Quantum is four businesses. Mapped.
The physics in four experiments you can run yourself, the gap between the machines we have and the ones we need, and where we think the first venture outcomes come from. The one everyone talks about, the quantum computer, is last in line.
"Quantum" is not one market. It is four, on four different clocks.
They share the physics and almost nothing else: different buyers, different capital needs, different timelines to revenue. Two of them sell products today.
Computing
Processors, cloud access, control systems and software for machines that are not yet useful at scale.
- Buyers today
- Labs, governments, enterprises funding experiments
- Capital
- Very high
Sensing
Clocks, magnetometers, gravity and inertial instruments that use atoms as the measuring element.
- Buyers today
- Defense, aerospace, geophysics
- Capital
- Hardware, moderate
Communications
Quantum key distribution and networks that move fragile quantum states between sites.
- Buyers today
- Governments, telecoms, research networks
- Capital
- Infrastructure-heavy
Security (PQC)
New encryption that runs on today's computers and survives a future quantum attack.
- Buyers today
- CISOs, regulated industries, government
- Capital
- Software-like
Computing gets the headlines and most of the capital. Security and sensing get paying customers first. In 2025 the industry raised more than twice what it sold, so the question for any quantum company is which buyer has a budget today.
One idea explains the opportunity and the difficulty: interference.
Not "a qubit is 0 and 1 at the same time." Four experiments, about two minutes, and you will understand what founders mean by coherence, fidelity and error correction.
Fire electrons at two slits
Each electron lands as a single dot. Fire enough and stripes appear, with bands where almost nothing lands. That only happens with waves. Now switch on the detector, which records which slit each electron used. The stripes vanish. Measuring the path destroys the interference, and anything that bumps a qubit, such as heat or vibration, does the same thing by accident.
Simulated with a simplified two-slit model.
Add the arrows, then square
Two routes lead to the same spot, and each alone gives a 25% chance. Ordinary odds would add to 50%. Quantum mechanics adds amplitudes, arrows with a direction called phase, then squares the result. Turn the phase and the same two routes give anything from 100% to 0%. A quantum algorithm is a way of setting the phases so wrong answers cancel.
Add a qubit, double the possibilities
A machine with n qubits works with 2n amplitudes at once. That is where the potential power comes from, and why ordinary computers cannot simulate one past a few dozen qubits.
Why 99.9% isn't good enough
Errors compound. Pick how accurate each operation is and how many the calculation needs. Useful algorithms need millions to billions, which is why error correction, combining many physical qubits into one reliable logical qubit, matters more than qubit count.
Simplified: assumes independent errors and no correction.
Interference is the product, and the environment is constantly trying to destroy it. When a founder quotes a qubit count, ask for fidelity, coherence time and logical error rate. Qubit count alone tells you very little.
The machines are getting bigger. The requirement is falling faster.
In 2012 the best estimate for breaking RSA-2048, the encryption behind much of the internet, was about a billion physical qubits. By 2025 it was under a million. The largest machines grew from a handful of qubits to about a thousand over the same period.
Estimated qubits to break RSA-2048 vs largest machines built
Needed: Fowler et al. (2012), Gidney & Ekerå (2019), Gidney (2025), a 2026 preprint claiming under 100,000. Built: IBM (2016, 2021, 2022, 2023), Google Sycamore (2019), Caltech 6,100-atom array (2025, trapped atoms, not yet a computer). Counts ignore quality, which matters as much as size.
Built today vs what useful jobs need
Built: Quantinuum Helios, Google Willow, IBM Nighthawk, IBM Condor, Caltech atom array. Needed: Google Quantum AI (Mar 2026) for 256-bit elliptic-curve keys, Gidney (May 2025) for RSA-2048, Google / PRX Quantum (2021) for the FeMoco catalyst. Estimates assume about 0.1% physical error rates.
The estimate has dropped by roughly an order of magnitude every few years, from better algorithms and error-correcting codes, not just bigger chips. Plan for the curve, not today's number. That is the urgency behind everything in the next section.
The threat doesn't need a quantum computer to exist yet.
An attacker can record encrypted traffic today and decrypt it later. Michele Mosca's rule: if the years your data must stay secret plus the years it takes to migrate exceed the years until a code-breaking machine arrives, you are already late. NIST finalized the replacement standards in 2024.
Is your data already at risk?
You can't replace cryptography you can't find
The algorithms are free and standardized. The hard part is that most organizations have no inventory of where cryptography lives.
- Websites (HTTPS)
- VPNs
- Certificates and keys
- Internal services and APIs
- Embedded and IoT devices
- Mobile apps
- Software libraries
- Backups and archives
- Vendor systems
- Hardware security modules
- Crypto wallets and signatures
Crypto is a special case. Google's March 2026 estimate covers the signatures that secure Bitcoin and Ethereum. Bitcoin merged BIP-360, its first quantum-resistant address type, this year. The harder problem is governance: who decides what happens to coins whose keys are already exposed.
Selling the algorithm is not a business. Owning discovery and migration is: the system of record for an organization's cryptography. The risk is that the big security platforms bundle it.
The bug is the feature.
The sensitivity that ruins quantum computers is exactly what makes quantum sensors precise. A computer tries to block out the world. A sensor tries to feel it.
Heat, vibration and stray fields are noise to be shielded against. Goal: block the outside world.
A tiny magnetic field or a change in gravity is the signal to be measured. Goal: feel the outside world.
Atomic clocks have been products for decades. The new wave is navigation that works when GPS is jammed, plus magnetometers and gravity sensors. Defense and aerospace account for an estimated 60–70% of quantum sensor revenue today.
The first quantum products with real buyers are instruments. The test is field data at the right size, weight, power and cost, against the best classical sensor rather than an old one. Lab precision that disappears in a moving vehicle is not a product.
Nobody knows which qubit wins. Sell to all of them.
There are five serious ways to build a qubit, and none has won. Every one of them needs control electronics, cryogenics or lasers, photonics, calibration software and fast classical processors to decode errors.
Five ways to build a qubit
| Approach | What the qubit is | Strength | Challenge | Examples |
|---|---|---|---|---|
| Superconducting | A tiny circuit cooled near absolute zero | Fast operations | Wiring and cooling at scale | IBM, Google, Rigetti |
| Trapped ions | A charged atom held by electric fields | Very accurate | Slower operations | Quantinuum, IonQ |
| Neutral atoms | Atoms held in place by focused lasers | Scales to thousands | Moving atoms reliably | QuEra, Atom Computing, Pasqal |
| Photonic | A single particle of light | Telecom-style parts, room temperature | Light gets lost | PsiQuantum, Xanadu, Quandela |
| Silicon spin | An electron's spin in a silicon chip | Could use existing chip fabs | Making devices identical | Intel, Diraq, Quantum Motion |
The picks and shovels
A company that sells to every hardware approach does not have to pick a winner. The risks are a small customer base and the chance that the big hardware makers bring it in-house. The part closest to our AI work is AI for quantum, which can be tested today by how much calibration time it saves.
What would change our view.
- Logical qubits doing real work. 100+ error-corrected qubits running a useful algorithm, independently verified, would pull computing forward.
- A peer-reviewed RSA estimate under 100,000 qubits. That would compress every post-quantum migration deadline.
- A post-quantum vendor with recurring revenue independent of the big platforms. The test of whether migration is a company or a feature.
- Quantum navigation fielded on a drone. At drone-sized weight, power and cost, against the best classical inertial sensor.
- The National Quantum Initiative reauthorization signed into law. Both versions passed committee in 2026; the Senate bill directs a federal post-quantum migration strategy.
Methodology, caveats and sources
Simulations
The two-slit simulator samples from a simplified model: two overlapping single-slit patterns with interference when unwatched, and two separate piles when watched. The error calculator uses success ≈ (1 − error rate)operations, assuming independent errors and no correction. The Mosca calculator uses example inputs; nobody knows when a code-breaking machine arrives.
Estimates
Qubit requirements depend on assumed physical error rates (about 0.1% here), code choice and runtime. Estimates keep falling as algorithms and codes improve. Counts of machines built ignore quality.
Companies
Examples, not a census and not endorsements. Inclusion reflects public information as of October 2026. Spot one missing? Tell us.
Sources
- QED-C, State of the Global Quantum Industry 2026
- Google Research, quantum resource estimate for elliptic-curve cryptography (Mar 2026)
- Gidney, How to factor 2048-bit RSA integers with less than a million noisy qubits (2025)
- Gidney & Ekerå, How to factor 2048-bit RSA integers in 8 hours using 20 million noisy qubits (2019)
- Fowler et al., Surface codes: towards practical large-scale quantum computation (2012)
- Lee et al., Even more efficient quantum computations of chemistry, PRX Quantum (2021)
- Mosca, Cybersecurity in an era with quantum computers: will we be ready? (2015)
- NIST, Post-Quantum Cryptography
- PatSnap, Quantum sensing technology landscape 2026
- Caltech, 6,100-qubit atom array (Sep 2025)
Built by Anthony Avedissian at Canonical · v0.1 · Educational tool. Not investment advice. Simulations are simplified models; data is point-in-time and not exhaustive.