So, hi. This is Ray Wong with Michael here at the CoreWeave conference, Fully Connected 2026. Can I ask you some questions? Welcome.
Thank you so much, Ray. It's great to be here.
So, look, scaling is hard. You scaled faster than any other infrastructure company in history. There's financing, there's getting the chips, there's the supply chain. Like, what can you tell other CEOs about getting to that scaling? And more importantly, with AI, it's not just about cutting costs, it's about having more at bats, more revs. What do you do to get to decision velocity?
So, look, you're exactly right. When we founded this business, we understood that in order to be of relevance in this AI revolution, you need to be able to operate at scale. Scale that's able to serve the largest consumers of this infrastructure in the world. And that's everything from training to inference and everything in between. In order to do that, you've got to really build your organization from the ground up. You've got to build your software, you've got to build your cloud from the ground up in order to be able to deliver scale. And that's everything from a software stack that is built in every step of the way with one thought in mind, which is, how do I deliver the best experience possible at a scale that is of relevance during this transition that's taking place.
Yeah, and you're doing security by design and security at scale, which is not easy as well. I mean, this is because everyone's counting on you at that infrastructure level.
Yeah, look, you're exactly right. When you're delivering AI infrastructure and you're delivering the software stack and you're delivering the cloud, you don't get to deliver part. It's got to be all of the pieces brought together and delivered in a comprehensive, functioning package that allows for people to be able to serve their models, to serve their companies, to be able to deliver their product. All of those things require 100% of the pieces to come together. You can't skimp or have any missing part of the puzzle.
No, that's true. And on one hand, you got hyperscalers spending billions, even trillions of dollars on infrastructure, depending on how you calculate that, building really generalized clouds. And then on the other hand, you've got you building really specialized clouds. Why the difference? What are people looking for?
Yeah, so one of the things that we thought about when we founded the company is that the size of compute that was going to be required to enable artificial intelligence, the software rails that were going to allow a couple of founders to build a company that would go from an unknown to one of the largest consumers of compute in a matter of months. All of these things were really foundational in how we thought about it. It's just a different kind of compute delivered through a different technology with different hardware. And in order to be effective, we really started from the concept that it wasn't going to be a generalized cloud. The analogy we use internally is that a generalized cloud was this wonderful minivan, right? Like it could do everything. But in order to be able to serve this one specific use case, you had to build an F1. You had to build something that was super focused to serve a specific purpose because of the scale of what was going to occur.
Well, it's not just that. I mean, to even know that there would be multiple startups going after the same thing, not just big behemoths and a bunch of infrastructure players that were going to remain. I mean, you really thought through that there would be anyone like one individual, a tiny team could go out and kick butt.
Yeah, look, you know, the market's a wonderful thing, right? You identify an opportunity and creative, thoughtful people start up companies to try and go after it. We saw it first. We kind of established size and scale of our operation, but there's nothing about it that's shocking to see other people moving into space. I mean, it makes perfect sense.
Well, we're here at Fully Connected 2026. You're about to launch Forge. What are you most excited about this launch?
Yeah, look, Forge is an important step for us because what it does is it brings together so many different pieces of the ecosystem that we're building, right? And so, when you're thinking about what CoreWeave was built to do, it was built to deliver a cloud for artificial intelligence. And in doing that, we needed to build out all of the different pieces that the most sophisticated consumers of artificial intelligence infrastructure are going to need, but also the enterprise as it broadens, as it deepens. And the Forge product really allows us to address that holistically, to be able to deliver our product across the board, to meet our clients where they are, to ensure that they're going to get the tools that they need to be able to build what they need for their companies, for their institutions.
You know, to anticipate that is a little bit of genius, it's a little bit of luck, it's a little bit of fitting all these things in the right place. So, congratulations there.
Thank you. I have such an incredibly talented team of people that have put so much thought into what we're building and why we're building it. And the ability to come to this conference and kind of exhibit how all of the different disparate pieces come together to serve our clients is just wonderful.
Well, it's interesting, right? You have a precision in your order of operations of what you prioritize, how you release. I think people do see that, they do fully appreciate that. You know, one of the hottest topics this summer has been really token economics, cost of tokens. We've seen that they're driving down in cost, which is interesting. But, over time, do you see this trend actually creating the same amount of demand? Are we going to see Jevons paradox in place? And more importantly, I mean, what's your GPU hour cost? Are they going to continue to hold up?
Yeah, so this has been a hotly contested topic.
To put this on Kelsey over under.
