Quantum computing's real bottleneck is compute per watt, not qubit count, argues one hardware CEO

[BiasField78]

Matt Rijlaarsdam, chief executive and co-founder of quantum hardware company QuantWare, has laid out a pointed argument against how the quantum computing industry currently measures and markets its own progress. His central claim is blunt, qubit count alone says nothing about economic returns on a system, and the industry's continued obsession with headline qubit numbers is actively obscuring the metric that will actually determine whether quantum computing ever becomes commercially viable, compute per watt.

The technical reality underneath current superconducting qubit processors makes his case for him. More than 90 percent of the physical surface area on a typical superconducting chip is taken up entirely by the wiring needed to control the qubits and read out their results, rather than by the qubits doing the actual computation. That ratio, Rijlaarsdam points out, has stubbornly failed to improve in nearly a decade of development, meaning most of a chip's real estate has always been dedicated to plumbing rather than genuine computational capacity, a structural inefficiency that scaling qubit counts alone does nothing to fix.

The networking penalty compounds the problem further. Connecting multiple smaller quantum processors together to reach a higher combined qubit count is, in Rijlaarsdam's framing, lossy and produces only sparse connections between the linked systems, requiring exponentially more power just to maintain coherence across the connected network than the actual computation itself consumes. In other words, chasing a bigger headline qubit number through networking multiple chips together can make the overall system considerably less efficient rather than more capable in any economically meaningful sense.

His broader economic argument draws a direct parallel to how AI hyperscalers already think about their own infrastructure. Those companies increasingly care intensely about compute per watt specifically because power availability, not raw capital, is now the binding constraint on data centre expansion, and Rijlaarsdam argues that exact same economic logic applies just as forcefully to quantum computing systems, even though the industry's marketing has not caught up to that reality yet.

His proposed fix is structural rather than purely rhetorical. He calls for the industry to adopt what he terms a Quantum Open Architecture, with specialised suppliers each building distinct layers of the technology stack rather than every company trying to vertically build everything itself, enabling genuine volume manufacturing and the kind of steady cost reduction through experience curves that transistor manufacturing achieved decades ago. The urgency in his argument comes from timing, supply chain development of this kind takes years or even decades to mature, and the industry level decisions being made right now will determine whether quantum computing eventually becomes a genuinely practical technology or stays permanently stuck as an expensive series of demonstrations.

Why quantum scales on compute-per-watt, not qubit countwww.techradar.com

[SockPuppet42]

The comparison to AI hyperscalers caring about compute per watt because power availability constrains data centre expansion is a smart parallel, and it is worth remembering those companies only started prioritising that metric once power genuinely became the binding constraint rather than capital. Quantum computing may need to hit that same wall before the industry actually shifts its marketing to match.

[Skibidi98]

Ninety percent of chip surface area going to control wiring rather than actual qubits is a statistic that should get repeated a lot more often whenever someone quotes a headline qubit count without any context. That ratio not improving in nearly a decade suggests this is a genuinely hard engineering problem rather than something scaling will just solve on its own.

[WWEHarry78]

The networking penalty point is the part that actually surprised me most here. I had always assumed connecting smaller chips together was a reasonably clean path to scaling total qubit count, not something that introduces this much of an efficiency tax on the system as a whole.

[Natalie]

Supply chains taking years or decades to mature is the part of this argument that gives it real urgency rather than just being an abstract efficiency critique. If the industry keeps optimising for qubit count headlines for another five years before pivoting to this framing, that is five years of supply chain development potentially wasted on the wrong priority.

[Y2J_WCW]

Coming from someone who runs an actual quantum hardware company rather than an outside commentator gives this argument real weight, though it is also worth noting QuantWare has an obvious commercial interest in pushing a framing that favours engineering efficiency over the qubit count race that bigger, more headline focused competitors currently win on marketing terms.

[Dragon49]

Quantum Open Architecture with specialised suppliers building distinct layers is basically calling for the same modular supply chain structure that made the semiconductor industry's cost curve possible over the last several decades. Whether quantum computing's underlying physics allows for that same kind of clean layer separation is an open question though