Free preview. This lesson is part of a paid course in the ETA Quantum Fluency. About this course
Five ways to build a qubit, and the specs that matter
In this lesson you'll
- Compare the five main ways to build qubits and their trade-offs
- Read a hardware announcement for the specs that matter
- Explain why no single approach has won yet
A refresher
Quantum Fluency: Foundations covered qubits, superposition, entanglement, interference, noise and the difference between physical and logical qubits. This course builds on that. If any of those terms feel shaky, review its first module before you continue.
Five ways to build a qubit
| Approach | What the qubit is | Strengths | Challenges | Examples |
|---|---|---|---|---|
| Superconducting circuits | Tiny electrical circuits cooled close to absolute zero | Very fast operations; built with chip-making methods | Needs large dilution refrigerators; qubits connect mainly to neighbors; wiring at scale | IBM, Google, Rigetti |
| Trapped ions | Charged atoms held by electric fields and controlled with lasers | Very high accuracy; any qubit can interact with any other | Slower operations; scaling to many ions and moving them around | Quantinuum, IonQ |
| Neutral atoms | Uncharged atoms held in arrays of laser "tweezers" | Large arrays; atoms can be rearranged | Atom loss; slower cycles; still maturing | QuEra, Atom Computing, Infleqtion, Pasqal |
| Photonics | Particles of light traveling through optical circuits | Works at or near room temperature for parts of the system; fits fiber and chip manufacturing | Photon loss; needs very large systems for error correction | PsiQuantum, Xanadu |
| Spin qubits in silicon | The spin of single electrons in silicon devices | Tiny; could reuse semiconductor factories | Uniformity and control at scale | Diraq, Intel |
A sixth approach, topological qubits, aims to build error resistance into the hardware itself. Microsoft announced a topological qubit chip in February 2025; researchers continue to debate how strongly the published evidence supports its claims. It's a good case for the claim check: an important idea, with results still being validated.
The roots of superconducting qubits
The 2025 Nobel Prize in Physics went to John Clarke, Michel Devoret and John Martinis for experiments in 1984 and 1985 showing quantum tunneling and energy levels in an electrical circuit large enough to hold in your hand. That work led to superconducting qubits, today's most widely used approach. Devoret and Martinis both went on to lead quantum hardware work in industry.
Specs that matter more than qubit count
When you read a hardware announcement, look for:
- Two-qubit gate fidelity: how often the most error-prone operations succeed. For example, Quantinuum reported 99.921% two-qubit fidelity across all qubit pairs for its 98-qubit Helios system launched in November 2025. Each extra "9" matters enormously for error correction.
- Connectivity: which qubits can interact directly. All-to-all connectivity (common in trapped ions) avoids extra operations to move information around.
- Speed: how long operations and error-correction cycles take. Superconducting systems are much faster per operation; slower systems may compensate with fewer errors.
- Logical qubits and their error rates: how many error-protected qubits the system supports, under which code, and how good they are (lesson 2).
- How you access it: cloud availability, software tools and queue times.
A system with fewer, better qubits can beat one with more, noisier qubits.
Why no approach has won
Each approach is betting on a different hard problem: refrigeration and wiring, laser control at scale, atom loss, photon loss or manufacturing uniformity. DARPA's Quantum Benchmarking Initiative, which is assessing whether any approach can reach a useful, utility-scale quantum computer by 2033, advanced 11 companies across several of these approaches to its second stage in November 2025. That diversity is a fair picture of the field: serious contenders, no clear winner.
Workplace example: a pharmaceutical R&D group
A drug discovery team gets pitches from three quantum hardware companies, each claiming to be "the most powerful." Instead of ranking by qubit count, the team builds a one-page comparison: two-qubit fidelity, connectivity, logical qubit demonstrations, cloud access and software support. Two of the three can't provide independent benchmarks. The team runs small learning experiments on the two platforms it can access through its existing cloud provider.
Try it
Pick two quantum hardware companies from the table. Ask an approved AI tool:
Compare [company A] and [company B] quantum computers as of late 2026 on: qubit
technology, largest system, reported two-qubit gate fidelity, connectivity, logical
qubit demonstrations, and how customers access them. Cite a source for each number
and mark anything you're not sure of as uncertain.Open at least two cited sources and check the numbers. Note any figure the AI got wrong or couldn't source.
Key takeaways
- Five main approaches (superconducting, trapped ion, neutral atom, photonic, silicon spin) trade speed, accuracy, connectivity and scalability differently; topological qubits remain under study.
- Read fidelity, connectivity, speed, logical qubit results and access, not just qubit counts.
- The 2025 Nobel Prize in Physics recognized the experiments behind superconducting qubits.
- DARPA's QBI advanced 11 companies across several approaches in November 2025; no approach has won.
Sources