The next phase of quantum computing will be defined less by headline-grabbing processor demos and more by whether the industry can solve a broader systems challenge across hardware, control, and architecture.

That was the message from a recent roundtable featuring 2025 Nobel laureate John Martinis, Quantum Machines co-founder and co-CEO Yonatan Cohen, Alice & Bob head of chip design Nicolas Didier, and MIT researcher Fabrizio Berritta. The panel argued that the path from promising processors to practical fault-tolerant machines will require advances well beyond the qubit itself.

Martinis said the industry cannot afford to treat scaling as a single bottleneck. Moving from hundreds of qubits to fault-tolerant systems with millions will require changes across manufacturing, wiring, control, and system cost. He pointed to the need to reduce the cost of control infrastructure and improve how qubits are built and connected.

Cohen made a similar argument from the control side, saying quantum systems will only scale if calibration and error correction are brought closer to the hardware and more tightly linked to classical compute. Calibration loops that take too long become a practical blocker as systems grow, he said, adding that quantum machines must be designed as hybrid quantum-classical architectures rather than isolated processors.

That systems view also underpins Quantum Machines’ newly launched Open Acceleration Stack, which is designed to connect quantum processing units with classical accelerators for calibration, decoding, and eventually applications. Cohen said the industry still faces open questions about which classical processors are best suited to different tasks, but argued that modularity will matter more than locking into a single architecture too early.

Berritta’s recent MIT work added another calibration dimension. He said the team used faster control hardware to track qubit energy decay, or T1, at much finer timescales and found it could change by an order of magnitude within tens-of-milliseconds. That was scientifically interesting, he said, but also “a nightmare from a calibration point of view,” because it revealed a fast-moving parameter that had previously been averaged out.

For Martinis, the results highlight how better measurement can expose new failure modes, not just improved performance. Faster calibration may help researchers understand qubits in real time, but it also reinforces the case for improving fabrication and reducing materials defects.

Didier argued that cat qubits could ease some of the scaling burden by lowering error-correction overhead. But even then, he said, the wider stack matters. Getting useful logical qubits sooner would allow the industry to evaluate the surrounding ecosystem earlier, including how quantum processors integrate with high-performance computing (HPC) environments and hybrid workflows.

The discussion suggested that scaling challenges are shifting from individual qubits to the systems that support them. Calibration speed, control infrastructure, decoding, and software are now emerging as key constraints as the industry pushes toward fault-tolerant machines.