Let's do that.
Great. So, Michael, we've agreed it's the end of the day and we have to give something to everybody here.
Keep it peppy. I mean, CoreWeave is a peppy story. I was just looking at my numbers. 2022, 16 million in revenues. This year I think the guidance is 13 billion, which by my maths is about a thousand x. So it's been quite a story. Let's start with what you see as CoreWeave. So you're one of the infrastructure companies of the AI world, and so you see what the big AI companies buy. You also have a tenth of the Fortune 500 as customers, I hear, and including companies like Caterpillar. So what do most people, particularly in Wall Street, not see in AI demand? What are they getting wrong about what AI demand looks like?
I don't think they're getting it wrong. I think they are cautious. It's an enormous investment that's moving into a space very, very rapidly. And so, you know, I think there's some level of trepidation, but what we're seeing at CoreWeave is two compounding drivers, right? You're seeing the scale of your existing clients, the demand for infrastructure that they have rising in a parabolic curve, and you have the breadth of different companies that have different use cases compounding that. And that's been one of the trends that I really watch, right? Because that's a function of diffusion of this technology throughout the space, throughout the economy, throughout society, and that is very, very encouraging for us. It gives us a lot of courage in terms of how we go about positioning the company on a go-forward basis.
Fully 50% of our clients this year, the first three quarters of this year, are not AI labs. They're not AI companies. They are representatives from across the economy that are consuming compute because they need that to drive their business to be more effective. And that is an incredibly bullish statistic, just like, you know, having 10% of the Fortune 500 use us as a provider of the compute and the infrastructure and the cloud that they need. I mean, that's really incredibly compelling.
Well, if demand is one side of this AI story, right? The other side of the story is how are we paying for the infrastructure that is required for all of this. And, you know, CoreWeave is part of this, you know, trillion dollar a year and growing infrastructure buildout. We're in Silicon Valley. Silicon Valley has always been an equity story, right? You share in the upside and it goes 10x, 100x, a thousand x. And you've come, you're in the East Coast, and infrastructure is a debt story. And CoreWeave is often seen as a debt story, right? Tens of billions of dollars, maybe more than $40 billion of debt right now. That's sometimes a frightening number, but you don't seem frightened by it. So what is the debt story and how do we make sense of it?
So I thought when you said one side of it is the demand side, we're going to talk about the supply side. But look, when you're thinking about CoreWeave, right? We deconstruct the business in terms of how we've gone about building the business, financing the business, right? So the headlines about CoreWeave are always focused on the infrastructure part of our business, the physical build of this enormous computing cloud that we've built, right? We have 1.7 gigawatts of live power, right? Like there's really just a very small handful of companies in the entire world that bring that much infrastructure to market.
Just what that 1.7 gigawatt number, because I know that CoreWeave was talking about megawatts and gigawatts before the traditional big three hyperscalers were talking about this. But just help us with a sense of what does a gigawatt cost and what does it make, and then you can go back and tell me all about the debt.
Well, when you can look at our revenue, you can tell, you know, that that's sort of what it makes. So, and that's being scaled through the year, right? Like, so it's a run rate kind of. Yeah. And I think that the last guidance we gave on run rate was 17 or 19 million, I forget which number it was, as an exit run rate for the company. But, you know, I mean, moving from megawatts to gigawatts once again is representative of an industry that has changed the way that it thinks about compute by an order of magnitude, right? Or maybe two.
Three orders of magnitude, right? And so it's giving you an idea of how people are going about thinking about the amount of compute that they need for either training or inference. You know, and that doesn't kind of lean into the other services, right? Like, you know, 90% of our clients use more than two of our software solutions, 75% use three or more, right? So like it's not just the compute. And this gets back to what I was saying: when we think about our business, you have an infrastructure component of it which is a depreciating asset, which should be, as far as I'm concerned, financed with debt, right? Like if we think about other infrastructure buildouts that have occurred around the world throughout time, that is the tool that you use because you can match your debt amortization to the useful life of the infrastructure. And that's a very powerful way to think about mobilizing capital to build very large, very expensive things. And it makes a lot of sense, right? If I'm going to live in the house for 30 years, I want to pay for it over 30 years. That's the mental model that you should be using. On the other hand, we have this incredible software layer, which we launched at our Fully Connected conference called Forge. And what Forge is, is really that loop of self-reinforcing software that allows you to build better, faster, larger, more dynamically. And that is a wonderful tool which our clients really plug into and take piecemeal for what they need, or they take the whole thing. And it's an incredibly powerful software suite which more closely resembles the equity-based model that folks have historically thought about within Silicon Valley.
But let's come back though to this question of how do you fund the, how does anyone fund the infrastructure during this buildout? And a lot of the concerns have been that we move from a world where there are a handful of companies funding this through their cash flows. Obviously, CoreWeave was a young company; you didn't have the cash flows of a, you know, 20-year search monopoly to support you, so you've gone a different way. But the market is now looking at a lot more debt being required over the next four or five years in general. So, you've got some experience over the last few years in building this. How do you see that emerging? Is the market already there? Does it professionalize more? Is there enough capital around to pay for all of this?
So look, CoreWeave was the company that was really, we were the tip of the spear in a lot of ways around how you go about financing this infrastructure as a small company trying to build large things. And we created the GPU-based financing tools. And what is important to understand about that is that we took existing financing structures and altered them to the profile of the infrastructure that we were building. So if you're building a dam or power plant, it lasts much longer. If you're building computing infrastructure, the obsolescence curve is shorter. And so you have to alter the structure. And what we talked about on our last earnings call was that the structures that we built have worked exactly as designed. And CoreWeave has paid off over $4 billion worth of the infrastructure that we built. And remember, that's the infrastructure that we built when the company was making, what was it, 16...