Quantum computing is often positioned as something distant, powerful, complex and still years away from practical use. That perception is misleading. Rather than arriving as a sudden breakthrough, quantum is emerging quietly through hybrid architectures, digital annealing techniques and integration with AI. Much like artificial intelligence a decade ago, the early signals are already visible. As with AI, by the time it feels "ready," the organisations that understand it will already have moved ahead. The question then, is not "when will quantum arrive?" but rather "will we recognise where it fits before it arrives?"
A Different Kind of Computing Problem
Quantum computing is important not because it is universally faster, but because it efficiently addresses a specific category of problem. Where classical systems step through possibilities (either sequentially or using massive parallelism) quantum can explore vast solution spaces in fundamentally new ways using quantum concepts such as "superposition" and "entanglement". The difference becomes clear when we look at real world systems.
Consider the challenge of optimising traffic flow across a busy intersection. At first glance, it seems like a natural fit for AI. Cameras and sensors can detect vehicles, machine learning models predict congestion and control systems adjust traffic lights dynamically. In practice, many cities are already doing this.
But when you scale the problem from an intersection to an entire city, the limitations become obvious. Thousands of intersections, each with multiple signal phases and timing options, create an explosion of possible configurations. AI can learn patterns and make local improvements, but it struggles to guarantee optimal outcomes across the entire network. Training such systems is computationally intensive, and even then the results are often "good enough" rather than provably optimal. This is where quantum and digital annealing approaches begin to show their value.
Where Quantum Fits: Optimisation at Scale
Problems like city-wide traffic control belong to a class known as combinatorial optimisation. They are not limited by data availability or software design, but by the sheer size of the solution space. This is the same class of problem found in airline scheduling, power grid balancing and financial portfolio optimisation.
Quantum and digital annealers are designed specifically for this type of challenge. Instead of learning policies or heuristics, they search directly for optimal or near-optimal solutions across massive possibilities. In the traffic example, each signal timing can be represented as a variable in an optimisation model with a cost function representing congestion or total travel time. The solver then explores the entire system holistically rather than incrementally. The result is not a replacement for AI, but a complement to it. The most effective approach is hybrid. AI handles the sensing and prediction layers by detecting vehicles, forecasting demand and identifying patterns. Quantum solvers take on the global optimisation problem, determining how the entire network should behave. Classical systems can then execute those decisions locally, handling any real-time adjustments and edge cases. This hybrid combination changes what is possible.
Digital Annealing: Acting Before Quantum Arrives
While fully fault-tolerant quantum computers are still evolving, organisations do not need to wait to start adopting this way of thinking. Digital annealing already provides a practical bridge. Digital annealers run on classical hardware but use algorithms inspired by quantum mechanics to solve optimisation problems. They are already being used in areas such as supply chain routing, workforce scheduling and financial risk modelling. In many cases, they deliver near real-time solutions to problems that would otherwise require hours of computation. However, they require a change in mindset. Problems need to be reformulated as optimisation challenges, often expressed mathematically as 'QUBO' models, a type of quadratic function. This pushes teams away from rule-based or purely predictive approaches toward a more holistic view of systems.
Why Digital Annealing Is More Sustainable Than Deep Learning at Scale
An often overlooked aspect of this hybrid shift is energy. Deep learning and LLM model training can be extremely energy intensive. Solving a problem like city-wide traffic optimisation using AI alone typically requires repeated training cycles over enormous state spaces. Each iteration consumes significant GPU resources and any material change to the environment (new roads, changing traffic patterns, policy updates) will trigger re-training. Over time this leads to substantial cumulative energy consumption. Digital annealing approaches the problem differently. Rather than learning through repeated trial and error, it formulates the optimisation challenge mathematically and solves it directly. This eliminates the need for prolonged training cycles and dramatically reduces compute overhead. The difference becomes particularly stark at scale. A deep learning approach might require days or weeks of GPU time to converge on a workable solution for a large complex system. By contrast a digital annealer can evaluate the optimisation problem in seconds or minutes. The energy consumption is tied to solving the problem once, not repeatedly learning it. There is also another efficiency advantage. Digital Annealers are designed specifically for combinatorial optimisation, meaning they avoid the inefficiencies of general purpose architectures being applied to specialised problems. In effect, they deliver more useful computation per unit of energy consumed. In scenarios where optimisation problems must be solved repeatedly e.g. logistics networks, energy grids or real-time transport systems, this difference compounds. Over time digital annealing can deliver orders of magnitude improvements in energy efficiency compared to AI-only approaches attempting to solve the same class of problem.
From Single Processors to Compute Fabrics
The future is not about replacing AI with Quantum, but about orchestration. AI systems will need to dynamically select the right form of compute (classical or quantum) based on the nature of the task. Modern compute architectures are no longer built around a single type of processor. Instead, they are becoming compute fabrics, combining CPUs for orchestration, GPUs for parallel processing, NPUs for efficient inference and soon QPUs for specialised optimisation tasks. Quantum processing units will not replace this ecosystem, they will join it. Already, platforms such as Azure Quantum and AWS Braket are making quantum and digital annealing solvers accessible alongside classical compute. Over time, developers will not need to think about qubits directly. Instead, they will define problems and the platform will determine how best to solve them. This is why quantum advantage is unlikely to appear as a dramatic leap. It will emerge incrementally as specific workloads begin to run more efficiently or produce better outcomes on hybrid architectures.
Why This Matters Now
For most organisations, quantum is not an immediate procurement decision. It is a strategic awareness challenge. If we learnt anything from the AI boom its that demand for understanding emerges long before demand for solutions. Organisations begin by asking where the technology applies, how to prepare and what risks exist. Those who can answer these questions shape the conversation early. Quantum relevant problems are already present in many enterprises. They appear wherever systems become too complex to optimise effectively, where trade offs multiply and where small improvements have large impacts. In these spaces, even small advances in optimisation can translate into significant financial and operational value. The risk is not adopting quantum too early, it is failing to recognise which problems are "quantum shaped" until competitors have already begun solving them differently.
Final Thought
Quantum computing will likely not arrive with a moment of disruption that forces immediate action. It will arrive gradually, embedded in platforms, integrated into workflows and abstracted behind familiar tools. While all attention is locked onto AI, it will quietly redefine what good looks like in optimisation, efficiency and sustainability. By that point, the advantage will already belong to those who understood the shift early.