When we think about a technology we can’t live without, the mobile phone is often near the top of the list. It has become far more than a telephone and influences how we work, communicate, conduct research, navigate, consume media, read, shop, and organize our lives. This outcome was far from obvious during the early 1990s. Mobile phones were expensive, coverage was incomplete, batteries were limited, and our range of applications was narrow. But the technology was already useful—it didn’t need to become a smartphone before it created value.

What followed wasn’t the result of one decisive breakthrough. Devices improved. Networks expanded. Costs fell. Signal processing became more sophisticated. Software ecosystems emerged. Each layer reinforced the others, and adoption accelerated. In the U.K., mobile connections exceeded the population by 2004.

Quantum computing has reached a similar point. It’s not waiting for a single future moment when it suddenly becomes useful. Its practical phase has already begun through hybrid quantum computing, in which quantum processors, quantum-inspired algorithms, artificial intelligence, tensor networks, central processing units (CPUs), graphics processing units (GPUs), and high-performance computing (HPC) are integrated into complete industrial workflows. More powerful quantum processing units (QPUs) will dramatically expand what these systems can achieve, and fault tolerance will accelerate the transition. But neither one is the starting point of the market.

Quantum computing is already in its practical phase

Much of the debate around quantum computing treats the technology as binary. Either a fully fault-tolerant machine capable of transforming entire industries is imminent, or quantum computing has no practical value at all. Both positions miss how technologies actually develop. In 1995, the mobile phone was neither a useless experiment nor the mature platform we know today. It was a constrained but already valuable technology embedded inside an ecosystem improving rapidly. Devices became better, networks expanded, costs fell, software evolved, and new applications emerged. Each layer reinforced the others. Quantum computing is following a similar path.

Today’s quantum processors remain noisy, and circuit depth is limited. They can’t yet execute many of the large, error-corrected algorithms expected to transform areas such as chemistry, materials science, simulation, and cryptography. But it doesn’t mean quantum technology has no practical role today. The mistake is assuming an entire industrial problem must run on a quantum processor before quantum computing becomes useful. In reality, the most practical architecture is hybrid. A real problem may involve data preparation, simulation, machine learning, optimization, inference, validation, and decision making.

There’s no reason every stage should run on the same type of processor. CPUs remain strong for general-purpose computation. GPUs provide massive parallelism. Tensor networks can efficiently represent certain high-dimensional systems. Quantum-inspired algorithms bring methods from quantum information science to classical infrastructure. QPUs can be used selectively for optimization, sampling, inference, and other stages where their native properties contribute value.

Hybrid quantum computing emerges

The key is to preserve the real scale and complexity of the problem rather than reduce it to a small proof of principle merely to fit on current hardware. Hybrid quantum computing combines CPUs, GPUs, HPC, tensor networks, quantum-inspired methods, and QPUs based on the needs of the workflow. Its value is assessed using the same criteria applied to any other technology: Solution quality, runtime, scalability, accuracy, cost, and operational impact.

A proof of principle shows a quantum circuit can run. Hybrid quantum computing demonstrates quantum technology can help solve a commercially relevant problem. More capable QPUs will enable larger portions of these workflows to run natively on quantum hardware. They will eventually improve performance and broaden the range of applications, but organizations don’t need to wait for them before generating value.

Fault tolerance will change the slope of progress. Quantum error correction protects logical qubits by distributing information across many physical qubits. Once physical error rates fall below the relevant threshold, increasing the code distance can suppress logical errors exponentially. More hardware can then be converted not merely into a larger system but into increasingly reliable computation. Google’s Willow result was important because it experimentally demonstrated this below-threshold behavior. Increasing the size of the error-correcting code made the logical system more reliable, not less. Willow was not yet a commercially useful fault-tolerant machine, but it strengthened the evidence that future quantum processors can become substantially more capable as hardware, control systems, decoding, error correction, compilers, and algorithms improve together.

The cryptographic target is also moving in both directions. In 2021, estimates for factoring RSA-2048 using Shor’s algorithm were around 20 million noisy physical qubits under specific assumptions. By 2025, improved algorithms and resource management had reduced the estimate to fewer than one million. More recent architectural studies have proposed regimes below 100,000 physical qubits and, under more space-efficient neutral-atom assumptions, within the range of 10,000 to 14,000 qubits, although with significantly longer runtimes. These remain theoretical resource estimates, not experimental demonstrations, which rely on different assumptions about error rates, connectivity, codes, cycle times, and parallelism. But the trend is unmistakable: Hardware is improving while algorithms, circuit design, error correction, and system architecture are simultaneously reducing the amount of hardware required. And it’s why linear extrapolation can be misleading.

Cybersecurity implications are already present. Under the harvest-now-decrypt-later model, encrypted information can be collected today and retained until sufficiently powerful quantum computers become available. Information that must remain confidential for many years is at risk before a cryptographically relevant quantum computer exists. Post-quantum security is consequently not a future research topic but a deployment challenge today. Organizations must identify where cryptography is used, prioritize long-lived and high-value data, introduce crypto-agility, test standardized post-quantum algorithms, and update products and systems that may remain operational for decades.

Organizations should begin identifying difficult computational problems, benchmarking hybrid approaches, building internal expertise, and integrating quantum-compatible architectures now. You can use quantum-inspired algorithms on existing infrastructure, selectively incorporate QPUs, and prepare systems to absorb more capable quantum hardware as it becomes available. The mobile phone of 1995 was not the smartphone of today. It was more limited and more expensive, but it was real, useful, and improved quickly. This is a different way to think about quantum computing.

Hybrid quantum computing can already address real problems at industrial scale and complexity. Quantum-inspired methods already operate in production environments. Present-day QPUs already contribute to selected stages of computation. Post-quantum security can already be deployed. Fault-tolerant quantum computing will dramatically expand this market, but it will build on an existing ecosystem.

Quantum computing isn’t waiting to become useful—its practical phase has already begun.