The question of exactly where a quantum processor should be located relative to classical computing infrastructure has ceased to be purely theoretical. A team of analysts from AWS, Nvidia, Lawrence Berkeley National Laboratory, and NASA has presented a model that allows for an objective assessment of the need for physical proximity of quantum equipment to a supercomputer.

The key takeaway: there is no universal answer. Infrastructure requirements are strictly tied to the type of problem being solved and the architecture of the algorithm.

Two Levels of Interaction: Control and Hybrid Computing

The researchers identified two fundamentally different scenarios for the operation of a quantum-classical pairing. The first is real-time equipment control. This includes calibration of physical qubits and, critically, data processing for quantum error correction. In this mode, the classical system must analyze control measurements and issue corrective commands within fractions of a microsecond. The slightest delay renders the correction meaningless. Obviously, for such tasks, remote access via cloud networks is absolutely unacceptable. Only a direct, physically close connection is viable.

The second scenario involves hybrid algorithms, where the classical and quantum processors alternately perform parts of the calculation. This is where things get more complex and interesting.

When Latency Kills Performance

The authors broke down the execution time of a hybrid algorithm into three components: classical computation, quantum processor operation, and data exchange. If the main load falls on classical processing, reducing network latency has virtually no effect on overall speed. An example is sample-based quantum diagonalization (SQD) for quantum chemistry problems. An experiment on a 77-qubit IBM Heron processor showed that the share of communication overhead is a negligible 0.0001. Such an algorithm can be safely run via the cloud.

A completely different picture emerges with the quantum-enhanced Markov chain Monte Carlo (QE-MCMC) method. Testing on a 10-qubit IBM device revealed that, with remote access, communication overhead exceeds the computation time itself by approximately 1000 times. However, the paradox is that this algorithm does not require a powerful supercomputer. A small but high-speed classical controller located in close proximity to the quantum chip is sufficient.

Error Correction as a Driver of Convergence

As we transition to fault-tolerant quantum systems with logical qubits, the role of classical infrastructure will only increase. The system must continuously process a stream of measurements, instantly detect failures, and send corrective commands. The speed of the classical part will begin to directly dictate the pace of the quantum computer's operation. The authors rightly expect that large fault-tolerant installations will require much tighter integration with high-performance computing systems.

My comment: This work shatters the myth that quantum computing is simply "another server in the cloud." The reality is that for a number of critically important tasks, including error correction, we will need a completely new class of hybrid architectures where quantum and classical processors coexist at the level of a single system. Investors and infrastructure developers should closely monitor this direction — this is where key technological barriers and opportunities will be shaped.