, and the One Deadline You Can’t Skip

Last updated: October 1, 2026. Every figure below carries a source and an evidence label (explained in the next section).

Short answer. In October 2026, quantum computing delivers vendor-reported gains on a narrow class of scheduling and routing problems, a handful of physics-grade “beyond-classical” demonstrations, and no independently verified business advantage. The one action that is not optional is post-quantum cryptography (PQC) migration. Its clock has moved forward over the past twelve months, and not because hardware improved. The estimates of what an attack would need collapsed.

The argument in one paragraph

Quantum computing in business is two stories that people keep merging into one. The value story (optimization, simulation, machine learning) depends on hardware that is still being proven. The threat story (breaking RSA and elliptic-curve cryptography) depends largely on algorithms and error-correction designs, which improve on paper, without new machines. That is why the threat clock has accelerated while the value clock has not. Budget for them separately. Don’t let uncertainty about the value story delay the threat work.

How to read the evidence in this article

Not all numbers deserve equal trust. Here is how each claim is labeled.

Where sources disagree or I couldn’t confirm something, I say so instead of smoothing it over.

Part 1: The clock that is moving, cryptography

What changed in twelve months

The June 2026 executive order: what it actually requires

The order, titled Securing the Nation Against Advanced Cryptographic Attacks, was signed June 22, 2026. Per Cybersecurity Dive, Federal News Network, and Cloudflare’s analysis:

• Federal agencies must move high-value assets and high-impact systems to post-quantum key establishment by December 31, 2030 and digital signatures by December 31, 2031.
• Agencies name a PQC transition lead within 30 days.
• The FAR Council has 180 days to propose a rule requiring covered contractors to meet NIST’s PQC-inclusive FIPS by the end of 2030.
• Commerce is tasked with pilot-testing PQC on selected NIST systems by the end of 2027.

What the order does not do. It binds federal agencies directly. For a private company, the pressure arrives through contracts, so the FAR rule is the thing to watch, not the order itself. I could not confirm whether OMB’s follow-on guidance has been published as of this writing.

The gap between estimate and hardware

Here is the calculation most coverage skips.

Reading this honestly: no existing machine is close, and even the most aggressive public hardware roadmap lands several-fold short of the attack requirement in 2029. Raw qubit ratios also overstate the gap, since error rates matter more than counts.

But the 2019 → 2025 drop in the RSA estimate was a 20× reduction from algorithms and error-correction design alone, with the same hardware assumptions. A shortfall of 6–7× is the kind of gap that further algorithmic progress has closed before. This is why serious infrastructure operators plan around 2029, and why I think anyone treating “it’s decades away” as a safe planning assumption is making an unforced error. Cloudflare says plainly that 2029 is a readiness deadline, not a prediction. That is the right framing.

Encryption is largely handled. Signatures are the problem.

This is the distinction most “prepare for quantum” articles miss.

Key exchange (protecting data in transit): the migration is further along than most teams realize. Over 50% of human traffic to Cloudflare was post-quantum protected by October 2025, and recent browsers, OpenSSL, Go, and Apple OSes enable X25519MLKEM768 by default. X25519MLKEM768 is a hybrid: a classical X25519 share plus an ML-KEM-768 share, so the session stays safe if either holds. Hybrid is an IETF standards-track practice (draft-ietf-tls-ecdhe-mlkem, May 2026), not a NIST mandate.

Signatures (proving identity): this is where the work remains, and the reason is physical. Post-quantum signatures and keys are far larger than what they replace.

A TLS handshake carries a certificate chain, so these sizes multiply. That explains why the federal signature deadline is a year after the key-establishment deadline, and why Cloudflare is building toward Merkle Tree Certificates for visitor connections by mid-2027 rather than just swapping algorithms.

My position: Google and Cloudflare both moved authentication to the front of their roadmaps after the 2026 elliptic-curve estimates. That is the correct priority. Harvest-now-decrypt-later attacks only matter for data that is still sensitive years from now. A forged signature or certificate is a live-fire attack the day a capable machine exists, and long-lived keys, such as firmware signing, device identity, and root certificates, are the hardest to replace afterward.

What to do: specific enough to change your priorities

1. Rank by key lifetime, not by system importance. Firmware-signing keys, device identity, code-signing roots, and anything with a 10-year-plus validity come first. A short-lived web session ranks last.
2. Check the hops your CDN doesn’t cover. A CDN may terminate post-quantum TLS at the edge while the origin connection, internal mTLS, VPN, and SSH stay classical. Audit those.
3. Test for post-quantum-intolerant endpoints before turning anything on. Hybrid key shares enlarge the handshake, and some stacks fail on it. Open-source probes such as pqprobe classify endpoints that reset connections from post-quantum-capable clients while classical clients still pass your existing health checks. If your monitoring only uses classic clients, it won’t see this failure.
4. Put PQC into vendor contracts now. Cloudflare’s own advice is to make post-quantum support a procurement requirement. Bain’s 2025 survey found only 9% of tech leaders had a roadmap for quantum cybersecurity risk (Secondary), so your vendors are probably behind too.
5. Design for algorithm swaps, not a single migration. Parameterize algorithm choice. Don’t hardcode RSA or ECDSA into anything meant to live past 2030.
6. Don’t rush the cutover. A counterpoint worth taking seriously: Stanford’s Dan Boneh, a co-author of Google’s 2026 white paper, has reportedly warned that a hasty transition is more likely to cause a catastrophic bug than a quantum attack is (Secondary report; I did not locate the original statement). Hybrid modes and staged rollouts exist for this reason.

Part 2: Optimization, what the production evidence supports

Quantum annealing is not the “universal quantum computer” from textbooks. It targets one problem family: combinatorial optimization expressed as QUBO (quadratic unconstrained binary optimization). Scheduling, routing, and resource allocation fit. Almost nothing else does.

The reported deployments

Calculation: Ford Otosan’s 30 → 5 minutes is an 83% reduction (25 ÷ 30), which matches the “83%” figure D-Wave cites elsewhere. The numbers are internally consistent. That is not the same as independently verified.

The one benchmark I could check

I opened the abstract of arXiv 2408.01641, a benchmark of D-Wave’s Leap Hybrid Solver against an industrial classical solver on a job-and-machine scheduling problem with up to roughly 150,000 variables. It reports that D-Wave’s results were “highly competitive and sometimes offer speedups.” That is a fair summary of the paper, and it is a single problem class. Note also what it does not say: it claims no general quantum advantage.

Correction to the earlier version of this article. The previous edition attributed a sentence to this paper saying no practical quantum advantage has been demonstrated for industry-relevant combinatorial optimization. That quote does not appear in the abstract, and I could not verify it elsewhere in the paper, so it has been removed.

The commercial scale check

D-Wave’s own Q2 2026 materials, reported by Investing.com, show first-half bookings of $35.5 million (including $2.3 million from the Quantum Circuits acquisition closed in January 2026) and a revenue miss. Bain’s market estimate is blunter: the hardware-and-services market was under $1 billion a year in 2025, growing to $5–15 billion by 2035 from early simulation and optimization uses. The $100–250 billion figure circulating in headlines is Bain’s estimate for a fully capable, fault-tolerant machine, which does not exist.

My position on annealing

It is real, it is in production at named companies, and none of the published evidence shows it beating a well-tuned classical solver on your problem. The honest framing is that you would be buying a solver, so test it like one.

A bake-off protocol (a method I’m proposing, not a result I ran):

Free trial access has been offered through D-Wave’s LaunchPad program (announced January 2025). I did not re-verify its current terms, so check before relying on it.

Part 3: Gate-model quantum, where “advantage” claims stand

My position: the Quantum Echoes result is the strongest advantage claim to date. Google’s classical red-teaming is unusually thorough, and as of mid-2026 it has not been overturned (Secondary; reported support from Scott Aaronson). But look at what it measures: a physics observable. It is not a business workload. Science News also noted that earlier advantage claims were frequently followed by improved classical methods, which is why IBM built an open tracker in the first place.

What to watch in the next 90 days: IBM’s end-of-2026 verified-advantage target is the nearest checkable prediction in the field. Judge it by two things: whether the tracker entries survive classical rebuttal, and whether the problem is one a business would pay to solve. A physics benchmark that passes verification does not change your roadmap. A verified result on a chemistry, materials, or logistics problem would.

For developers: learn the fundamentals and use cloud access (IBM Quantum, Azure Quantum, Amazon Braket) for experiments. Do not build production dependencies on gate-model hardware.

The question this article must answer: when is Q-Day?

Nobody can date it, and anyone who gives you a year with confidence is selling something. Here is what can be said.

• Planning dates in use: Google and Cloudflare target 2029 for completed migration. Microsoft’s reported target is 2033 (Secondary). Cloudflare explicitly says its date is not a forecast.
• The measurable trigger: the number to watch is logical qubits with sustained low error rates, not headline physical-qubit counts. The ECC estimate needs ~1,200–1,450 logical qubits. Roadmap targets sit near 200 for 2029.
• What would resolve it: a demonstration of error-corrected logical qubits in the high hundreds with gate counts in the tens of millions. Until that appears on a roadmap with a delivery date, the honest answer is a range, with estimates that have fallen roughly 20× in six years.

The asymmetry that should decide your budget: being early on PQC costs engineering time, and hybrid key exchange is already free on most modern stacks. Being late costs a forged-identity problem you cannot patch retroactively. Being early on quantum optimization costs one pilot. Be early on the first and skeptical on the second.

Frequently asked questions

Is quantum computing used in business today?
Yes, narrowly. Annealing and hybrid solvers are used for scheduling and network optimization at companies such as Pattison Food Group, NTT Docomo, and Ford Otosan. The results are vendor-reported, and I found no independent audit.

Can quantum computers break encryption right now?
No. Today’s largest machines have on the order of 100 to roughly 1,000 physical qubits. Published estimates for breaking RSA-2048 or 256-bit elliptic-curve cryptography assume hundreds of thousands of physical qubits and over a thousand logical qubits.

What is “harvest now, decrypt later”?
An adversary records encrypted traffic today and decrypts it once a capable quantum computer exists. It matters for data that will still be sensitive in 10-plus years. Post-quantum key exchange (ML-KEM) addresses it, and much of the web already uses it.

Do small companies need to migrate?
Mostly, you inherit the migration from browsers, CDNs, and cloud providers. Your job is what they don’t cover: origin connections, VPN, SSH, code and firmware signing, and vendor requirements.

Should I buy quantum optimization?
Only after a head-to-head against classical baselines on your own data (see the bake-off protocol above).

What is the difference between ML-KEM and ML-DSA?
ML-KEM (FIPS 203) establishes shared keys. ML-DSA (FIPS 204) creates digital signatures. They solve different problems, which is why the federal deadlines differ.

Glossary

• QUBO: the math format annealers solve; many scheduling problems can be translated into it.
• Logical qubit: an error-corrected qubit built from many physical qubits. The number that matters for cryptanalysis.
• Hybrid key exchange: combining a classical and a post-quantum algorithm so the session is safe if either one holds.
• CRQC: cryptographically relevant quantum computer, one capable of breaking today’s public-key cryptography.
• Crypto agility: designing systems so algorithms can be replaced without rewriting applications.

What I could not verify

• Original Pattison claims. An earlier version cited “25 hours to two minutes” and “50,000 workforce hours saved annually.” I searched D-Wave’s published pages and found only the ~80% effort-reduction figure, so the others were removed.
• NTT Docomo scale. “250,000 base stations” and “80 million subscribers” appeared in the earlier version. I could not confirm them, so they are removed.
• Executive order number. Sources disagree (14412 in some, 14409 in one). I’ve used the title and date instead. Check the Federal Register entry before citing a number.
• Whether any quantum-optimization result beats the best classical solver at industrial scale. The published evidence does not settle it.

Updates and corrections since the April 2026 edition

Rewritten for October 2026. Replaced the outdated “June 2025 executive order” framing with the verified June 22, 2026 order. Removed unverifiable figures (a survey percentage with an undisclosed population, a market range from a single aggregator, a CEO quote, and a quantum-sensing claim I could not trace). Updated hardware status (IBM’s Kookaburra moved to the 2026 roadmap; Rigetti’s 2023-era fidelity figure removed as stale). Replaced a low-authority developer guide with Cloudflare, IETF, and NIST-linked sources.

Key sources

• NIST, Post-Quantum Cryptography project
• Gidney, How to factor 2048 bit RSA integers with less than a million noisy qubits (2025)
• Google Quantum AI, cryptocurrency white paper on elliptic-curve resource estimates (2026)
• Cloudflare, State of the post-quantum Internet in 2025 and analysis of the 2026 executive order
• IETF, draft-ietf-tls-ecdhe-mlkem
• Google Research, A verifiable quantum advantage
• Bain & Company, Quantum Computing Moves from Theoretical to Inevitable (2025)
• arXiv, D-Wave Leap Hybrid Solver scheduling benchmark (2024)
• D-Wave, customer success stories (vendor-reported)