← All articles

What a quantum computer actually looks like

On this page

1. Five layers in a typical system
2. The golden chandelier: superconducting qubits
3. Trapped ions: atoms held by electric fields
4. Neutral atoms: grids drawn with laser light
5. Photons: light on a chip
6. The ordinary computer is half the machine
7. How people actually use one
8. What the picture can't tell you
9. Common questions
10. Sources and further reading

Most photos of a quantum computer show a gleaming gold structure hanging from the ceiling. That object is real, but it isn't the computer. It's mainly the refrigerator around one small chip. This piece goes through the whole machine piece by piece: what the qubits sit in, what controls them, the ordinary computers that run alongside them, and how people actually get to use one. The aim is that the next time a vendor, a news story or a board paper shows you a picture, you know what you're looking at and which questions it raises.

Five layers in a typical system

There are several competing ways to build qubits, described below. Underneath the differences, a typical quantum computing system can be described in five layers, from the person using it down to the qubits. Real systems vary in the details, but the layers are a useful map:

The qubits need extreme isolation because, as Why qubits are so fragile explains, almost anything from the outside world (heat, vibration, a stray signal) disturbs them. So each design wraps its qubits in a protected environment, reaches in with carefully timed pulses to run gates, and reads the answers out at the end. What that looks like depends on what the qubits are made of.

The golden chandelier: superconducting qubits

This is the design behind the famous photos, used by IBM and Google among others. The qubits are tiny electrical circuits on a chip about the size of a coin. Keeping them far colder than outer space stops heat from disturbing them and lets them work reliably as qubits, so the chip is mounted at the bottom of a dilution refrigerator.

The "chandelier" is that refrigerator with its outer cans removed. It's a stack of metal plates, each colder than the one above. The top plate is at room temperature; the bottom one, where the chip is mounted, reaches about 10 to 15 thousandths of a degree above absolute zero. Much of the metalwork is copper, which carries heat well, and cooling the whole stack down from room temperature can take about two days.

Bundles of cables run down through the plates. They carry microwave pulses, at a few billion cycles per second, from racks of electronics outside the refrigerator to the chip, and carry faint readout signals back up through amplifiers. Those racks, plus the pumps and gas-handling equipment that run the refrigerator, take up far more space than the chip itself. A whole system fills a room rather than a desk.

Trapped ions: atoms held by electric fields

Companies such as Quantinuum and IonQ use individual charged atoms, called ions, as qubits. Electric fields hold the ions in place. In many current designs the fields come from electrodes on a small chip, with the ions floating a few hundredths of a millimetre above its surface; other designs use different trap shapes. Every design needs an ultra-high vacuum, so that stray air molecules don't knock into the ions. Some run with the vacuum chamber at room temperature; others cool it, to around ten degrees above absolute zero, to cut electrical noise.

Instead of microwave cables, the main tools are lasers. They cool the ions, run gates on them and read them out. At the end of a calculation, a laser makes each ion glow or stay dark depending on its state, and a camera or light detector records the pattern. Some designs move ions around the chip between zones, bringing pairs together when they need to interact. The surrounding optics, lasers and control electronics fill several equipment racks.

Neutral atoms: grids drawn with laser light

Companies such as QuEra, Pasqal and Atom Computing use uncharged atoms instead. A cloud of atoms is cooled by lasers to less than a thousandth of a degree above absolute zero, inside a small glass vacuum cell. Then tightly focused laser beams, called optical tweezers, each catch a single atom and arrange them into a grid a few thousandths of a millimetre apart. In 2025 a Caltech team held more than 6,100 atoms this way. That was a demonstration of holding and controlling many atoms, not a working 6,100-qubit computer.

The atoms are cold, but the machine around them isn't: these systems don't need a dilution refrigerator, and researchers describe fitting them into standard equipment racks. Gates are run with lasers that briefly push pairs of atoms into a highly excited state so they interact. To read the result, the atoms are lit up and photographed; each shows up as a bright or dark spot. In many systems, measuring an atom knocks it out of the grid, so the grid has to be refilled for the next run. Newer techniques can read an atom's state while keeping it in place, so it can be reused.

Photons: light on a chip

A fourth approach, pursued by companies such as PsiQuantum and Xanadu, uses particles of light. Photons travel through channels etched into chips, a bit like light in optical fibre, and are steered and combined there. Unlike superconducting qubits, photons don't need a deep freeze to keep their quantum state, and they can travel between chips over fibre. That doesn't always mean no refrigeration: the most sensitive light detectors can need it. In 2025 Xanadu described a system linking 35 photonic chips across four server racks; everything ran at room temperature except its 36 light detectors, which sat in two dilution refrigerators. The hard part is that photons are easily lost along the way, and Xanadu named that loss as the biggest obstacle still to overcome.

The ordinary computer is half the machine

A quantum computer never works on its own. Ordinary computers sit beside it at every stage:

• Before the run, software translates a program into the gates the hardware actually supports, then into precisely timed pulses for each qubit.
• During the run, fast electronics fire those pulses with timing accurate to billionths of a second. Some machines also measure selected qubits partway through and feed the results straight back into the next steps. In the experimental machines that correct their own errors, ordinary processors read these error checks and decide on fixes in real time, while the calculation is still going. That capability has been demonstrated, but it isn't yet standard.
• After the run, the results come back as samples: the same circuit is run many times, and the answer is the pattern across those runs.

So it's more accurate to think of a quantum processor as an accelerator that works alongside ordinary computing, the way a graphics processor speeds up certain tasks for an ordinary server, than as a replacement for it.

How people actually use one

Almost nobody who uses a quantum computer has one in the building. Most access is over the cloud. A user writes a circuit on an ordinary laptop, often in a toolkit such as Qiskit, sends it to a provider, waits in a queue, and gets the results back. Hardware makers run their own cloud services, and large cloud platforms such as Amazon Braket and Microsoft Azure Quantum offer machines from several makers side by side.

For a bank, that has two practical consequences. First, experimenting with quantum computing is a cloud and supplier question before it's a hardware question, so the usual third-party, data-handling and data-location reviews apply to anything sent to a quantum service. Second, a quantum computer is not a server you can add to a data centre to speed up everyday systems. It's a specialised instrument, used for particular calculations, that stays with the provider.

What the picture can't tell you

A photo shows size and engineering effort. It doesn't show how good the machine is. Two systems that look alike can differ enormously in how often their gates make mistakes, how many qubits can interact directly, and whether they can correct their own errors. Those are the numbers that matter, and they're the subject of a separate piece on comparing quantum hardware. The picture also says nothing about cryptography. No publicly demonstrated quantum computer, of any design, can break the public-key cryptography banks rely on, such as 2048-bit RSA or common elliptic-curve keys. Symmetric encryption such as AES faces a different and much smaller quantum threat. Why the public-key side still needs migrating now is covered in Why your bank is already preparing.

Common questions

• Why does it need to be so cold? Heat is random jiggling, and random jiggling scrambles qubits. Superconducting chips need the deep freeze to work at all. Ions and atoms are cooled by lasers instead, so the equipment around them can stay much warmer.
• How big is it? The qubits are tiny: a coin-sized chip, or a grid of atoms far smaller than a fingernail. The supporting equipment fills racks, and in the largest superconducting systems, a room.
• Will there be one on my desk? Not for the foreseeable future, and for most uses there doesn't need to be. Access over the cloud is how the field works today.
• Are these the only designs? No. Others are in development. One example is silicon spin qubits, which store information in single electrons on chips made much like ordinary computer chips; Intel released a 12-qubit research chip of this kind in 2023.
• Which design is winning? No design has won. Each trades speed, error rates and ease of scaling differently, which is why it pays to compare them on their published numbers rather than their photos.

Sources and further reading

Tags show what kind of source each one is. A standard or government guidance is an official document; a peer-reviewed paper has been checked by other experts; a preprint has not been peer-reviewed yet; an experiment reports a real-world demonstration; a company announcement is the company's own account; vendor documentation is a supplier's own technical guide; a news report is a journalist's account. Dates and figures were checked against these sources on 16 October 2026. Spotted an error? Email hello@plainquantum.com and it will be corrected, with a note.

1. News reportIn photos: Journey to the center of a quantum computer (IBM's dilution refrigerator)Charlotte Hu, Popular Science, 2022
2. CourseCryogenics parallel sessions: dilution refrigerator temperature stagesHollister, Tatkowski, James and Dang, Fermilab (US Quantum Information Science Summer School), 2023
3. PreprintIon-based quantum computing hardware: performance and end-user perspectiveStrohm et al., arXiv, 2024
4. Peer-reviewed paperNeutral atom quantum computing hardware: performance and end-user perspectiveWintersperger et al., EPJ Quantum Technology, 2023
5. ExperimentA tweezer array with 6100 highly coherent atomic qubitsManetsch et al. (Caltech), Nature, 2025
6. PreprintFast, continuous and coherent atom replacement in a neutral atom qubit array (includes non-destructive readout and atom reuse)Li, Bao, Peper, Li and Thompson (Princeton), arXiv, 2025
7. ExperimentScaling and networking a modular photonic quantum computer (Xanadu Aurora)Aghaee Rad et al. (Xanadu), Nature, 2025
8. ExperimentQuantum error correction below the surface code thresholdGoogle Quantum AI and collaborators, Nature, 2024
9. Vendor documentationClassical feedforward and control flow (dynamic circuits: measuring qubits mid-circuit)IBM Quantum documentation, 2026
10. Company announcementIntel's new chip to advance silicon spin qubit research (Tunnel Falls)Intel newsroom, 2023
11. Company announcementAmazon Braket: access to quantum computers from several makersAmazon Web Services, 2026
12. CourseBasics of quantum information (free course)IBM Quantum Learning