John Martinis, a pioneering physicist whose work twice re-charted the course of quantum computing, has embarked on a new mission: to create a machine so advanced that it will redefine the phrase “state of the art.” Below, we examine what he has achieved, why his latest project matters, and how it could accelerate the arrival of fault-tolerant quantum computers.
Who Is John Martinis?
Martinis spent decades at the University of California, Santa Barbara (UCSB), where his laboratory became one of the first to demonstrate high-fidelity superconducting qubits. In 2014 he was recruited by Google to spearhead its Quantum AI team, ultimately leading the 2019 quantum supremacy experiment—an event that proved a programmable quantum processor could outperform the best classical supercomputer on a specific task.
The Two Quantum “Revolutions” He Already Led
1. Pioneering Superconducting Qubits
• Martinis and colleagues demonstrated that Josephson-junction–based qubits could reach coherence times long enough for quantum algorithms.
• They introduced techniques such as capacitive shunting and 3D microwave cavities, significantly reducing decoherence.
• Their work provided a template for virtually every industrial superconducting-qubit program that followed (IBM, Rigetti, Google, and others).
2. Google’s Quantum Supremacy Experiment
• In October 2019 the 53-qubit Sycamore processor sampled random quantum circuits in 200 seconds—a task estimated to take the world’s best classical supercomputer 10,000 years.
• Although the practical utility of the specific calculation was limited, the result was a watershed moment, establishing a clear performance gap between quantum and classical machines.
The “Third Revolution”: Building an Ultra-Powerful, Fault-Tolerant Quantum Computer
After leaving Google in 2020, Martinis began rethinking the hardware stack from the ground up. Rather than add qubits piecemeal, he argues that the next era demands fully integrated error correction and a path to millions of physical qubits.
The Technical Roadmap
Surface-Code Error Correction
Martinis plans to employ the surface code, which tolerates approximately 1% error rates. The goal is to reduce two-qubit gate errors below 0.1%, thereby lowering the overhead of logical qubits.
Materials & Fabrication
• Switching from aluminum to niobium-titanium-nitride (NbTiN) can suppress loss channels caused by surface oxides.
• Through-silicon vias and 3D integration will route control lines without introducing thermal or electromagnetic noise.
Control Electronics
Integrated cryogenic control chips could move classical processors from room temperature to 4 K, trimming latency and cabling complexity.
Scaling Strategy
Rather than fabricate monolithic wafers with thousands of qubits (which suffer from yield problems), Martinis advocates a tileable module approach: a small fault-tolerant unit (≈1,000 physical qubits) that links to neighbors via superconducting resonators or photonic interconnects.
Why “Most Powerful” Is More Than a Marketing Claim
The systems Martinis envisions would not merely run larger instances of today’s quantum-supremacy benchmarks. Instead, they aim for:
- Logical qubits with error rates <10–15, enabling day-long computations without mid-circuit resets.
- Quantum volume orders of magnitude beyond current devices, sufficient for chemistry, logistics optimisation, and certain cryptanalytic tasks.
- Algorithmic breadth: the ability to execute resource-intensive algorithms such as quantum phase estimation, HHL linear-solver routines, and large-scale variational simulations.
Key Challenges Ahead
Thermal Load: A million-qubit cryostat must remove tens of milliwatts of heat at millikelvin temperatures, a daunting requirement even for dilution refrigerators.
Fabrication Yield: A single defective junction can compromise an entire tile, demanding near-semiconductor-class clean-room standards.
Classical Processing: Error-correction decoding at kilohertz cycle times will require petaflops of real-time classical computation colocated with the cryogenic stack.
Potential Impact on Science and Industry
• Drug Discovery: Exactly simulating active sites of complex biomolecules (≈100 – 200 spin-orbitals) could shorten R&D timelines for antibiotics and antivirals.
• Materials Science: Predicting high-temperature superconductors or next-generation battery compounds from first principles.
• Finance & Logistics: Quantum Monte Carlo and optimisation algorithms may offer exponential or polynomial speedups, depending on problem structure.
• Cryptography: A sufficiently large fault-tolerant machine could run Shor’s algorithm, breaking widely deployed RSA-2048. This accelerates the race toward post-quantum cryptographic standards.
What Comes Next?
Martinis’ immediate objective is a demonstrator system with roughly 100 error-corrected logical qubits—enough to outpace classical simulations of quantum chemistry. He projects this could be achieved within five to seven years if fabrication milestones hold.
Whether or not Martinis ultimately delivers the world’s most powerful quantum computer, his track record suggests that the effort itself will push the field forward. Each attempt to tame noise, scale up qubit counts, and streamline control electronics makes quantum computing less of a laboratory curiosity and more of an engineering discipline—exactly the transition needed to move from proof-of-concept to practical utility.
