Mythos-class Model Claude Fable 5 Early Reviews, How Nasdaq Landed SpaceX's Mega IPO

20:00Stephanie Palazzolo:Right. Great. Well, Michele, I want to thank you for coming on. That is Michele Katassa, president and head of AI at Replit here on TI TV. OpenAI is making another big effort to secure data center capacity. And my colleague Anise Gardizi, our cloud and compute reporter, published exclusive reporting on a major deal the company is closing in on. I want to bring on Anissa to share with us more about what she learned. Anissa, welcome back to the show. What do we know about this new data center deal?

20:31Akash Pasricha:OpenAI is in late stage negotiations on leasing a 10 gigawatt data center in Ohio. Now, it's very important to know that this data center hasn't been built yet, so it's still an early stage project. But OpenAI is in talks to bid on the site. And if it were to win the bid and this data center were to be built out, it would 100 % be OpenAI's largest AI data center facility. This location is in Southern Ohio, and it's near a power plant, a new power plant that's being built. So yeah, that's what we reported yesterday.

21:06Stephanie Palazzolo:So let's unpack it. So 10 gigawatts, that's Stargate, right? I mean, Stargate was, the announcement said that it was going to be a joint effort, right? It was Masayoshi Son was there, Trump was there, Larry Ellison was there, Sam Altman was there. That was going to be a big joint effort. This is opening eyes saying, we'll just do it on our own and we'll lease it independently.

21:32Akash Pasricha:Essentially, I would say the original Stargate that you just mentioned with Oracle and SoftBank has kind of still gone on. So OpenAI is doing around six gigawatts with Oracle under that 10 gigawatt umbrella. And then they've counted another small project with SoftBank Energy, getting them kind of close to that 10 gigawatts. So this is a completely net new 10 gigawatts. And yeah, you're right.

21:54Stephanie Palazzolo:In addition to that. Okay.

21:56Akash Pasricha:And in this case, it is OpenAI leasing a data center, which, you know, if they actually end up going through with it will be as close as they've come so far to sort of controlling more of their own infrastructure. So yeah, they essentially are, this is Stargate maybe 2.0. Okay.

22:12Stephanie Palazzolo:And so now in your story, you outlined that this effort, I mean, it's a$500 billion effort to operate this much compute or construct it. Help us understand what that figure refers to. That's not the amount that OpenAI will be paying itself, right?

22:31Akash Pasricha:Yeah, there are really big numbers here. So I can sort of walk you through what this all means. So$500 billion is the number on the total cost of building out 10 gigawatts of data center capacity based on today's prices. And so OpenAI isn't going to be building the data center. That is all on SoftBank Energy. And we've reported that they're going to have to potentially, you know, borrow project financing in order to build out this 10 gigawatt site for OpenAI. Now, on the OpenAI side, they're going to be on this 20-year lease if the deal goes through. So we estimate that would be in tens of billions of dollars over the 20-year period just to lease this data center capacity.

23:15Akash Pasricha:And then the other thing that OpenAI would be unhooked for is actually the largest line item here, which is the chips that go inside of the data center. And so based on that$500 billion number, the amount of hardware and other types of chip gear that you might have to buy, networking, that would be around$350 billion. And so these are just astronomical figures. And we expect that if OpenAI progressed to that stage, that they would likely seek a financing partner to help them fill this data center with chips. And it'll obviously get built out in phases as well. Right.

23:49Stephanie Palazzolo:Right. And so now this is where NVIDIA comes into the picture, right? Because you reported that NVIDIA is putting itself on the line to be a financial backstop of sorts for this leasing arrangement. So what would NVIDIA conceivably be on the hook for? When would they need to give money?

24:11Akash Pasricha:Yes. So the NVIDIA angle here is that they are in talks to be what is known in the data center community as like a financial backstop. And so it doesn't, I don't think, you know, a lot of these backstops happen because the party trying to access capacity, in this case, OpenAI, is not investment grade. So for example, it'd be very hard for SoftBank Energy to get people to lend them money if they say that their tenant is OpenAI. You know, as much as people might like OpenAI and think that this data center is going to be successful, lenders don't really care. They just care about whether the tenant is investment grade.

24:47Stephanie Palazzolo:So what it means, meaning if it's a public company where bonds can sort of be freely traded?

24:55Akash Pasricha:Yes, and has a credit rating. And that's something that OpenAI as a private company doesn't have. Anthropoc doesn't have it. They're in the same boat. And so what NVIDIA is proposing to do is say, you know what, if we will backstop OpenAI's lease payments. And that is something that makes lenders feel a lot more comfortable in lending money for a project. So that's NVIDIA's role. And obviously, the data center would be filled with NVIDIA chips. So this could potentially be one of NVIDIA's largest data center sites as well for its chips.

25:24Stephanie Palazzolo:Okay. And just to clarify one other detail here. So you mentioned that the OpenAI's lease payments, we don't know the figure. It'll probably come to the tune of tens of billions of dollars. The$350 billion-ish that it would take to secure the chips that you said, who would be paying that money? Would that be SoftBank Energy to putting the chips in the data centers?

25:52Akash Pasricha:No, as we understand the deal talks today, it would be OpenAI doing the chips. And so, yeah, the question is, or how are they gonna be able to pay for that? And this is definitely something that I think they're discussing, but is a little bit down the line in terms of when they actually have to figure that out. But one thing that could be plausible is that NVIDIA does a vendor financing with OpenAI. Obviously, that'd be a very large vendor financing. But yeah, I wouldn't be surprised if maybe that's what we see later down the line. The first phase of the data center is expected to be done in 2028.

26:24Akash Pasricha:So there's a little bit of time for OpenAI to figure out how to get that 350.

26:29Stephanie Palazzolo:Right. Now, we should say SoftBank Energy, SB Energy, as you refer to them in the story, they are the lead developer on this Ohio plot of land, sounds like. SoftBank is, of course, a big investor in OpenAI. Do we have any sense for how that relationship maybe could make things easier for OpenAI? Are these going to be, I don't know, favorable rates on the lease or maybe... I don't know the answer. I'm just thinking, everything is so interconnected here, right?

27:09Akash Pasricha:Yeah. What we learned about this specific site is that we haven't even talked about the connection to the Department of Energy and the Department of Commerce and how the U.S. government is kind of playing a role in getting this data center off the ground. And so this new power plant that's going to be built for the data center is being built on Department of Energy land. And one thing that the Department of Energy asked SoftBank Energy was to run a fair process to find bidders for the data center. And so as far as we know, multiple companies looked at this site. And if OpenAI is the one in advanced discussions, maybe they offered a higher price than other people.

27:48Akash Pasricha:We don't exactly know. But I think it was, from what I heard, the government was interested in not just saying, you know, because SoftBank Energy, which has a relation to SoftBank, is doing this site, it's 100 % going to go to OpenAI.

28:04Stephanie Palazzolo:Right. Yeah, well, I mean, and that's, you know, that's a key aspect to this too, because in this world of circular financing, you could imagine a scenario where if the government wasn't evolved, you know, maybe they make things easier. But that's a great point. Now, I just want to ask you about the capacity number here, 10 gigawatts. I mean, we know that OpenAI is obviously a fast-growing company. 10 gigawatts is a lot of capacity. Did you get any understanding from people around, does OpenAI need all of this today? Is this capacity that they're building for the future? Just put the 10 gigawatts into context for us.

28:41Akash Pasricha:This is definitely future capacity for OpenAI. You know, the first phase of this project in Ohio isn't set to be ready until the 2028 timeframe. So in today's world, you know, the companies might view that as a near-term compute, but the full 10 gigawatts that we're talking about here is likely many years away. So this is definitely a bet on opening eyes part that they need to get the ball rolling on a large site like this early. Because, you know, if it comes to be 2032, and they don't have enough compute, then they at least have this data center where they have a 20 year lease and a lot of flexibility around compute.

29:15Akash Pasricha:And then just to put the 10 gigawatts into context, this is the largest data center under development in the US. And, you know, I think that also means that we should take it with a dose of skepticism because a lot of opening eyes, big compute announcements don't pan out. So even though they're talking about this and even might sign the lease, we'll definitely have to see, you know, do lenders buy in and is it able to reach that 10 gigawatts? Or maybe it ends up somewhere around four or five, which is still so large. But there's a lot up in the air here, even though they are nearing this initial deal.

29:50Great.

29:51Stephanie Palazzolo:Well, Anissa, I want to thank you for coming on. That is Anissa Cardizzi, our cloud and compute reporter here at The Information. SpaceX's IPO will have ripple effects on the broader stock market, and it also has implications for the companies who are in the business of operating the stock market. To that end, it is noteworthy that SpaceX picked the NASDAQ over the New York Stock Exchange to list its shares my colleague cory weinberg published an in-depth look at changes nasdaq has been making within its own business to help secure listings like these i want to bring him on to talk more of all about it cory welcome back to the show it's great to have you here hey cory why did spacex pick the nasdaq over the new york stock exchange Well, NASDAQ happens to be the only stock exchange that has basically the power to drive billions of dollars worth of stock in SpaceX's direction.

30:53It's the only stock exchange that is directly tied to an index that controls a ton of passive money. And so while we don't know exactly the decision-making process behind SpaceX and why they chose NASDAQ, they had a very strong incentive to do so if they want their stock to move higher.

31:14Stephanie Palazzolo:Okay. We do know, though, that the NASDAQ has been making some changes, though, to how it includes companies in its indexes, right, which could have had an impact here. yeah nasdaq was the most aggressive and among the first of these wave of index providers to essentially say hey we're going to make some changes to which companies we allow in the indices to make way for these mega ipos that are in the pipeline this year the first of which being SpaceX. So a bunch of index providers, including sort of MSCI, Russell, CRISP, which is owned by Morningstar, and NASDAQ, all made changes to essentially make it a little bit easier for companies like these mega caps that are going public to enter the indexes faster than they would have otherwise.

32:08Usually there's a seasoning period. Usually there's some other requirements that companies need to have before they can be automatically bought by funds that track the index. NADAC was among the most aggressive. It had this unique feature where it will weight SpaceX in the index at a disproportionate amount compared to how many shares are actually trading. So that should give it a sort of disproportionate weight in its index. And it also happens to be the one where SpaceX is actually listing its stock.

32:39Stephanie Palazzolo:And what is the reason that they give you or what's the public reason they say they are making these changes? Does it have to do with companies staying private longer or what's the rationale on their end? That's one of the stated reasons, yeah. I mean, I may have a point. Look, I mean, the index providers are responding to an environment where you have for the first time companies coming public with trillion dollar mark you know plus market caps like we've never seen this before the structural changes that they're talking about in terms of hey we need to make um some adjustments to how we usually keep you know sort of include companies in the index um that's real uh for sure critics would say um look that doesn't change the fact that a lot of these companies are still incredibly risky bets for investors to make um that they're losing tons of money that their future prospects have some hair on them and that by putting uh sort of them in the index faster you are essentially putting investors at risk and this decision to accelerate uh how fast spacex could be included in the indexes i mean that decision in and of itself uh you had some interesting reporting on how people are feeling about that it's not not everyone is agreed that it's a good thing, right?

34:03Yeah. I mean, it's a controversial decision. We essentially worked with a company called Peak Metrics to measure sentiment around the online conversation surrounding the SpaceX IPO. And one of the most negative themes was this notion that the index providers are forcing SpaceX stock down investors' throats, essentially. That was an insanely negative theme online, on Reddit, on X itself. And then folks that we talked to in the index industry, including people who've actually worked with companies to essentially pick an exchange, have said, like, this seems like a bridge too far. This seems like the index, especially NASDAQ, because they want to win the listing are making some big changes to please one or a handful of companies.

34:56And that threatens the market structure overall.

35:01Stephanie Palazzolo:Now, I understood this idea that you just explained that there are funds that track the NASDAQ 100 index. And so if SpaceX is included in NASDAQ 100, then the funds that track that will have to purchase stock to sort of balance things accordingly. And that will obviously be good news for SpaceX because it means that they will have more buyers of their shares. The other part of the story I want to ask you about, though, is you mentioned that Adina Friedman, the CEO of Nasdaq, said essentially Nasdaq has become a 3 % to a 4 % owner of companies that sit in the Nasdaq 100. How is it that Nasdaq has become such a big shareholder, I guess, in these companies?

35:42They are literally a shareholder. Essentially, what Adina Friedman, who's the longtime CEO of NASDAQ, is referring to there is essentially referencing, I think, sort of the influence that NASDAQ wields over the funds that track the index.

36:00Stephanie Palazzolo:Got it. So by proxy, I mean, based on what they do, those funds own that much. Correct. NASDAQ is not an owner. Right. Okay. So let's just be really clear there. But they are indirectly or I mean, they are directly influential on sort of how those funds actually where those funds actually go, because the weight that NAPDAC determines will determine where, you know, really influential ETFs like Invesco's QQQ series, sort of how they allocate funds. And that interview that Adina Freeman gave in March, which I found really striking, it was the first time I had seen sort of her or really any NASDAQ officials talk openly about the relationship between the index and the exchange.

36:54that essentially, as one person put it to me, this index is essentially a marketing weapon for NASDAQ to be able to win more listings business, especially large listings, which they have been doing recently.

37:08Stephanie Palazzolo:Right. Now, the S &P 500 decided not to go the way of NASDAQ and accelerate this process of getting it included in its index. what did that tell us about about this IPO but the decision about how the future could look you know will these two indexes take drastically different directions will they look different I mean what are the implications there I mean yeah they're they're gonna look pretty different I think especially if NASDAQ 100 has you know if Anthropic and OpenAI decide to list there as well you're going to have a pretty stark difference in sort of how the NASDAQ 100 performs compared to the S &P 500.

37:57Like these are going to be volatile stocks and everyone has an opinion on them. I'm sitting here today as you know someone who's talking to people involved in the SpaceX IPO all time no one really knows how it's going to trade right like like this is a polarizing stock with a pretty low float um and if it's not in the s &p 500 and neither are open a on ion anthropic um you're going to be see some pretty stark differences in performance there for better or for worse i think the s &p has historically been seen as a little bit more conservative um obviously they have a uh a much broader base of funds that track them uh nasdaq 100 is heavily you know heavily in vesco qqq so they have essentially uh they don't have one master to please but they they you know sort of are much more disproportionate to to kind of one customer there um s &p is much broader and s &p doesn't have an exchange business they they aren't trying to win the listing for these companies.

39:04They're kind of a broader data intelligence company that is, you know, obviously have had its own history with trust and trying to ensure stability in markets. And they decided to go a much more conservative route than NASDAQ. Right.

39:20Stephanie Palazzolo:Well, we can thank Elon for certainly muddying the waters here between all of these businesses and market dynamics. Corey, I want to thank you for coming on. That is Corey Weinberg, our senior reporter, here at The Information. Our next segment is with our partner, Nebius. NVIDIA says its Verorubin chips are in full production. Nebius will eventually deploy those chips in its data centers. I had the chance to talk with Nebius' chief revenue officer, Mark Boroditsky, while he was at Nebius' AI conference in Flexion. We spoke about Verorubin demand, that he's seeing lessons from the Blackwell rollout and how CPUs are coming up in his conversations with customers.

40:03Stephanie Palazzolo:Here is that conversation. Mark, welcome back to the show. It's great to have you here. Akash, it's a pleasure. I'm glad to be back. So you are at Nebius' big event. You've got customers on the ground with you. You've got investors, analysts, everybody in the Nebius orbit. The topic that I want to start off today with is Verorubit. They are the next family of NVIDIA chips. Everyone is looking ahead to see how much demand they get. You have a front seat to how much demand these chips are ultimately going to get. What are the big questions that your customers are asking you about that family of chips?

40:43Well, first of all, they all want to get access. There's a desire to get their hands on them. People are asking about test racks. People want to start to take their existing models or their existing software and see what happens when they put on Vero's. Second of all, everybody wants to know when can they actually start to get access to buy. They're recognizing that the next evolution of whatever they're building, they're going to need to be prepared with Vero's. And that means setting themselves up to, you know, have the capital or necessary financing to be able to get that next set of chips and remain competitive or take on the additional capacity they need to continue to build their business.

41:27Stephanie Palazzolo:And so as they look ahead to the next family of chips, I mean, one question I've thought about is, as these chips get better, how much demand will there remain for the older generations of chips, the Blackwells, the Hopper family, the A100s? I mean, you know, I imagine that the chips just need to hit a certain threshold and then they are able to do, you know, enough for customers. So are you getting inquiries saying, hey, we don't need the Vera Rubens. We're actually okay with the Black Welter or the Hopper. Can you give us any of those right now? Really depends on the customer and the kind of workload that they have.

42:03And you're completely correct in the way you described it. There's still a large number of customers that are very satisfied with the older generation of chips. As a matter of fact, there's a number that say, hey, any hoppers come available? Here's my number. And you can actually see it in the pricing dynamics. Hopper prices aren't dropping. And in some cases, they've been going up. So it's interesting. You look at these customers and you wonder, are they going to evolve? And some of them aren't. They're fine with the capabilities they have from the chips. Others, they're actually running multiple types of chips for different workloads that they have within their requirements.

42:37So they hold on to the previous generation as they move on to the next generation. They use it for other things than maybe the most advanced model making or the highest scale inferencing.

42:49Stephanie Palazzolo:Other than price, what are the other levers that customers will try to talk to you about when it comes to the negotiations and their considerations for which chip to use? Well, I mean, the sort of rudimentary commercial terms are price, duration of the contract, and the amount of cash up front. So those are factors that we consider when we're looking at the transactions. We also look at the kind of customer. If the customer is a startup that's just raised their first seed round, they represent a different degree of risk than, say, a AAA credit enterprise. So the demands that we may have in terms of terms might vary slightly based on the kind of customer that's there.

43:37Stephanie Palazzolo:Right. And bringing it back then to Vera Rubens, I mean, we reported the information on some of the challenges that the Blackwell had in terms of implementation. And we should note that, and we noted this in the story, I mean, this is a scale of rollout that has never been seen before. And so in some ways, it'd be crazy to think that there wouldn't be challenges with implementation. But I wonder how you are thinking about the lessons from the Blackwell implementation that you are then applying to the Vera Rubin rollout. Well, first of all, the reality is that every chip generation has some degree of challenges associated with the implementation.

44:15And it differs based on each of the generations. What we're understanding, and this is preliminary, is that the architectural shift is not as dramatic as the Blackwell shift was. So our hopes, and I know hopes are not a plan, but our hopes are that we're not going to experience the same degree of challenges. But that said, we're making sure that we're getting our most important customers early access so they can touch them and have experience with them. Likewise, if that's not possible, then we're talking to customers about POCs once we actually get production chips available for customers. And it's likely that many customers that have had, say, a previous experience, they're going to be a little more cautious.

45:00They might be asking for, you know, ultimately test validation before they implement for their commercial term.

45:08Stephanie Palazzolo:You know, the other class of chips I wanted to ask you about are not GPUs, but CPUs. So we are in the era of a memory chip shortage. And I wonder how that has affected your conversations with customers and what they are asking of you. Well, so memory chips, different than CPUs, but I'll address that. My apologies. Yeah. Memory chip. Well, let's stick with CPUs and then I want to get to memory chips afterwards. Awesome. I'm happy to answer about both. It's interesting, in the last couple of quarters, we've seen a pretty significant increase in requests for more CPUs, especially with the reinforcement learning and large-scale inferencing and some of the agentic requirements that people have.

46:00Having CPUs near the GPUs is critical in order to be able to service the workload. And in some cases, the requirements are a high multiple of the GPUs. As a matter of fact, NVIDIA has Vira CPUs coming out. And so we're looking forward to seeing how those ultimately perform in the market. The obvious benefit being that the bundling with the Vira Rubens will be an interesting combination to potentially address some of these workloads.

46:35Stephanie Palazzolo:Who is your dominant CPU vendor right now, though? I believe we're mostly AMDs, but I can't say for certain. My apologies. I should know better. No, no, but I mean, this was the question is, you know, as the agentic workloads increase, and we're seeing that CPUs are becoming much more important to that story. and given that NVIDIA has their own CPU offering, do you see them dominating more share, I guess, in the CPU space, even in your own workloads? We're already looking forward to the Bureau of CPUs and having conversations with NVIDIA about getting our hands on them for test purposes and getting them to customers as well.

47:19There's going to be obvious benefit to NVIDIA's bundled capabilities. So we're looking forward to seeing if there's going to be a significant performance improvement or some kind of overall TCO benefit that our customers can rely upon.

47:36Stephanie Palazzolo:Right. Okay. And now let's go to memory chips. How is the memory chip shortage affecting your discussions? Well, it's obviously an impact on COGS. You know, cost of goods sold. So prices are going up and, you know, we are reflecting that in the pricing that we're bringing to market. And there's also, at the same time, a desire to get more memory. This is not too different than the CPU conversation we just had. Customers are loading more data that they want to have near adjacent or adjacent to the GPUs to be able to take on these more and larger and larger parameter models or more and more agentic workloads.

48:14So the price is going up and the demand for more is taking place at the same time. So ultimately, cost of goods will affect price. Price will move. I'd say, you know, if I had to be really honest, the pricing dynamic is more heavily driven by supply and demand right now. Just the fact that there's so much more demand than supply. Right. We're probably increasing prices more because of that than just the cost of goods sold.

48:45Stephanie Palazzolo:And so to that point, I mean, the most recent quarter that Nubius reported 684 percent revenue growth. It is certainly a rocket ship and a good time to be the CRO of the company. But I think everyone is understanding that this is a moment. And, you know, you can't always grow 684 percent every quarter, although I'm sure it's what – you hope for. Take me inside sort of your playbook here. How are you structuring your team and your offering to account for the fact that it's not always gonna be 684%, but we have to keep our offering sticky in some cases? Well, so let's start with the offering and then I'll talk about the go-to-market.

49:29On the offering side, we've been expanding our capabilities so that we're delivering more and more of the functionality AI builders need. You know, the earlier experiences that we've had in the market, we've been able to win the model builders with the tools that we provide them for model creation and model fine tuning. We've augmented and added to that our token factory inferencing solution. So as people put models into production and they want to commercialize them, they can use token factory to do that. So this is the software around the office. This is not just data center space. It's software.

50:03It's software. It's software that ultimately relies on data center capabilities. But the point being that as you look at the market, we want to be diversifying into more workloads per customer. This is cross-selling at the end of the day. So when a model builder is taking their model to market, we want to be able to take on and help them with their commercial requirements. And you can think of it, the evolution of the market has gone from models to inferencing, and next is agentic. And so we're building out a full set of capabilities to deal with each of those. You know, most recently in the agentic space, we acquired Tavali, which is an agentic search solution that is very popular with developers that are needing to have that API access to the internet for search and access to data APIs and other agents.

50:50Right. And then what about the go-to-market team side of the equation? A hundred percent. Thank you for bringing that back up. So we've been expanding our coverage of the market. Today, we have now teams in the Americas and EMEA, and we just stood up a team this past quarter in APJ, in Asia Pacific and Japan. So reaching more of the geographic market. We're also remaining extremely committed to the AI native portion of the market. That's been a very successful market segment for us and one that we have a right to win. But we're also making investments in the rest of the market where we believe as the market may cool down in the foreseeable future, we think that software vendors and enterprises are going to represent the lion's share of opportunity.

51:42We're making investments now so that we can meet that demand as it matures. So building a go-to-market engine to be able to know who the main enterprise customer opportunities are, being able to target specific workloads, and win the lighthouse accounts, establish credibility, and establish trust in the category so that as the market starts to heat up, we'll be able to continue to grow at a better-than-market growth rate. Maybe not 684 % like you suggested, but I'm going to continue to aim for it.

52:18Stephanie Palazzolo:Great. Well, I look forward to speaking with you more as we get more and more quarterly results in the months to come. Mark, always a pleasure. Thank you for coming on. That is Mark Boroditsky, the CRO at Nebius here on TITV. That does it for today's show. A reminder, we are on this stream Monday through Friday at 10 a.m. Pacific, 1 p.m. Eastern. If you can't make it then, episodes are available on theinformation.com, on our YouTube channel, or wherever you get your podcasts. Make sure to follow us on social media on X, Instagram, and TikTok, and LinkedIn. I'm already excited for our next show tomorrow, but I am even more excited for the New York Knicks to take, hopefully, a 3-1 lead tonight.

53:01Stephanie Palazzolo:They are in New York City. Fingers crossed. Have a great rest of your Wednesday. We'll see you tomorrow. Bye-bye for now.

53:11Thank you.