Data processing unit

So before we talk about artificial intelligence, AI, remember that AI is physical before it becomes digital.

So you may hear people talk about terms like CPU, GPU, NPUs, and DPUs.

Well, write those down, right? If you have those written down, they're going to become part of your verbiage as you make those decisions because you don't have to memorize what every letter means.

What I want you to understand is this concept that different types of hardwares are becoming specialized for different types of work.

Then GPUs became extremely important because certain AI workload requires different kinds of processing.

So now you are increasingly hearing about NPUs, specializing AI processors appearing inside of laptops and phones and other devices.

You may also hear about DPUs, particularly when conversation becomes more focused about data centers, networking, and infrastructure.

You don't need engineers for any of them.

it comes to technologies in the quantum space in fact we believe as well as others in the industry believe that you will not have a singular quantum technology modality that is going to be successful in the future you will have multiple modalities that will have different use cases that you can deploy on those modalities and it's no different than in the classical space.

You have CPUs and GPUs and TPUs and DPUs and all kinds of XPUs in the classical space, all doing different things.

We believe it's going to be the same thing on the quantum side.

So your point

He's got, you know, Ethernet switches.

He has two kinds of DPUs.

You know, he's got...

You know, we've gone from just, um, scale out networking being a thing.

So, you know, you think about Nvidia, they've got their AI factories that they build.

You know, uh, that's what, what Jensen likes to call them, and that really consists of CPUs, it consists of DPUs, which is, you know, networking, uh, gear, includes switching, it includes rack scale architectures, their accelerators, and their GPUs.

So these are complete systems which are designed to work together.

And actually, if you look across the landscape, what you're seeing is that AWS is doing the same thing with its Trainium products.

Mm.

with those players right across the board.

Um, whether it's the AMD DPU or the CPUs and DPUs in Nvidia, whether it's the DPUs and acce- uh, CPUs that are part of the Graviton system, the CPU that we've built in partnership with Meta.

In every case, you know, we're sort of foundational in helping to bring those systems together and, and sort of meeting them where they, where they need us to be in order to kinda make those systems a reality.

We've been working with VAST to make sure that those technologies bolt well together with VAST storage technologies.

And then we've been building that on top of technologies on the networking side, like RDMA, where we can avoid the overhead of running TCP IP stacks, and we can offload the TCP IP stacks to things like DPUs.

or advanced like NIC technology, and then also using technologies like NVMe for the backend storage where we get that advantage of low latency protocols talking to highly performant non-volatile memory technologies in order to make sure that that objective of low latency with low CPU overhead is met.

I think when we think about networking, there's a few aspects there that are critical.