Quest for Qubits: Chipmakers Join Quantum Fray

Chipmakers are increasingly seeking to use their experience developing classical CPUs and GPUs to stake a role in the quantum computing ecosystem. However, compared with quantum startups and the technology multinationals dominating cloud quantum services, they are taking a more cautious approach.

Nvidia set out its quantum stall at its recent GTC event, including a world-first GPU-accelerated quantum computing system, the Nvidia DGX Quantum. By contrast, Intel has developed a silicon-spin quantum processor.

In this Q&A, Bob Sutor, vice president and practice lead for emerging technologies with research and advisory firm The Futurum Group, discusses the significance of chipmakers joining the quantum race. Sutor was previously with IBM for 39 years before joining quantum computing and sensing company Infleqtion.

Enter Quantum: What’s your take on Nvidia’s recent quantum announcements?

Bob Sutor: Obviously, the new chip is a good thing. Anything you might do with a very powerful GPU that factors in previous results you had with GPUs and semi-quantum simulation, that's completely state of the art.

GPUs are going to get better. Nvidia certainly has the best quantum circuit simulators based on what they can do on GPUs and the amount of memory they have access to. You just can’t do this on a CPU.

As quantum computers grow in capability, why are companies like Nvidia investing in simulators?

Tim Costa (Nvidia’s HPC and quantum computing team leader) was asked this question at GTC. It's one of these questions that you don't want to answer because it's the main thing you do.

First of all, you can't debug real quantum computers very well. If you can simplify your problem to get some confidence using a simulator, maybe in a small model of what you're doing, then you feel better about running it on a real quantum computer.

Simulators are also very good for education and experimentation, so they will continue to have value. I could easily run a small quantum simulator on my desktop; I  don't have to send it off and wait until a result comes back.

One of the angles to think about this is this combination of high-performance computing with quantum computers. They have laid out a very specific architecture, with quantum cloud in particular, that’s a set of services meant to bridge the classical HPC with quantum. I think they've done that very well.

Intel developed a 12-qubit silicon-spin qubit leveraging its silicon fabrication expertise and made it available to researchers last year. How does that compare with Nvidia’s approach?

I really like what Intel is doing with the silicon spin qubits. When you see a list of different modalities for quantum computing – superconducting neutral atoms, ions, photons and so on – there are far fewer people taking the silicon approach than, say, superconducting.

But it seems to be a slow and steady effort. They’re not rushing to market in favor of doing good research. Think of their perspective as a semiconductor company. If you take this long view on quantum computing – by long view I mean decades – it'll get there, but they’re not rushing to be the first to get to 1,000 logical qubits.

But over the last few years, they’ve seen some very nice improvements in scalability. They’re still small numbers, but they're tackling the problems one by one in a very systematic way, to be consistent with what they're doing from a chip level.

What factors will shape the next phase of quantum computing?

Once quantum computers become commoditized, users will need to know why one is better than another. You could look at how long it would take to run a problem on one platform compared with another.  Do you really care if it's one hour versus two hours? If it uses less energy, you might be willing to take more time.

As we enter the middle phase of quantum computer development, I think sustainability is going to become much more important. So far, with quantum computers, we’re doing what we've never been able to do before so it uses a lot of energy. I think that will become a very compelling factor in choosing between platforms.

Those are hard questions, and I think we need to hold manufacturers’ feet to the fire on this.