In The Arena by TechArena - Infleqtion and the Rise of Hybrid CPU GPU Quantum Systems

Episode Date: July 3, 2026

In this episode of Data Insights, hosts Allyson Klein and Solidigm’s Jeniece Wnorowski sit down with Pranav Gokhale, CTO of Infleqtion, to explore where quantum computing is already delivering value and how hybrid CPU, GPU, and QPU systems will shape what comes next.

Welcome to Tech Arena, featuring authentic discussions between tech's leading innovators and our host, Alison Klein. Now, let's step into the arena. Welcome in the arena. My name is Allison Klein. We're coming to you from the Accelerated Compute Show in New York, and this is a Data Insights episode, which means Janice Norowski is with me. Welcome, Janice. Hi, Allison. We're going through a ride of tech innovation here at Accelerated. Tell me about what this episode is a video. so it is about and who you brought with you. We are all things knee-deep in quantum computing. So we are. Just catching up, but I'm super excited.

Today we have Pranav Golclai, CTO of Inflection. Welcome to the program, Pranav. Thanks so much, Alison and Janice. Yeah, we're excited to just dive right in. So Pran, this is the first time that Infliction's been on the show. Why don't you introduce the company and what it means to be the CTO of a quantum computing company? Sure. So Inflection builds quantum technology.

We build both quantum computers and quantum sensors. Those sound like very different things, computing versus sensing, but in fact, at their core, they're the same underlying technology, which is called neutral atoms. If you've heard the term qubits, quantum bits, we make qubits, neutral atom, qubits, and then we use those same atoms to sense things like time, like radio frequency communications,

and like acceleration or rotation. But the crown jewel of what we're building is the quantum computer, the quantum computer, and I'm sure we'll get to talk more about that over the next few minutes. That's awesome. Yeah. So you've spent a lot of time with software. You mentioned that a moment ago, building quantum computing software. But now you're at a leading company in Flection, which is so freaking cool. I'm glad you brought that. But how do you see quantum computing shifting and lifting into the broader compute landscape alongside GPUs, HPC, and just overall classical systems? Sure. Maybe I can almost weave that into a bit of my personal background, too.

I did my undergrad in computer science and worked in Silicon Valley for two years at tech startups, which I loved. There was a moment where I woke up at 4 a.m. and I had this kind of epiphany of I love my job and working on at Quora, question answer website, doing some AI work. That was really cool. But what struck me was that everything I was doing was one to end. And I had in my undergrad, in fact, all the way since middle and high school, been very interested in quantum computing and did some research at it in NIST, where I grew up in Maryland. And it was a technology that offered a zero to one paradigm for computing that would unlock some fundamentally new scaling curves. So today, a lot of computing has been running at an amazing breakthrough pace, but we're hitting some saturation on how fast AI models can scale up, how fast materials can be crunched or new chemistry can be solved.

And quantum computing enables us to, for the first time, unlock a certain class of problems that have been completely inaccessible to humidity. And that's a big ambition, but this technology is now at the point where it's starting to knock on that door. That's amazing. Now, one of the things that I read in my research is that you have talked about how quantum

is not a replacement for classical computing, but the two will work hand in hand. Tell me about your vision there. Sure. There's a lot of precedent for this just in classical traditional computing, too. With GPU has not replaced CPU, but it's been a co-processor.

And in the same way, we think that CPU and GPU are going to be co-processors to QPUs or quantum processing units. And the best example I could give of this is just last week. I was at GTC, the NVIDIA conference, because for the first time in their conference floor history, they've had quantum computing on site at the NVIDA booth. And specifically, they had three machines, one of which was inflections at the NVIDIA booth, connected using a technology called NVQ link to GPU. So big picture, what we see happening is that the world's hardest computational problems, very high value problems, are you going to get sliced and diced. Some pieces are going to get handled by GPU. Some pieces will be handled by CPU,

but the kernels that are most computationally demanding will live on the QPU, and you'll have software that orchestrates between these three compute technologies and gives you kind of a unified compute fabric for the programmer, but under the hood, specialization to solve each piece. Nice. So within that kind of hybrid stack, where do you see the earliest point of synergy emerging between AI systems and quantum hardware in real world deployments? The best example I can give you is a publication that we put out just about five months ago, which was Nvidia with inflection for the first time ever solving a material science

application using a combination of our quantum computer with air protected logical qubits, a little bit of jargon, using in coprocessing, Nvidia's GPU technology. So that's a bit abstract. Let me break that down a bit. A lot of the world's hardest materials challenges, things like building a battery for your phone that would last 10 years instead of 10 hours. That's a really exciting application, but we can't study it with the world's biggest GPU clusters because if you really zoom into that problem. It's electrons and a battery interacting with other electrons. A quantum computer is really good at that task. And so I think this is one of the first where QPU plus GPU starts to address something that has a lot of commercial relevance, has a lot of day-to-day relevance.

Can you imagine if your battery would last 10 years? And that was the first domino to fall. It's not yet at a scale where we've found that battery that lasts 10 times longer or outclassed, what a large GPU cluster can do. But the beauty of quantum computing, is that every time we make a little bit of progress, it doubles and quadruples and TEDx is the performance of the quantum computer. So we think that around 2028 in a resisting roadmap, we will be at this inflection point where the combination of NVIDIA's GPU tech plus inflection's QPU tech will tackle these important problems first in material science, but later down the road in areas like AI, pharma, drug discovery, and beyond.

You've also identified contextual machine learning as an early use case in the AI world. Tell me why that one is a target and what kind of practical implications that would have. Sure, glad you asked, because that is squarely in this, how does quantum help with AI topic? And at this conference today, the accelerated compute conference, I've seen many times people are talking about the limits of context, especially as we get to agentic workflows. So as some of the speakers here are talking about, one of the issues on modern Voddrich language models is that they have a pretty aggressive limit on how many tokens they can process at once. There's a context window of about a million tokens, whether it's Google Gemini or Anthropic Cloud or opening a chat chTP. And one of the amazing discoveries about quantum competing is that Q bits, quantum bits, have more context, more memory than traditional bits like in your laptop. And what we found through our advisor, Dr. Eric Anshutz, was a way to parlay that into a context window extension effectively by a quantum computing.

And that's an algorithm that was published just about a year and a half ago, which showed 10x outperformance of Transformers, the TN ChatGPT, using a quantum model. And the craziest thing, which we debuted just a year and a week ago at GTC, was the ability to take that model and actually bring it forward to GPU in a quantum-inspired sense. So this is one of the, I think, most fertile grounds for where quantum is working to accelerate AI, which is that there's algorithms that offer enormous promise when we have large quantum computers, but even before then, we're finding ways to extend the amount of memory that machine learning models process. So we've been working with the US Army, the US-Nabias customers for this technology because they have large sensor data streams that easily pass a million tokens in a millisecond kind of time frame.

And so I think that's where quantum AI are going to first come together. That's amazing. You touched on this, but can you give a real-world example of where a quantum computing model with the same memory as classical would produce a dramatically different result? Sure. I gave the example of sensor data streams, but let me give one in biotech, which is the human genome. So right now, the human genome has been this grand challenge of modern AI. Obviously, if we could understand the human genome better, that would be incredible for human longevity and for just humanitarian applications and whatnot. The challenge is that the human genome is 6 billion-based pairs long, and that's 6,000 times longer than what ChatGPT or Gemini can take.

And it's not just that you can't solve it by just throwing 6,000 the amount of compute at it. Every time you double the context window, you have to 4x the GPU. So right now, existing machine learning models really struggle with long genome sequences. And we were actually able to take our quantum-inspired bottle. There's a predecessor of this that's in our publication, and then the hallmark result from us came out a year ago, where we actually have been able to set a new record on genomic sequence processing using this contextual machine learning technology. Again, it's taking a concept from quantum mechanics of all places, but applying it to the genome.

And the reason this matters is that whether a DNA-based pair in position one is AT or G or C actually has dramatic impacts on downstream biology, 6 billion-based pairs later. Sure. Sure, that makes a lot of sense. Now, when you're talking about this, it all sounds pretty heady, but when you think about the practical implications of near-term real-world value, when is that near-term in your mind? And how do you see that coming together? For applications like processing sensor data, it's already here. So for instance, we have deployed these models to Nvidia Jetson, which is their Edge GPU.

And the U.S. Navy is contracting with us as the U.S. Army to apply that to things like sensor data fusion. So for instance, you can imagine that in places like the Strait of Remodos right now, where they don't have valid GPS signals because GPS is being played around with by adversaries. There's an opportunity to use this type of technology that is edge deployed to instead rely on things like computer vision or even celestial navigation, believe it or not, that looking at the stars and figuring out the map of where I am is a valid way to navigate. So that's a very here now application of this technology. And then just around the corner, we think, is design of better materials, design of chemistry, drug discovery, and whatnot. that is far beyond what we've been able to do

with the world's biggest competing clusters that were pre-quantum. So you've kind of described GPUs and AI systems as essential to real-time quantum error correction. How important is AI in making fault-tolerant quantum computing commercially viable? Good. And maybe I can start for the audience

with what is met by fault-taric quantum computing. We throw it a lot in the quantum world, but it's basically a useful quantum computer. If you don't have fault-tolerance, that means that as you run the quantum computer, it's going to start to succumb to errors. Fault tolerance really means you eclipse that point and can run a computation for a long time. So it's an important point that AI is going to help us achieve fault-tolerant quantum competing. And the main way is GPU can help us solve this inversion problem.

Bit of jargon, but basically it is saying that as the quantum computer runs, it's going to produce some errors because we live in a noise-prone world. and the GPU can go and look at the output of the quantum computer and infer where did the error happen and correct it. And this sounds a bit abstract, but it's actually something that we interact with in our daily lives when we turn on Wi-Fi or turn on 5G, because the packets that we're getting on Wi-Fi or 5G are they have some noise, and ultimately you get a pristine signal on your laptop, your phone. We're using AI to do the exact same thing, which is we're taking noisy quantum bits, qubits, and turning them into a very pristine signal, using AI to detect where did something potentially

go wrong? And if so, how do we most efficiently fix it? So going back to GTC last week, this is why a big part of Nvidia putting their GPUs next to inflection as quantum computers, vice versa, was because they know that their GPUs can help us climb up that curve to fault-tolerant quantum computing much faster and squash those errors so that we can run a computation not just for six microseconds, but six hours or even six days. Now, everything that you're describing makes me feel like we are well on the path to getting to the moment where there's a viable quantum computer. When do you think, I know this is the million dollar question,

but when do you think this moment is going to happen if you look at that next five to 10 year time horizon? And what will we know when we hit it? Sure. I think it's a trillion dollar question arguably. And we think that year 2028, is the moment that we get to 100 reliable quantum bits, which are fault-tolerant logical qubits. And we in turn think that's the point where we get to these real-world applications

that vastly outclass what the biggest supercomputer in the world can do right down. Now, there's still work to get there, but we have an engineering roadmap to get there. And how do we know that moment has occurred? I think one thing we'll see is that we won't even necessarily talk about it being a quantum computer anymore. It'll just be something behind the scenes that customers will say, yeah, of course I'm going to buy this instead of compute PlatformX because this solves my problem much better. It's funny. I think in some ways, GPU has gone through the same thing where we forget now that the gene GPU's graphics, but people who are working on AI, it's a no-brainer for them to pick the GPU. So I think in

sometime around 2028, perhaps a little bit after that when the market saturation really hits in, we'll reach this point where customers will pick the icon that says QPU for the obvious way to run their workload most efficiently, and hopefully even forget that it's quantum. It's at the end of day it's just going to be a way to solve their problems much faster than GP or CPU can. Very cool. Yeah. This is a fascinating conversation. I know our listeners are going to want to know where else to go.

Where should folks go to connect with you and or learn more about inflection? Sure. Our website is inflection.com. That's inflection with a Q because we're a quantum company. And we've got profiles on LinkedIn, X, Twitter, Instagram, I think, too. And I welcome people to follow us there and stay tuned with the exciting things that are going on. And as a computer scientist, I am just geeking out at the level of amazing physics and hardware innovation that is now becoming accessible to developers and programmers. And so I think it's an exciting time for our field.

Fantastic. Thank you so much, Praninom, for coming on the show. And that wraps another episode of Data Insights. It's Janice. Always a pleasure. And what a wonderful conversation today. Thanks so much. Thanks for the new questions. Thanks for joining Tech Arena. Subscribe and engage at our website, Techarena.aI. All content is copyright by Tech Arena.