Nobel Prize in Physics Winner: The Quantum Leap That Changed Everything - John Martinis

Summary

David Friedberg interviews John Martinis, a 2025 Nobel Prize in Physics laureate, about the 1985-86 Berkeley experiment that showed quantum tunneling and discrete energy levels in a macroscopic Josephson-junction circuit, the foundation of today's superconducting qubits. Martinis walks through quantum mechanics, superconductivity and how that circuit evolved into Google's 53-qubit quantum-supremacy machine in 2019. He says current systems have roughly 50-100 noisy qubits versus the ~1 million needed for general-purpose, error-corrected computing, puts useful machines about 8-10 years away and admits the field carries more hype than reality. He describes his startup's plan to move qubit fabrication onto Applied Materials 300 mm processes and voices concern that China is near parity with Google's latest results; no investment recommendations are made.

• Martinis shared the 2025 Nobel Prize in Physics for a 1985-86 experiment with adviser John Clarke and Michel Devoret showing macroscopic quantum tunneling and quantized energy levels in a superconducting Josephson-junction circuit.
• He explains wave functions, probabilistic position and momentum, tunneling through 10-20-atom-thick insulators, and how Cooper pairs let supercurrents flow without loss.
• The Josephson-junction LC resonator became the superconducting qubit; his UCSB group built 5- and 9-qubit machines and Google's team ran the 53-qubit quantum-supremacy experiment in 2019.
• State of the field: about 50-100 fully controllable but noisy superconducting qubits today, neutral atoms as a promising newcomer, and roughly one million qubits needed for general-purpose error-corrected machines.
• Timeline: he targets useful systems in 8-10 years while acknowledging repeated 10-year predictions and more hype than reality.
• His startup plans a new fabrication generation on Applied Materials 300 mm processes with Synopsys and Hewlett Packard Enterprise as partners, aiming to leapfrog current methods and protect a US lead.
• On China: Chinese groups replicated Google's supremacy results and appear near parity with Google's latest work; Martinis worries results are withheld until Western publication.
• AI may assist with modeling and error-correction decoding, but he argues clean hardware and control remain the main performance drivers.