Quantum-assisted artificial intelligence is beginning to move beyond research demonstrations and towards commercially deployable systems. But the emerging model is not one in which quantum computers replace GPUs. Instead, quantum processing units (QPUs) are increasingly being positioned as specialist accelerators working alongside conventional AI, GPU and high-performance computing (HPC) infrastructure.
Key Players
One example from earlier this year comes from Kipu Quantum. Kipu has developed an approach in which quantum feature extraction is used during AI model training, but the resulting information is transferred to a classical surrogate model for deployment. This allows subsequent inference to run on conventional hardware, potentially avoiding the cost and latency associated with repeatedly accessing a QPU.
The infrastructure needed to support such hybrid approaches is also becoming more tangible. Quantinuum and Oracle have announced plans to deploy the Helios quantum computer in a US Oracle Cloud Infrastructure AI data centre. The arrangement would make quantum processing available alongside existing cloud and AI resources, supporting hybrid quantum-classical workloads.
Meanwhile, SoftBank and Quantinuum are exploring the longer-term commercial model for quantum-AI data centres. Their work considers applications including materials discovery and graph-based analysis, such as telecommunications fraud detection, and asks which applications could become commercially feasible as quantum hardware improves. A further indication of the likely future architecture comes from Japan, where RIKEN and Quantinuum are integrating quantum computing with the Fugaku supercomputer and GPU-based infrastructure. Rather than assigning an entire problem to a quantum computer, different parts of a computational workflow can be directed to HPC systems, GPUs or QPUs according to their strengths.
NVIDIA, meanwhile, is approaching the convergence from the opposite direction by using AI to improve quantum computing itself. Its expanding quantum ecosystem applies AI to areas including quantum-processor calibration and error-correction decoding, alongside tools for developing and orchestrating fault-tolerant quantum applications.
These developments suggest that the near-term opportunity for quantum-assisted AI lies primarily in hybrid infrastructure, scientific AI, computational chemistry and materials, specialised feature extraction, and optimisation of quantum hardware, rather than general-purpose acceleration of large language models.
IP Angle
For intellectual-property watchers, areas of particular interest include quantum feature extraction and classical surrogate models, QPU-GPU workload orchestration, AI-based quantum error correction and integrated quantum/AI data-centre architectures.
The important qualification is that practical quantum advantage for mainstream AI has yet to be convincingly established. The commercial story is therefore one of convergence rather than replacement: AI + GPU/HPC + QPU, with quantum processors increasingly envisaged as specialised components of a broader computing architecture.
.png)
.png)