up:: Start Here MOC
So what are quantum computers even for?
They’re being built mainly to model the behavior of atoms and molecules, which is why the US Department of Energy names chemistry, materials science and high-energy physics as the fields the first reliable machines are expected to change. That list is short deliberately. Chemistry is the physics these machines are made of, so a quantum machine is the natural tool for modeling it.
There’s a second half of this story that almost never gets covered. Quantum sensing is already commercial, it has been for years, and one example is above your head right now: the satellites your phone reads to put a dot on a map carry atomic clocks, and an atomic clock is a quantum device.
And the plain state of the computers, as of 2026: no quantum computer has yet beaten an ordinary computer at a problem somebody actually needed solved. That sentence is accurate, it comes from reading the government’s own descriptions rather than from a skeptic, and most marketing material is written to avoid it.
Source: Testimony of Dr Tanner Crowder, Office of Science, US Department of Energy, before the House Committee on Science, Space, and Technology, 22 January 2026, energy.gov.
The short version:
- Quantum technology isn’t one thing on one timeline. Sensing works and sells today. Computing is early, expensive and unproven on real problems. Treating them as one field is why people conclude the whole subject is either finished or fake.
- The DOE names 3 target areas for the first reliable machines: materials science, quantum chemistry, and high-energy physics. All 3 are problems about modeling how matter behaves.
- Quantum sensing is in daily civilian use. In the DOE’s own words, atomic clocks and atom interferometers “are being commercialized and providing technological solutions today.”
- The best 2026 result is a benchmark, not a job. IBM and the University of Chicago demonstrated a verified advantage in July 2026 on a task chosen because it’s hard for ordinary computers, rather than because anyone wanted the answer.
- Most of what’s marketed isn’t on the DOE’s list. Finance, logistics, “quantum AI” and drug discovery have no demonstrated advantage on a real problem.
What can quantum technology already do?
Sensing, and it’s the strongest fact in this whole subject because it’s the one nobody expects.
Quantum sensing works because quantum systems are extraordinarily sensitive to their surroundings. In a computer that sensitivity is the problem to be fought, which is why those machines live inside refrigerators colder than deep space. In a sensor, the sensitivity is the product.
| Device | What it measures | Where it already goes |
|---|---|---|
| Atomic clock | time | satellite navigation, telecom timing, financial timestamps |
| Atom interferometer | gravity and motion | mapping what’s underground, navigation without satellites |
| Diamond magnetometer | magnetic fields | measuring biological activity, working inside high-radiation environments |
Two of those are worth stating slowly, because they sound like sales copy and aren’t. A diamond-based sensor can reach a resolution “capable of detecting single neurons firing,” which is the same measurement problem as scanning a brain or a heart from outside the body. And atom interferometers measure gravity precisely enough for “terrestrial subsurface sensing,” meaning you can see what’s buried under a road without digging it up.
None of that is 2035. All of it is quoted from testimony a federal official gave to Congress in January 2026.
Source: Crowder testimony, 22 January 2026, energy.gov.
Why does sensing work when computing doesn’t yet?
A sensor needs one quantum system to behave for a moment. A computer needs thousands of them to behave together, repeatedly, for a whole calculation, while the environment constantly tries to disturb them.
That difference is the entire gap. The physics is the same. The engineering burden isn’t close.
It’s also why 2 opposite headlines can both be published truthfully in the same week. “Quantum is decades away” is wrong about sensing and defensible about computing. “Quantum is here” is right about sensing and unearned about computing.
What are the computers actually being built for?
The DOE’s quantum lead names 3 areas, and says the first reliable machines “will serve as novel scientific instruments.”
Chemistry. Modeling how molecules really behave. The example almost nobody thinks about is fertilizer. Making ammonia for it takes enormous heat and pressure, and the International Energy Agency puts ammonia production at roughly 2% of the world’s total energy use, with about 70% of it going to fertilizer. Close to half the people alive are fed on crops grown with it. Bacteria in soil run the same reaction at ordinary temperature and we still can’t fully model how they do it.
Materials. Batteries, superconductors and catalysts are all questions about how electrons arrange themselves. Today we answer them by building things and testing them, which is slow. Every argument about electric cars, phone battery life and the power grid sits downstream of this.
High-energy physics. Pure science, and a legitimate reason to build an instrument on its own terms.
Source: Crowder testimony, 22 January 2026, energy.gov; International Energy Agency, “Ammonia Technology Roadmap,” 2021, iea.org; Our World in Data, “How many people does synthetic fertilizer feed?”, ourworldindata.org.
Why chemistry and not banking or delivery routes?
Because the advantage comes from a match between the machine and the problem, and only some problems have it.
Richard Feynman made the argument in 1981 and it hasn’t needed revising. An ordinary computer modeling a quantum system pays a cost that explodes as the system gets bigger, because it has to track every possibility separately. His answer was to stop approximating and build a machine that obeys the same rules as the thing being modeled.
Chemistry is quantum mechanics, so it gets that match for free. Scheduling trucks and balancing a portfolio don’t. Those are ordinary problems with decades of very well funded ordinary solutions behind them, and the theoretical quantum margin is thin.
Source: R. P. Feynman, “Simulating Physics with Computers,” International Journal of Theoretical Physics 21, 467–488, 1982, doi.org.
Has a quantum computer beaten a normal one yet?
At a benchmark, yes. At a problem somebody needed solved, no. Keeping those apart is the single most useful thing a reader of this subject can do.
On 30 July 2026 IBM and the University of Chicago announced a verified quantum advantage: a computation finished in about 15 minutes that leading ordinary methods can’t practically reproduce. It’s a genuine scientific milestone, and the verification part is new and important. It’s also a sampling task, chosen because it’s hard for ordinary computers rather than because anyone wanted the output.
The caution is earned by history. In 2019 Google claimed a task took its machine about 200 seconds against an estimated 10,000 years for a supercomputer. IBM answered within days that an ordinary machine could do it in about 2.5 days, and later work narrowed it further. The headline number didn’t survive people trying.
And ordinary computing isn’t waiting. A 2026 result from the University of Osaka and Fixstars ran a very large chemistry simulation on 1,024 graphics processors and beat the quantum version on every measure tested.
Source: IBM Newsroom, 30 July 2026, newsroom.ibm.com; F. Arute et al., “Quantum supremacy using a programmable superconducting processor,” Nature 574, 505–510, 2019, doi.org; IBM Research, “On quantum supremacy,” research.ibm.com; University of Osaka and Fixstars, 2026, eurekalert.org.
What’s being oversold?
The marketed list and the government’s list barely overlap, and the gap is where most commercial quantum material lives.
| Sold as nearly here | Where it stands |
|---|---|
| Finance and portfolio optimization | Heavily marketed, weakest technically. Ordinary solvers are mature and improving |
| Logistics and route planning | Same shape. Demonstrations are small |
| Quantum machine learning | Real research, no demonstrated advantage on real data |
| Drug discovery | Sits downstream of chemistry, so it inherits chemistry’s timeline |
| ”Quantum AI” | Mostly a naming exercise. Ask which of the 2 words is doing the work |
That doesn’t make these companies dishonest and several do serious research. It makes the application claims premature, which matters because those claims are what budgets and headlines rest on.
How do I check a quantum claim without being a physicist?
Two questions do most of the work.
- Is that a logical qubit or a physical one? Physical qubits are raw, noisy hardware. Logical qubits are error-corrected assemblies of many physical ones, and they’re the unit that decides what a machine can really run. A headline quoting a big physical number next to a claim about encryption is comparing 2 different things.
- Was the problem picked because it’s useful, or because it’s hard for ordinary computers? The first shows value. The second shows capability. Both are worth reporting and mixing them up is how a benchmark becomes “quantum computers now beat supercomputers.”
The longer version is at How Do You Tell Real Quantum Progress From Hype.
Isn’t the point of quantum computers to break encryption?
Breaking encryption is unusual in this list rather than central to it. It’s the one use that was proven on paper before the machine existed. Peter Shor published the method in 1994 and it’s been checked ever since, so it’s the only application already waiting on hardware rather than waiting on somebody to find a job for the machine.
It’s also narrower than most coverage suggests. The kind of encryption that breaks is the kind that lets 2 strangers agree on a secret and proves a website is genuine. The kind that scrambles your files survives with a larger key, and the NSA advisory that retires the vulnerable math for national-security systems keeps AES-256 as its required cipher.
Full detail: What Can a Quantum Computer Actually Break, and on why the deadline isn’t the same as the arrival date, Harvest Now, Decrypt Later (HNDL).
Source: NSA, “Announcing the Commercial National Security Algorithm Suite 2.0,” nsa.gov.
Has anything like this happened before?
The DOE’s own quantum lead reached for the comparison. Speaking at Quantum USA 2026, Crowder placed quantum computing at its “ENIAC moment,” still needing applications and far more performance.
In 1943 the US Army was designing artillery faster than it could work out how to aim it. The aiming tables were calculated by about 200 women whose job title on the paperwork was literally “computer,” because that’s what the word meant. One trajectory took a person 20 to 40 hours on a mechanical calculator, and a single table needed roughly 1,800 of them.
So the Army paid for a machine to do that one narrow job, and ENIAC was built at the University of Pennsylvania. When it was finished in 1945 the first serious calculation run on it was a problem for Los Alamos, not a firing table. What the machine turned into after that was not written down by anyone funding it in 1943, because nobody could have written it down.
The lesson people usually take is that every expensive instrument becomes a consumer product, which is the wrong lesson and the one vendors prefer. The defensible one is narrower: the uses arrived after the machine. That cuts against confident enthusiasm and confident dismissal equally, and it means anyone naming what quantum will be worth in 2050 is guessing, whether they’re selling it or dismissing it.
Questions people ask
Is quantum technology really in my phone? Not in the handset. The satellites it reads for position carry atomic clocks, so the quantum device is in the system rather than in your hand.
When will a quantum computer do something useful? Nobody credible names a year. The DOE’s framing is that the field still needs both applications and far more performance.
Will quantum computers replace normal ones? No. They’re special-purpose instruments expected to work alongside ordinary computers, which is also how the DOE’s own Genesis Mission platform describes the setup.
If chemistry is the target, why does every ad mention finance? Finance and logistics have buyers with budgets and short purchasing cycles. Chemistry research doesn’t buy machines the same way, so the marketing follows the money rather than the physics.
Does the 2026 IBM result change the encryption timeline? Not directly. It ran on 70 error-corrected qubits, and breaking today’s public-key encryption needs a machine orders of magnitude larger.
Does a quantum computer try every answer at once? No, and this is the most common wrong picture. Getting a useful answer out requires cancelling the wrong ones and reinforcing the right ones, which is why useful quantum methods are rare and hard to invent rather than automatic.
Does more qubits mean a better machine? Not on its own. A qubit count with no error rate attached describes very little.
If you’re writing about this and want a figure checked before it runs, the corrections page has the contact route, and everything here is free to quote with attribution.
Last verified 2026-09-04 · Maintained by Addie LaMarr, LaMarr Labs.