00:00How difficult has it been to keep up with demand in the last 12 months?
00:04So it's not the last 12 months. It really has been a sustained issue across the space.
00:11So not just for us, but for really all of the companies that are trying to serve infrastructure to AI models that are being used by enterprise, that are being used by governments, that are being used by retail.
00:24It's really been a relentless and voracious requirement to continue to build infrastructure as fast as you can.
00:34What are the big constraints at the moment?
00:36So the constraints are really around our ability to bring up powered shells fast enough to be able to kind of manage this incredible spike in demand.
00:49You know, over time, you know, as as the markets do what markets do, which will be to to to address the the infrastructure issues, the constraint will move to power.
01:01The constraint will be from fabs in their ability to to generate more chips and memory.
01:08And, you know, it's so it moves around. But right now, the single most difficult issue is really to focus on the powered shells.
01:16There's some skepticism around the industry just at large, but in particular around your company.
01:22There was a recent lawsuit having to do with overrepresenting capacity on the part of Corweave ahead of the IPO last year.
01:31What's your response to that type of skepticism and that claim in particular?
01:35The company will vigorously fight that claim.
01:40We believe that it's baseless and, you know, there there's an enormous amount of information out there about the company, about demand and about our ability to build it and when we're going to build it.
01:51And so, you know, these things happen, but we're not terribly concerned about that.
01:55Your IPO is in March. Your shares are up 130 percent.
01:58It's not a terrible performance.
01:59I will say there is a question about whether there are constraints building in financing markets.
02:05When you had that two billion dollar debt sale, there was a sell off in shares, a question around how well you could finance this in an efficient way.
02:13How are you dealing with that? Does it challenge your ability to use the debt markets going forward?
02:17So there was a brief sell off which actually recovered through the rest of the day.
02:21So, you know, I think that there is a significant amount of appetite in the markets for products that can be underwritten, where the offtakes are from creditworthy entities, where you have a company like CoreWeave that has built an incredible reputation around our software
02:43and about our ability to integrate, you know, NVIDIA's GPUs, which are the best silicon solution in market, with the best software solution in market to deliver the best product in market.
02:56And, you know, the capital markets have been very, very receptive to that and will continue to be receptive to that.
03:02As long as the products are structured correctly and the counterparties are strong, we don't see any lack of demand in the capital markets to build the infrastructure we need.
03:11As we've been talking all morning, there's really this push-pull at Davos when it comes to the politics and geopolitics and AI.
03:17The president is going to talk about affordability, and you recently had governors from blue states and red states at the White House basically telling tech companies they need to subsidize American consumers.
03:27When you see these hyperscalers, how do they make sure that they are maintaining the grid, not just for themselves, but keeping prices lower for Americans?
03:35Yeah, so look, the politicians, they're balancing many priorities as they're thinking about artificial intelligence, as they're thinking about affordability, as they're thinking about geopolitics.
03:49All of those things are coming to bear on how they think about and how they frame this discussion.
03:53Around the affordability component in particular, we believe and we have aggressively taken the position that it is important that companies that are building AI pay their fair share of the requirements to deliver power to these data centers.
04:13It's going to be important in the short term and it's going to be important in the long term as artificial intelligence continues to permeate the economy.
04:22And, you know, there's a great opportunity for us to really kind of build out a system to be able to ensure that we're playing that role as we continue to scale this business, which is so important for so many different aspects of society.
04:36Just quickly, and I understand it's a complex question, so forgive me for the time constraints, but how do we value the lifespan of a GPU?
04:44It feels like that's like the key question right now in financial markets and you're at the heart of it.
04:49Yeah, so look, I'm pretty aggressive about this, right?
04:54Like the way that I conceptualize how to value the depreciation curve or the obsolescence curve of technology is a function of our clients and what they are willing to pay for.
05:08There is no argument that will get me to concede that a GPU becomes obsolete when I have clients who are the smartest, most experienced users of this compute that come in and buy compute from us for six years.
05:22They know new technology is coming.
05:24So the idea that this technology becomes obsolete or useless after three years or two years, it's bunk.
05:31It's a narrative that's been put out there by people who are trying to influence the discussion around the stock.
05:37But the truth of the matter is, you know, if Microsoft and Meta and NVIDIA are buying compute for long term contracts, they know new technology is coming and they're going to buy it anyway because they know it's got value.
Comments