⚛️ From Majorana 1 to 2: Microsoft unveils a quantum chip 1,000 times more reliable

Microsoft has just unveiled a quantum chip whose quantum bits (qubits) remain stable 1,000 times longer than those of its predecessor. Instead of a few milliseconds, coherence now lasts on average 20 seconds, and sometimes up to a minute. A leap that makes researchers dream, hoping for a reliable quantum computer by 2029, much earlier than expected. However, this announcement raises doubts among specialists.


What exactly is it? The Majorana 2 chip uses so-called topological qubits. Their principle is based on a 90-year-old idea by physicist Ettore Majorana: a particle can be its own antiparticle. By combining a semiconductor and a superconductor, Microsoft creates "topoconductors." The qubits would be more stable and less energy-hungry than those based on superconducting metals used by IBM or Google.

The transition from Majorana 1 to Majorana 2 required changes in materials. Instead of aluminum, lead now protects the qubits from external disturbances (electromagnetic waves, cosmic rays). The semiconductor has also been improved with an alloy of indium arsenide and indium antimonide, doubling the "topological barrier" that isolates the qubits from noise. Result: reliability multiplied by a thousand.

To design this chip, Microsoft used artificial intelligence. Engineers had to add impurities to the crystal structure with atomic precision. Too much or poorly dosed, they would have disrupted everything. Thanks to AI agents, they were able to simulate thousands of configurations in parallel and find the ideal recipe in one go, instead of taking weeks. AI also automated measurements, significantly accelerating experiments.

Despite these advances, the scientific community remains cautious. Some physicists doubt that Microsoft has truly proven the existence of Majorana zero modes (MZM), the basis of these qubits. The preprint of the study has not yet been peer-reviewed, and the previous one of the same type was never published in a journal. "We must wait for independent reproduction before making a judgment," tempers Yuval Boger, a researcher at QuEra.

Microsoft claims to halve the time separating us from the first "fault-tolerant" quantum computer, capable of correcting its own errors. The goal is now 2029. This machine could solve problems inaccessible to current supercomputers. But other approaches exist: superconducting qubits, neutral atoms, photons... Each has its advantages.