Why quantum scales on compute-per-watt, not qubit count
AI never planned for its own success. Quantum still can, if the industry moves now.
Power usage and cooling systems are being upgraded in response to the sudden surge in demand from AI applications, as reported by UN researchers. Quantum computing stands ready to address this growing need. AI hyperscalers prioritize the amount of useful computation delivered for each dollar spent, with power usage and depreciation making up the bulk of the total cost of ownership.
As a result, data centers are being constructed with power constraints in mind, potentially even in space if growth persists. However, the industry should shift its focus from the number of qubits to compute-per-watt, as qubit count alone does not indicate economic returns. Superconducting qubit processors, a popular quantum computing platform, have been limited to around 100 qubits for over a decade due to the majority of chip surface being occupied by wiring and qubit control.
Increasing processing power through networking multiple small processors is inefficient due to lossy connections and sparse relationships. To alleviate this issue, the quantum industry should concentrate on experience curves, where costs decrease as production volume expands. The classical history of solar panels and batteries demonstrates this principle, and quantum computers will follow suit if built at sufficient scales.
The transistor exemplifies this trend, with costs plummeting from a dollar to a fraction of a cent over decades of improved manufacturing and technical progress. For quantum computers to become economically viable, the cost per qubit must fall by at least a hundredfold. To achieve this, an open architecture involving various companies specializing in different components, such as processors, cryogenics, and control electronics, is essential.
With the transition from research to engineering, there is a growing need for supply chains and manufacturers, which are currently scarce. Quantum computers hold immense potential for transforming drug discovery, material development, and machine learning, particularly in relation to AI. However, the focus must shift towards economic metrics like compute-per-watt to prevent the creation of costly demonstrations rather than practical applications.
Written by urgent.news from TechRadar's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
