Superconducting, Ions, Atoms, Photons or Spins: Which Qubit Will Lead?

[ProperJobs50]

One of the most exciting things about quantum computing right now is that nobody knows which technology will win. In classical computing, the silicon transistor took over decades ago and never looked back. Quantum computing is still in the era of competing designs, with several serious approaches all making real progress. DARPA's decision to move four very different companies into the final stage of its Quantum Benchmarking Initiative, alongside two that were already there, shows that even the experts are keeping their options open. So let us look at the main contenders and what each brings to the table

Superconducting qubits are the best known. They are tiny electrical circuits cooled to a fraction of a degree above absolute zero, and they are used by IBM, Google, Rigetti, IQM and many others. Their big strength is speed, with gate operations measured in tens of nanoseconds, and they can be made with techniques borrowed from chip manufacturing. Google's Willow chip showed in late 2024 that adding more qubits to an error correcting code could reduce errors rather than increase them, a key milestone. The challenges are wiring, cooling and the fact that each qubit is slightly different and needs careful calibration

Trapped ions take a different route. Individual charged atoms are held in place by electromagnetic fields and controlled with lasers or microwaves. Quantinuum and IonQ are the leading names, and ions have long held the records for the most accurate operations. Every ion of the same element is identical, which removes much of the variation that troubles superconducting chips. The downside has traditionally been speed and scale, as operations are slower and moving many ions around is hard. Companies like Universal Quantum and ZuriQ are working on chip based traps that could change that

Neutral atoms have been the surprise of the last few years. Uncharged atoms are held in place by focused laser beams called optical tweezers and can be rearranged during a calculation. Atom Computing, QuEra and Pasqal all use this approach, and Atom Computing has already passed 1,200 qubits. In 2023, a Harvard and QuEra team ran algorithms on 48 logical qubits, an impressive result at the time. The flexibility to move atoms around makes some error correcting codes much more efficient, which is why startups like Oratomic think a fault tolerant machine could need far fewer qubits than expected

Photonic quantum computing uses particles of light. PsiQuantum, Xanadu and Quandela are among the leaders, and the big appeal is that photons travel easily through standard fibre and can be made with existing semiconductor foundries. PsiQuantum is building large manufacturing sites in Brisbane and Chicago on the bet that its approach scales well. The challenge is that photons do not easily interact with each other, and losing photons along the way causes errors. Recent work on deterministic fusion of photonic states is chipping away at those problems

Spin qubits in silicon are the quiet contender. They store information in the spin of single electrons, often in structures very similar to ordinary transistors. Diraq, Quobly and Intel are working in this area, and Diraq's place in DARPA's final stage shows the idea is being taken seriously. The huge potential advantage is density, since millions of tiny qubits could in principle fit on one chip made in an existing fab. Fidelities have improved a lot in recent years, though connecting and controlling large numbers of spins is still a work in progress

There are other approaches worth mentioning too. Diamond based qubits using nitrogen vacancy centres can work at room temperature, and recent experiments have run small algorithms without any cryogenics at all. Microsoft is pursuing topological qubits, which could be naturally protected from errors if they work as hoped, although its claims have been debated. Some groups are exploring molecules and other exotic systems. The variety is remarkable

What I find most encouraging is that the competition is pushing every approach forward. Superconducting groups are borrowing ideas from ion traps, neutral atom teams are learning from photonics and interconnect research is starting to link different qubit types together. IonQ and Quantum Source have both shown interconnect work that could, in principle, connect different kinds of processors. The future might not be a single winner at all, but a mix of technologies doing different jobs. Fast superconducting processors could handle some tasks while accurate ions or flexible atoms handle others

If I had to bet, I would say superconducting and neutral atom systems look strongest for the first large fault tolerant machines, with trapped ions close behind on quality. Photonics and silicon spins are longer bets with enormous upside if they crack their scaling problems. But I hold that view loosely. Five years ago, few people would have put neutral atoms near the top of the list, and they surprised almost everyone

It is also worth thinking about what winning even means. The first technology to run a useful fault tolerant calculation might not be the one that dominates commercially twenty years later. Early leaders in classical computing, such as vacuum tubes and magnetic core memory, gave way to transistors and silicon chips that scaled better. The same could happen here. A platform that is slower to start but easier to manufacture could overtake an early leader once the industry moves to mass production

Cost and practicality will matter as much as raw performance. A machine that needs a building sized cryogenic plant will suit national labs and big cloud providers, while one that runs closer to room temperature could reach many more users. Power consumption, maintenance and the availability of skilled engineers all play a part. Data centres care about these details more than headline qubit counts. The eventual winners will probably be the approaches that balance performance with the realities of running a machine day after day

Which qubit technology do you think will lead, and why? Do you expect one clear winner, or a mix of approaches working together? And is there an outsider you think people are underestimating?

Four more teams enter Quantum Benchmarking Initiative's final stage | DARPAwww.darpa.mil