Technology, Connected - Quantum And AI Just Joined Forces

Episode Date: February 3, 2026

Matt Kinsella runs Infleqtion, a company building quantum computers. The biggest misconception about quantum computing is that it will replace classical computing. It won't. Quantum processors will sit above GPUs in data centers the same way GPUs sit above CPUs today. NVIDIA just built the bridge to make this work. It's called NVQ Link, and it changes how we think about the future of compute.NVIDIA announced NVQ Link in October 2024. It's the bridge between quantum computers and classical GPU clusters. Workloads pass seamlessly between them.Here's how it works in practice. Infleqtion and NVIDIA solved something called the Anderson Impurity Model - a photovoltaic problem in material science. Parts of it were solved on a GPU cluster. Parts that couldn't be solved by GPUs were solved on Infleqtion's quantum computer. Then they recombined to give the answer. This isn't commercially useful yet. But expand that over time and you could be looking at the future data center. One with three layers. CPUs at the bottom for general computing. GPUs in the middle for parallel processing and AI. QPUs at the top for problems that are quantum mechanical in nature. Workloads come in, get chopped up, each piece goes to the part of the stack best suited to solve it. Then results recombine.This is already happening. Infleqtion just announced a contract with the Army called Sapient Secure AI for PNT - position, navigation, and timing. It runs their quantum-inspired software on NVIDIA's Jetson edge GPUs. Small GPUs that don't have much memory. The software lets them ingest far more streaming data than normal. Video, speed, inertial motion. Then it recreates what GPS gives you - where you are in the world - by extrapolating from all those signals. Without GPS.Please enjoy the show.Cheers, Mark & Jeremy.--Other ways to connect with us:⁠Listen to every podcast⁠Follow us on ⁠Instagram⁠Follow us on ⁠X⁠Follow Mark on ⁠LinkedIn⁠Follow Jeremy on ⁠LinkedIn⁠Read our ⁠Substack⁠Email: [e-mail]--

This is a five minute short on quantum computing. If by the end of it, you haven't learned something, then email me personally at mark at thinkingonpaper.xyZ, and I'll make it up to you. One of the most important things to understand about quantum computing is it's not going to replace the things or the workloads that we're doing in classical computing. It's going to open up whole new workloads. And so I don't see this as a necessarily a disruptor to,

GPUs or classical computing at large, it will be an expansion of what humanity can throw compute at. Let's call it that. So the vision of the future data center that I have is just like GPUs layered in on top of CPUs to enable new use cases, you're going to have quantum processor units or QPUs sit above those. And workloads will get sent into the data center in the cloud and they'll be chopped up. And the parts that the various parts of the stack are best suited to solve, the workholds will be sent to those parts. And so the underlying infrastructure to do that is something that we focus a lot on and we focused a lot on it with Nvidia. Invidia is really taking a leadership position in that bridge between quantum and classical.

And so they announced something called NVQ link in October of this year, which is really exactly that. it is the bridge to allow workloads to pass really seamlessly between quantum computers and classical GPU clusters. And a good example of how that might look in the future is something we did with them just about a year ago, where we solved a very basic material science problem called the Anderson Impurity Model. It's a photovoltaic model. Parts of it were solved on a GPU cluster, but parts of it that wouldn't have been able to be solved by GPUs were solved on our quantum computer. and then they recombined to give the answer. And we talked a lot about how nature does quantum better than anyone.

Quantum is really the computational method of nature. And so the types of problems that quantum computers are going to be very good at solving are those that are really rooted in nature themselves. And so the ability to model the interactions of electrons when you're combining different molecules, that's inherently quantum mechanical in nature. and even the most powerful GPU cluster at the end of the day is boiling everything down to zeros and ones, right?

And that's not how nature fundamentally works. And so these types of nature-based problems break on classical computers because they can't be boiled down to that heuristic of zero or one. And so a lot of the problems, a lot of the parts of that Anderson-Mdreda model could be, you know, solved by the classical computer.

So it was the combination of the two that made us able to do something that had never been done before. I think about that as a very not commercially useful, example yet, but you expand that over time. And that's how we're going to build better batteries. So your iPhone having to be charged every night, you can build a battery that lasts for a year maybe, right, or much longer. And or help solve the Haber Bosch process, which is what we use to build fertilizer. And we've been trying for 100 years to find a better catalyst. But it's just too complex to model that interaction. That's the type of thing that quantum computers will be able to do in tandem with classical computers. Is it fair to think about classical computing as a bit of like a brute force modality

and then quantum as more of a finesse driven modality? It's not a bad way to think about it. You are boiling all of life's problems down to Boolean zero or one logic. And what we can do by boiling life's problems down to that type of logic is incredible, right? It's amazing what you can do with doing that zero or one billions of times. but that's not fundamentally how nature works. And so, yes, we're kind of brute forcing these problems. And that's why I think you see the power consumption issues

when we're starting to throw classical computing at problems that it maybe wasn't well suited to solve. Could you speak a bit about the algorithms or the software that you're running on those as well? I do think software isn't talked enough about in quantum computing. Even at the most basic level, we're shooting atoms at lasers, but it's not like I'm sitting there with a laser gun shooting. shooting the atoms, we are programming these lasers with software to do this at very, very precise, you know, literally atomic level precision. When you write software and going back to software for

quantum computers, you have to write it in a fundamentally different way because you actually have to code around some of the laws of quantum mechanics. And one of those laws is called the no cloning theorem, which is very interesting. You can't copy and paste quantum data, which copying and paste and data is like, it's all over the place. And class. classical software, right? And so when we write our software for our quantum computers, we write this kind of very different software that doesn't do some of the things that would be very natural to do in classical software development. And that means we have to sort of re-architect the memory. But what we did was we started to run this re-architected memory software on GPUs. And it turns out

that that actually provides some very interesting performance enhancements to GPUs, largely around the enlargement of the context window. which is one of the fundamental scaling bottlenecks to GPUs. And that's because the memory is totally re-architected. It's quantum in nature. And so we've kind of been trying to bridge the gap between classical computing and quantum computing hardware by software and giving some of that quantum advantage to classical computing

kind of via software and rewinding or bringing forward the time to quantum advantage by software. So that's a kind of really interesting bridge we've been working on. That's really cool. You'll have to keep us posted there. That's really interesting. We just announced a really cool contract with the Army. It's called Sapient, Secure AI for PNT. And this is based on our what we call contextual machine learning software.

So it is our contextual machine learning software running on edge GPUs like Nvidia's Jetson, so small GPUs that don't have a ton of power or memory on them, and allowing them to ingest far more streaming data. So think of like video data, speed data, inertial motion data, and process all of that multimodal sensor data right there on the edge, which would have normally overwhelmed the memory capabilities of that GPU. But this allows them to expand it and then recreate some of the basic things that GPS is giving you, like where you are in the world, by extrapolating by all the different external signals

that can be brought in onto this GPU. And so that's where the name Secure AI for P&T comes from. And so that's a kind of very interesting near-term use case that we're working with the Army on. But yeah, it wouldn't be possible without this software.