Quantum computing in the cloud: what can you do with it today?

A few years ago, quantum computing was mostly for research labs. Today, a business can buy compute time on real quantum computers through the cloud, much like renting a server. But what concrete value does it offer today? And where should you be careful?

How does quantum in the cloud work?

A quantum computer sits in a specialised lab, often at extremely low temperatures or in a vacuum. You don't need to buy or maintain one. Through platforms such as IBM Quantum and Amazon Braket, you send a program, a so-called quantum circuit, to the hardware. It is scheduled, executed, and you get the results back.

You write such a circuit with open software like Qiskit or PennyLane, usually in Python. The same code can often run on hardware from different vendors using different technologies: superconducting qubits, trapped ions or neutral atoms.

What can it do today?

To be honest: today's quantum computers are still small and make errors. For most business problems, a good classical computer is still faster and cheaper today. There are, however, meaningful uses:

• Exploring optimisation. Planning, scheduling, routing and portfolio problems can be translated into a form (a QUBO model) that quantum hardware can solve. In practice you combine this with classical methods. That is called hybrid computing.
• Simulating molecules and materials. This is where quantum computers will eventually make the biggest difference, for example in pharma, chemistry and battery research. Today it still involves small systems.
• Building knowledge. Teams that start learning to think in quantum algorithms now will be ready when the hardware matures. For research, finance and logistics teams that is a strategic investment.

The role of simulators and GPUs

Much quantum work doesn't happen on a quantum computer at all, but on a simulator: regular hardware that mimics what a quantum computer does. It is cheaper, faster to iterate on and error-free. That lets you perfect an algorithm before spending expensive time on real hardware.

Simulating qubits does take a lot of memory. Every extra qubit doubles the amount needed. That is why serious simulations run on GPU servers with plenty of memory, using software such as NVIDIA CUDA-Q and cuQuantum. Depending on the algorithm, you can simulate dozens of qubits this way.

How to approach it

1. Pick a concrete problem, such as a planning problem that currently takes a lot of compute time or manual work.
2. Set the classical benchmark. How well and how fast does your current method solve it? Without that comparison you never know whether quantum adds anything.
3. Simulate first on GPU servers and refine the model.
4. Test on real hardware from one or more vendors.
5. Decide based on numbers: move to production, park it until the hardware matures, or solve it classically.

What to watch out for

• Cost. Time on quantum hardware is billed per task or per unit of time and adds up fast. Budget limits are not a luxury.
• Hype. Be wary of promises like "a thousand times faster". Always ask for the comparison with the best classical method.
• Vendor lock-in. The hardware is evolving quickly. Use open standards so you can switch.
• Data. Where is your data processed? For sensitive data, staying within the EU matters.

How we help

Nuraghi Cloud arranges access to quantum computers from multiple vendors, provides GPU servers in European data centres for simulation, and manages your workloads in an EU cloud. One point of contact, one invoice, and always an honest comparison with the classical solution first. Curious what it could mean for your problem? Request a quantum pilot.