Photons, Majorana and silicon
Lesson 17 of 23 in the course “Quantum”. Free, with animations and a quiz.
A qubit made of light
A photon can be a qubit too: for example, 0 is one path, 1 is another, like in the interferometer from level one. Pluses: a photon barely interacts with its surroundings, so it’s low-noise, and it travels through fibre by itself, so qubits are easy to connect. The minuses are the flip side: photons don’t want to interact with each other either, and a two-qubit gate needs clever tricks. The main enemy here isn’t noise but loss: the photon just disappears.
PsiQuantum: all or nothing
The boldest photonic bet belongs to PsiQuantum. The company doesn’t release small demo machines; it’s going straight for a million-qubit computer. Its chips are made at an ordinary semiconductor fab, GlobalFoundries. In September 2025 PsiQuantum raised $1 billion at a $7 billion valuation, is building sites in Australia and Chicago, and reached the final phase of a separate DARPA programme. The risk is that until the big machine runs, it’s hard to verify the whole approach works.
Xanadu: build it from blocks
Canada’s Xanadu took a different route. In January 2025 it showed Aurora: 12 qubits on 35 photonic chips, linked by optical fibre into one machine. Not many qubits, but the idea matters: a big quantum computer can be assembled from identical modules, like servers in a data centre, and the links between them need no fridge. Xanadu also made DARPA’s Stage B, which means its plan is taken seriously.
Topological qubits: protection by design
All the qubits above fight noise with error correction, spending many physical qubits on one reliable one. There’s a more elegant idea: store information not in one spot but “spread out” between two distant particles called Majorana modes. Like a knot in a lace: light shaking won’t undo it; you have to untie it on purpose. If such qubits work, far less error correction will be needed.
Microsoft Majorana: watch the peer review
Microsoft has bet on topological qubits for years. In February 2025 it announced the Majorana 1 chip, but many physicists found the data insufficient to prove the qubits really are topological. In June 2026 came Majorana 2: the material was switched to lead, and lifetime rose from milliseconds to about 20 seconds. It’s still a preprint without peer review, and critics say more measurements are needed. Lesson: a loud announcement isn’t proof yet; wait for peer review and replication.
Silicon spins: a qubit like a transistor
One more contender: the spin of a single electron trapped in silicon under a tiny electrode. The device looks a lot like an ordinary transistor, so these qubits are small and can be made in existing fabs. In theory millions could fit on one chip. The approach is younger than the others, but three such companies made DARPA’s Stage B: Diraq, Quantum Motion and Silicon Quantum Computing. So keep an eye on it too.
Check yourself
- What’s the main problem with photonic qubits?
Correct answer: Photons get lost and barely interact with each other. A photon is low-noise but easily lost, and getting two photons to “talk” is hard. Hence complex gate schemes. - How should you treat the Majorana 2 announcement?
Correct answer: With interest, but wait for peer review and replication. The lifetime jump is impressive, but it’s a preprint, and critics point to missing measurements. The right stance is to wait for verification. - Why are silicon spin qubits interesting to investors?
Correct answer: They resemble transistors and are made in existing fabs. Small size and compatibility with the chip industry promise millions of qubits per chip, if quality can be raised.