Analog quantum computers and sustainable AI: the overlooked argument for why they might matter sooner than expected

[TomTiz]

One prediction from TQI's expert roundup for 2026 has received almost no mainstream coverage: as AI's compute appetite surges, analog quantum computers offer a more sustainable and efficient path forward, which could deliver the first meaningful quantum-enhanced AI applications sooner than many expect.

Analog quantum computers, including quantum annealers like D-Wave's systems and certain neutral atom configurations, do not use discrete gate operations but instead evolve quantum states continuously to find low-energy solutions. They are less general than gate-based quantum computers but consume dramatically less power for specific optimisation problems.

With AI data centre power consumption becoming a genuine infrastructure constraint and the carbon footprint of training runs increasingly scrutinised, the energy efficiency argument for quantum-assisted optimisation deserves more attention than it currently gets.

TQI's Expert Predictions on Quantum Technology in 2026

https://thequantuminsider.com/2025/12/30/tqis-expert-predictions-on-quantum-technology-in-2026/

[Ben]

D-Wave's annealing approach, often dismissed as not-really-quantum-computing, is the closest thing to a deployable quantum computer for specific optimisation problems right now. The Flatiron dispute about quantum advantage is about benchmarks. The energy efficiency claim is about practicality

[CMPunk]

The energy argument for quantum computing in the AI context is the one that could accelerate enterprise adoption faster than pure capability arguments. If quantum-assisted optimisation produces equivalent results at 10x lower energy cost that is a CFO conversation not just a CTO one

[PlanckLimit81]

Neutral atom analog approaches like Pasqal's are developing in exactly the direction this argument points. Their systems run at room temperature for sensing applications and have analog quantum simulation modes that are energy efficient relative to gate-based cryogenic systems

[BretHart]

The power consumption of AI training is now a genuine limiting factor on development pace. Any technology that lets you do equivalent inference or optimisation at lower energy cost gets funding. Quantum may enter AI infrastructure through the energy door rather than the capability door