원문 보기
Yesterday and today's action is what it was like working and trading the financial collapse except that lasted for 2 years. Um we we we lost the midterm rally in 6 minutes and you know trillions of dollars have have been vanished. Like is this the beginning of something or is it you know on Tuesday everything will be good again?
I Well, I don't know if you guys got the notification, but AI got officially cancelled. It's not going to happen ever again. Putting the gene back in the bottle. Clearly, that's what the market's telling us right now. But I mean, it's Hey, look, we can it's I was going to say it's just it's tricky times.
I think right now um nothing about the AI buildout has changed. The only thing has adjusted is the range of outcomes has widened and because of that it's causing a lot of people just to get a bit more defensive. Uh but I'm part of the camp where I just think this is kind of a narrative follows price action a bit and I'm a net buyer today.
I was a net buyer yesterday unfortunately as well. Uh I'm crying with my ads right now. But I am still confident this buildout's going to happen. It's going to happen here or Malaysia. Google is a perfect example of that this week. Like first morning after the midterms, what did Google announce? We're going to invest $15 billion in data centers in Malaysia.
So this this this is going to happen regardless. Demand's not going anywhere geographically. It's going to change, but regarding AI, it's not going anywhere. Yeah, I agree with Shay. I think this has been brewing for quite some time as there's so many investors who have been poking at this perceived AI bubble, almost hoping it's going to burst.
And I've seen some of the narratives this morning suggesting that now the AI bubble is bursting. But I still contend we're just in the very early innings, maybe even batting practice of a double header head as it relates to the AI revolution. So this type of volatility shouldn't necessarily come as a surprise to anybody.
Perhaps the extent of it in one day following the midterm election bounce comes as a surprise. But there's going to be more choppiness ahead. Whether it's them poking fun at how much the hyperscalers are spending or how little they're spending in some cases, chip demand, backlog and memory demand, the data center push back in terms of that political narrative, the lack of power, the lack of water.
pick whatever number you want on that bingo card, but if you go to the sidelines now and try and time this market and abandon the AI revolution, I think you're going to regret it in the long term. I agree with Kevin. I I completely agree. The buildout is happening whether people want to fully appreciate in the market or not.
I mean, the amount of spend that's just happened just this year in 2026 has been massive and the spend projected for next year is even greater than that into the trillions. So, I mean, if you look at just that momentum and just the price action that's happened just this year up into even just the last few months, there's people that are very excited and they're seeing that type of demand.
And I think you're going to see this, you know, move forward into the future. Yeah. You know, if you're a long-term investor, which like all my clients are, um this may be a good buying opportunity. Just, you know, price correction here. Um just the opportunity to buy it a little cheaper. Um, and I agree like if you know if you're going to build something this big, you're going to have some hiccups along the way, including politics.
Now, I'm gonna throw a little something different here. Um, which is that I think there's under the hood of the, you know, the big NASDAQ headline there. Um, you do have some, uh, diversion of performance here. And, um, I kind of see it as, you know, what does what does this election do? It slows down the data center build out.
Okay. Well, if it slows it down, who is impacted? How? If you are in construction, if you are selling the components of the data center, it's just going to be slower, but it's going to get built. But if you are waiting for the data centers to come live in order to collect revenue, that's a little bit more painful, right?
And if you're also levered on top of that, then you know, what do you do while you rate in terms of cash flow? So, I think there's a little bit of a differentiation here in what the markets are doing under the the big number. Um, and I think you can see that in individual stock prices today. And the only thing that I would point out there, and I totally agree, but the last data that I looked at, there's 3,100 existing data centers that are fully operational in the United States today.
Another 400 under construction, and then another,00 planned. Perhaps those other,00 planned will now be delayed depending upon the state that they're looking to build them in. But let's remember those other 3,100, let's assume the other 400, 3,500 [clears throat] existing data centers need power. They need cooling and they're supplying compute power.
So regardless of the outcome of the midterm elections, I still think there is benefit and growth opportunities with those picks and shovels. Right. The thing the thing that scares me the most cuz I'm I'm I'm going to say I'm the most bullish. Shy might might, you know, rival me, but this is what scares me.
We all agree and we're all kind of thinking the same thing, which I don't know. That's it's a little a little a little scary. But what I thought was fascinating was I was reading all of the morning recaps that were either written last night or first thing this morning. as the market was basically, you know, bot like just, you know, didn't have a bottom and all of the morning reports that I was reading, people kind of with, you know, with with their chest out were like, "Yeah, of course the market rallied and like they were coming up with reasons." Um, and so I can I can just imagine what we're going to be reading tomorrow, right?
Um, and it just reminded me that nobody knows anything. Correct. What I also think there's a bit of a You said a really interesting point though. There's like a bit of a shock factor of like why is the QQQ for example down 10% from the midterms when like if you go on Polyark or Koshi for example like it was kind of known it was going to be a blue wave.
Maybe not a blue tsunami but it's going to be a blue wave. Maybe the delta of the beat was just so wide especially regarding anything associated data centers. Now people are caught off guard but end of the day if you're a government like let's just pick on Texas because Texas is obviously the mecca of data centers.
26% of uh the data centers are in that state. If you're a governor there and you're gonna pause data center development, how are you gonna fill that vacuum of jobs that are going to be lost all of a sudden? You're going to have to answer to that next two years. You're not. If just go watch the HBO show VEP, like it's all politics.
It's theater. It's not real substance. Yeah, it's the best. But again, this is just a negotiation tactic that they're doing now that this she won. she's going to do something on a negotiation to maybe have the big tech to spend up a bit more, do something more for the community and the beta buildout's going to happen.
But guess what? If it doesn't, West Virginia is looking great right now for data center development. So, I think it's just going to find a different states. And let's not forget about the conversions of existing warehouses and distribution centers to data centers, right? I think about a company like Iron Mountain, big old boring Iron Mountain, which used to store paper and store tax documents for seven years in case we get audited.
Now, as a society, we don't store as much paper. So, what did Iron Mountain do about two and a half years ago? They pivoted and they lease out their excess data, their excess warehouse space to data centers, and they became part of the AI revolution and their stock took off. Will they be part of this moratorium in certain states?
Will other big large cap technology companies look to companies like Iron Mountain to lease out their excess warehouse space now for their data center needs? Time will tell. I think this may slow things down as it relates to new data center construction, but it's not going to slow down the need and the demand for data centers whether they're operational or converting.
Right. So, and and maybe Anna, maybe you can answer this because you kind of touched on this, but like we all agree that it's it's unstoppable, right? Does that mean it just got more expensive for them? Like, are they going to have to jump through more hoops and and the price or the to to pay to get it done got more expensive?
Is that or am I making stuff up? Well, yes. No, no. Um I think there might be some capital reallocation. Right. So one of the things that's happening here in the state of Ohio where I am is okay data centers can be built but we're going to have to negotiate. Um and it might be that the you know the the hyperscalers or whoever is building it may have to subsidize some of the energy costs that are higher because of the data center buildout.
Okay. Then capital is is definitely going to some other place right? um in terms of energy subsidy. But you know one of the let's talk about the bottleneck that everyone's talking about which is memory chips. So if things slow down then that bottleneck it it it releases a little bit right. So I don't know what's going to happen to the prices necessarily but logic just tells you there's a little bit of a relief from the price.
So I think there's going to be shifting of prices shifting of capital that we don't anticipate necessarily happening going forward. But so you mentioned Ohio um and what was it? Pqua Ohio. They they didn't just ban data centers. They were called the mayor. Correct. Am I okay? And the commissioners who approved it.
So Willie or Anna, can AI get built if nobody wants it in the door? Can can AI get built if it's that much opposition or is it going to a different state like Shai was implying? Yeah, it could go to a different state certainly. I think one of the Dakotas basically said, "Yes, please bring bring it to our state." Um, I I think I someone said this, I can't remember, but it's it's going to be it's politically popular to say we're going to ban it, but at the end of the day, when you see the states and when you see the cities and municipalities that actually participate and there's job growth and there's investment in the in the uh the municipality, I think they're going to be scratching their heads and thinking, "Okay, maybe we need to open up."
Yeah, exactly. I think it's going to be Oh, sorry. Exactly. Exactly that where people are going to these politicians and just the states are going to be jostling for you know what is the right deal you can give to these folks for uh ability to build data centers there. But I think what's interesting is we're not talking about whether or not there's demand for the data centers.
There's clearly the demand. I think what's going to happen is you're going to see price fluctuations, especially for those on the NeoCloud side, hyperscalers, that as this this uh constraint continues to happen, if it does continue to happen, depending on how the politics play out, there's going to be a lot of pricing power that these folks, especially the larger folks have that have these massive data centers, that have the power purchase agreements to build these data centers and install the GPUs, they're going to have a lot of pricing power to drive AI uh AI revenue for themselves.
So it'll be a very interesting dynamic to see over the next call it uh say like 12 months as this sort of plays out. Can you can you explain when you like when you say pricing power specifically like what do you mean exactly? I mean I know what pricing power but in this instant. Yeah. So, one of the things and we I'm sure we'll get to this eventually, but if you look at Anthropics, you know, recent amended S1, what you'll see, and people didn't really talk about this, but it's pretty incredible, is if you look at how much they how much compute that they had to bring in for Q3, call it like around 10 billion.
About half of that was in short-term contracts. So, they realized that they didn't have the long-term contracts to meet the demand that they were seeing, which speaks to two things. One is maybe it speaks to how far behind we are on this buildout, but the other part it speaks to was how much of an inflection we've seen in AI usage that they needed to bring on that expensive short-term capacity to meet that demand.
And so what you're seeing is and they they noted this is that had an impact on the short-term margin for that third quarter is that they had to take on more expensive compute, but they saw it as value accretive because they could still make a margin out of that, but the pricing is obviously higher for these short-term contracts.
So that's what I mean is and you're seeing it in in you're seeing it across you know the startup space where where we play as Neostellar where people are willing to pay these higher prices for short-term compute because they're able to have so much pricing power on the application side that people are willing to pay that right now.
I also I mean to add more color to what he to what he just said like 30,000 view just what changed I think right now the market like clearly the thing that changes the cost and process of building these AI infrastructure that's the one thing that actually changed this week but what didn't change like what he just mentioned is the actual demand and I think that distinction is really important right now because people are going the narrative right now in the market is people are going from communities want more control over data centers to somehow the AI cycle once in a lifetime cycle is over.
That's a massive dramatic leap right now. The politics clearly changed because communities are pushing back on the cost, water usage, tax breaks, but these enormous data centers are going to continue showing up and all the communities just want something more in return and that's what big tech's going to provide.
But you're going to say something. Oh, I'm sorry. Yeah. No, I was just going to say what about the demand for money? Well, that I think that's money. Yeah, that's a large part in that. I mean, that's a large part of it and that's why you're seeing Microsoft, Amazon, Meta, they're all near all-time highs.
I think before today, like obviously they're down a couple percentage, but I think that's also why they're they can self-fund and I think you're going to see a lot more maybe dilution on their end instead of going to the debt markets. Like Meta uh for example, they can just dilute three 3 or 4% a year and they can raise a stupid amount of capital by just doing that every single year.
That's in my eyes self-unding a bit. I know there's dilution eventually in the me media, but they don't have to go to the capital markets. These rounding errors for these big tech like $5 billion they can just raise. Well, these NEOClouds, these tier three NEOClouds, that's literally their bank for the next five, six years.
Like that's why you're seeing like this diversion a bit on winners or losers of this buildout where if you can fund it yourself and you've already done so you have that scarcity of these data centers being built out where you can charge this premium now on intelligence what he just said as well on that's not going anywhere this demand so you're going to be able to increase your pricing power all of a sudden that alone can fund this like community tax you have to deal on building these data centers.
Yeah, and I would just add to what Shay's just saying there that I don't think mega cap tech companies have been deterred spending wise as a result of the midterm elections. As my father always used to say, politics is very political, right? And I think the data center outcry became a political narrative that fit into this midterm election cycle.
And I think as time passes, as we move into 2027, and as all the panelists have said, when these states realize that there's going to be less jobs created, if there is a prolonged moratorum, less tax revenue coming in if there is a prolonged moratorum, and these large cap tech companies are going to go where their money is treated best.
They're going to go to those states that will allow them to develop, recognizing that in many cases, they're going to need to supply their own power. And that brings in small modular reactors and other nuclear capabilities off-rid. Or they're going to strip out the ratables such that the residents in the communities where the data centers are built, their utility costs stay fixed while the data center construction companies pay more.
I believe these companies are willing to do that and they have the money to do that. So I think each pullback that we see as it relates to this AI ecosystem creates attractive buying opportunities. portfolio managers like all of ourselves. Not that certain companies aren't spending too much, not that certain companies aren't financing their deals in the proper manner.
It's right to poke at those companies. But overall, I think the health of the AI ecosystem is still strong and I think the spending will continued although maybe now the development is slightly pushed out in terms of the data center side and new construction. I I also want to add sorry I I was just gonna add like right now I do think the barrier barrier of these bad agents of this buildout are it's going to improve like I think that it's going to be really good that the corre of the world for example they can't raise hundreds of billions of dollars because that's what causes a catastrophic
bubble for the industry so I think that right now it's slowing down due to community bureaucracy whatever you want to label it as but it's going to improve the quality quality and the bad agents cannot be bad agents anymore and that alone is great for the longevity of this bull market because the center of gravity is still AI and now I think I am getting a little more confidence that this won't turn into a bubble in next couple years
you know earlier when I talked about leverage uh problems with leverage I was actually thinking about Neocloud because you know they're the they're in that maybe it's not so sweet spot which is they're highly levered and they can't get paid until the data centers are built out and it's live and running right so you know they could potentially take it down but I you know there is sort of a safety net in that circular financing we've been witnessing for the last you know year over a year um so that all these big companies that are able to raise capital that rounding error is just gigantic number they have their backs right so I do think that that that there's a reason for that circular financing because everyone knows if you're a neocclad you have one revenue source and if that goes down that the whole thing can go down.
So yeah, no I think I love the I love the term you use center of gravity. I think it is a very very powerful momentum. I I love that. And Shai, oh, sorry. I was just going to say Shai, like I loved how you you framed it um in terms of like maybe protection against a bubble because the way I've been thinking about this, I'm I'm sober 17 years and when I was in the throws of addiction, like I would have paid anything,
you know, to to to get my fix, right? And I sort of feel like the data centers are in this position. they will pay anything and I think people are going to start to figure that out. Um and and I do think things are going to get a lot more expensive. Sorry, I stepped on did Willie were you about to say something?
Yeah, I mean I was just going to go back to the original question which was you know where is this money going to come from if it's not going to come from the debt market? I think that's a very interesting question and it's something that Cororeweave and Nebius these companies that are I think in the last quarter or two have their business models have started to evolve is starting to ask a very interesting question which I think will be what will drive the market in 2027.
So this so this is what it is it's the answer. Yeah. Yeah. So, so this is what I think is is going to be really important in 2027 is you've had these me mega hyperscalers basically do these massive contracts with core and nebas and in the last quarter you've seen their business model start to evolve so that they're starting to serve these larger enterprises.
What these enterprises were doing previously were doing, you know, spot contracts or spot compute with a corewave or Nebus and trying to do that. And what you've seen on recent earnings is them starting to take out larger contracts. So like one to threeyear contracts which have much higher margin for Core Weaver, Nepas.
And what the demands are going to be is you want this compute because you're going to start to build a product around this whatever mid-enterprise software company. you want us you want this compute, you're gonna have to pay a higher price, but you're gonna also have to pay more upfront in order to get that compute because we don't want to take on that debt or the debt marketers aren't there to help support that.
So, what I think the major question is going to be for 27 is if this AI demand is there, we want to see that be proliferated outside of just the major hyperscalers that are building this for anthropic or open AI. is is that going to start proliferating out to you know all these other enterprise software companies you start seeing it with service now you see it with Palance here but what about everybody else and I think that's an important question that is going to be asked in 2027 we see that in the private markets because they're much more innovative they're willing to spend a lot of that money because it's a lot of investor capital to secure that compute in order to start building but what the public markets haven't seen is exactly that behavior so I think either as you see some of these private companies start to go public and you see these pretty large contracts that they're willing to sign in order to get that compute.
I think when you see that evidence and then you also see the evidence in the public markets of these larger enterprise software companies starting to take down larger contracts, that's when you'll see a little bit more of where this money is flowing and where that revenue is starting to come from. I I mean I think you said it really well.
I'm glad you beat me too because you said a lot of things I wanted to say, but you said it in a much better way than I would have. So I I I do think that um also like the elephant in the room like you think principal Jensen is going to let the neocloud segment go bust. No. Like by principal Johnson I mean Nvidia.
Uh so I think that Nvidia for for example like 8% of their revenue is coreweave. Do you think that's they're going to allow the narrative that they are going to have a local peak on their pricing power and they're going to start having their growth erode happen right now? No, not at all. And Nvidia is printing $200 billion of cash starting in 2027 annually.
What are they going to do with thating? You we can go on and on about the circular financing if that's healthy or not. Fine. I get that. But it's coming from an actual cash flow that Nvidia is creating, not the capital market. So I think starting next year, there's going to be a lot more creativity on how this bailout's going to happen for the tier two and tier three names.
Also, like Nebius has showed us signs already on what's under the hood of these spot market prices on their bare metal. Like they post it online, you can see it every single day. And they also made the comment 50% of their uh customers do prepayments now for them. That could be a new signal going forward as well.
So, I think that there is a multiple lanes of capital. I think Jensen is going to save the day. He's been also awfully quiet all of a sudden past two weeks. I don't like that. I'm annoyed by that. I think it's cuz they're reporting earnings in a week or so, but worst timing ever for him to just all of a sudden go hang out with the family on some island somewhere, which I don't think is happening right now.
But man, I would love Principal Jensen to kind of speak up a bit right now and buck this narrative. So, since Shay mentioned it, the godfather of AI, Jensen Wong, I think once again what could provide more conviction to the AI revolution and reassure a lot of investors once again is when Nvidia reports on November the 17th.
Let's say for argument sake they beat on topline. They beat on bottom line. They guide higher again. They show that they have a diversified revenue stream between chips and between their data center business. They actually show that there's still insatiable demand for their chips and what they're doing with high bandwidth memory to power their next generation chip, that being the Reuben chip.
What if they check all those boxes again? Will that be enough for investors or will they say, "Ho h still not enough. Let's sell off further." or does Nvidia once again come to the rescue? I think that's going to be a very interesting day for the markets. Mark on your calendar, November the 17th. November 18th is my wedding anniversary.
That could be a very interesting day as well. But November 17th is going to be an integral part for the the current sustainability of this bullmark. And I would love to hear what all of you had to say, your thoughts about what you think Nvidia might or might not report on November 17th. Before we do that, what what are what is your thought?
So, I think once again they're not going to knock the cover off the ball. They're going to set records in terms of revenues. They're going to beat on earnings. They're going to guide slightly higher and if Jensen's going to talk about the the lack of supply of what he needs, and that's high bandwidth memory to power their next generation Reuben chip.
So, could there be delays there? And could that be a warning sign that these delays are now starting to impact the roll out a lot of these companies? So, that's my thoughts. Where where do you land, Anna? Um, I think a lot of the more of the same, which is lots of robust demand. Um, sorry. Okay. Lots of robust demand.
Um, great guidance. Um, somebody's going to ask him about, you know, the midterm elections and I'm sure he'll have something to say. We want to hear it today, but we might have to wait. Um, but, you know, my I pay attention to his price action to that, right? So, you know, several quarters it's been the same thing.
Demand is robust. Things look great. Um, and then of course the prices don't necessarily reflect good news, right? So, um, I think what the market is pricing in yesterday, today, and maybe even with good news on the 17th is beyond it, right? So, what happens in 2029, what happens in 2030? Equities are a long duration instrument.
So, investors naturally look out, right? And of course, you know, if you do, you know, discounted cash flow because of the time value of money, if it's delayed, even for Nvidia, you know, the the stock price may actually reflect some of the delays there, too. So, um, that's really what I'm watching for, but from Jensen Wong, I'm sort of expecting more of the same.
Sure. Um I I I do think they could save the day, but we were all chatting before and Shai um mentioned some chatter that's on the street uh about a delay in uh Anthropics IPO. That's a curveball. And then obviously OpenAI announced, you know, that they are going public in 2027. Enterprise is now um leading the way revenue and sales.
And so that's that's a whole another curveball um that we're looking at. Yeah. I mean I I personally think Nvidia is going to be a nothing burger. Again, like it's there shouldn't catch anyone by surprise. Like they've been a nothing burger all year. I think they're going to validate everything that's happened to AI demand.
It's going to be record numbers. They can say the moon the pathway to the moon is so extensive for AI. Nobody really cares because unfortunately the narrative on AI isn't really AI. As ridiculous as it sounds, it's not AI demand. It's just more rate markets. Like what's the Fed meeting going to say in December?
What's this anthropic IPO going to look like around Thanksgiving? Is it going to actually get validated on what the multiple is going to look like for a frontier model for the first time? Because we've only gone private equity stories for the past couple years on what anthropic should be worth. Now it's actually there.
Let's see what the market's actually going to be willing to price it. If it's at goes public at 1.5 trillion, which again is like uh it's supposed to go at 2 trillion, 1.5 and it's still red and goes to towards 1.2, I think that's going to be a repricing event for the whole AI trade. And I'm not saying the AI demand is going to erode.
I'm just saying the AI trade is going to get multiple compression. And I think the anthropic IPO, Fed meeting, that's a bigger um kind of saving grace, I guess, or a clearing event on what's happening in AI. man, if the Fed gets hawkish, like I I might just like close the computer the rest of the year.
Uh or it might ruin my holidays honestly at this point. So, I think uh I don't think that will happen. But also like Nvidia can't disappoint either. I mean that would be catastrophic. That would be the first time in what three or four years uh that Nvidia would disappoint. So that's you expect greatness.
If it's not greatness, that's going to be way worse. If it is, nobody cares. I totally agree with Shay here. three critical moments for the balance of this year. One, Nvidia earnings report on November the 17th. Two, what the Fed decides to do or not do at the beginning of December. And three, whether or not Anthropic still goes public this year or delays going into next year.
I think if they delay, that could be perceived by some investors as bearish. I think that if they don't delay and their IPO fails or comes in below expectations, that could be perceived as bearish to certain investors. Damned if you do, damned if you don't. But if in fact anthropic, let's just let me allow me to paint this for a second.
These are a lot of ifs, but let's say they do go public and there is demand and they meet their price and it exceeds and stays at that price. If the Fed does meet in December and they stay on pause, at least for the bounce of this year, and if Nvidia beats on the top line, beats on the bottom line, guidance higher, promises the moon as it relates to AI growth going forward and the stock doesn't sell off, boy, would that be a strong three tailwinds to guide us forward through the balance of this year and provide a great balance and close out to the year.
Will those three things happen? I don't know. But certainly three things for all of our uh listeners to pay attention to. Yeah, actually I I don't know if I agree with you, Kevin. I think that the thing that the only thing that matters for the rest of this year is the entropic IPO. Nvidia, like she was saying, they're going to say what they're going to say.
They they've not wagged the market in in some time now. It went from Nvidia to the hyperscalers. Now it's going to all these other pieces of the infrastructure. They're going to say what they're going to say. They have orders out to 2028. people are trying to secure capacity for data centers into 2028.
So it would be astonishing if they said anything other than what they're supposed to say. I think what really matters is exactly what we've been talking about which is what's going to happen with the anthropic IPO and people that are looking at the entropic IPO are looking at two things is one do you think that there's the durability and the AI growth story not just today but into the future and the second question is do you think they're going to have a meaningful market share of that AI the AI story and growth those are the two questions that every investor is asking within their investment committee every single day before the IPO And there's all these other components that are in their S1, their amended S1.
The numbers are what the numbers are. It's really the value that's going to be driving whether or not Enthropic is worth that much or not is people's perspective on those two questions. And so I think that's what everybody needs to ask themselves. And you see in the public markets there is trepidations on whether or not either anthropic has that ability to capture that market share or people want to see more evidence of where that growth factor is going to come for AI.
into the next year or the couple years after that. So I think that's what's really important into the story today is what is the next leg of the AI growth can we prove that whether that's anthropic open or whoever and I think that story is what needs to be developed. I I think all very valid points, but if Nvidia guides lower and if the Fed raises rates in the face of what's happening right now, that could ruin my holiday season as well as the what Shane said earlier,
but but but also like I before I pass additional color just real quick is like we already got Nvidia's earnings, the hyperscalers, the hyperscalers already reported. We know what's going to happen. So like uh also TSMC also reported ASML like we all we got so much validation already on earning season that if Nvidia does guide down what's the dislocation all of a sudden what's the new information that Jetson sees that the rest of the ecosystem doesn't that would cause mass fear because that doesn't make it doesn't correlate with everything else we're seeing.
Like I there's no I give it a 1% chance they miss. Like maybe I'm not saying it's going to happen, but if it did, right, that would be like Yeah. un unheard of. But Willie, I'm curious. Is versus yesterday to today like is is there a chance anthropic like delays? I think if you see the market continue in the last whether today is an indication of where the market's going to move over the next few days, I would find it very difficult that they would be able to go out.
I I'm not saying that in a way that Anthropic isn't a great company and they're not driving a lot of AI innovation, but I mean the public markets, they have to accept and want to be able to pay that price for anthropic. And I think people are asking that exact question is is this AI growth durable and whether that's driven by the compute if they can get enough access to compute if that's going to be constrained or if they are going to have the market share that they say they're going to have.
I think those are the questions everyone's trying to ask. Anna, what do you do you think there's a chance that Anthropic I mean to turn around the Titanic in the in the back room discussion somebody must have you know alerted the team that you know they would be IPOing right after midterm elections.
I don't know. Um but I I I do agree that it would be very difficult. Now the question is is the selloff going to continue to happen? you know be difficult to go public November. I'm sorry. Sorry. I'm just clarifying. You're saying it'd be difficult to go public or difficult. Yes. No, it it would be difficult to go public.
I mean, it's I think they're going public. Yeah. Right. I hear you. But I think I think all all systems had Sorry, I I I cut you off. Keep going. Oh, me? Yes. Yeah. So it it it would be difficult because the multiples are shrinking right now and and I think the investors the private investors some of them can get out day one and if if the market is like this and the the the reception probability is not very high for a good reception in the public markets would you really want to IPO if you can get out that day I mean I'm not entirely sure that they would be ready to do that
well also like just look at the track record recently cerebrra is like what 50% below their IPO price. Uh 80% below their post IPO high. I think SpaceX is trading below their IPO price. I know for a fact from like my buddies who work in the space. If you get priced at 1.5 trillion and you close below that post IPO, there's a lot of funds that don't touch the stock,
won't get exposure. It's mispriced and they just like have they have a rule in their fund that they can't really buy that stock if it's below their IPO. So I think there's a gamemanship where man if AI continues fading like right now I'm seeing QQQ is down 6% still if it continues at this pace for another couple days because guess what Fridays are never near trends changing days if we don't bounce today tomorrow is going to be red nobody wants to hold this ahead of the weekend and now all of a sudden week of your IPO
maybe I'm deluded um but like I think I think anthropic is bigger than the market and I think anthropic is going public. Like, I mean, you guys know better than I do, but I I just can't see them. I I understand the math, but they're bigger than the market and they're going public and you don't have a say over it.
If you're the CEO of go public, that's what I'm saying. Like, focus on Dario cuz Dario is an like N of one. Like, he doesn't care about public opinion at all. He's kind of the middle middle finger of AI a bit. So, like you're right. Mike just like who cares what you guys say? Yeah, exactly. He's very much like Karp.
Uh Karp delivers a little bit more. Dario does fear. But yeah, so I think they are kind of a little aligned with each other, but he's unpredictable. You never know. And again, it's not if they go public, it's when they go public. Right. Right. Of course. And Willie, you're not going to like this, but I have an analogy.
Sure. Okay. Anthropic is to Coca-Cola as open AI is to Pepsi. Am I wrong? No. I think you're 100% wrong. I think when you look at the complexion of what OpenAI is doing. Yeah. Like I think I I understand that analogy. I I think if you asked me six months ago what the private markets were feeling, I think you would be spot on.
I think that the private markets felt like OpenAI was losing a bit of the the momentum and that and profit was eating their lunch. And I think today what you're seeing is that narrative start to flip almost entirely, right? Open AI has this amazing consumer access vehicle and as you see news come into proliferation.
They have this great distribution with dot with with their ability to proliferate dots to do the personal agent piece. Robin doesn't have that that channel. That's not their strong suit. That's not what they've done. They've been focused on the enterprise. And now you see reports that OpenAI's enterprise business is greater than their consumer business.
Right? So they're clearly winning on the enterprise side. What I think everybody is missing in the public markets where we get a lot of confidence in the private markets, this is the whole question of whether or not there's this AI growth second leg that everybody needs to get comfortable with in the next couple years is when these large language model companies release models, they release the smartest one that's the most expensive.
Then they release a cheaper one that can do 80 90% of the intelligence and then so on and so forth until you get a model that's very very cheap that can do that level of efficacy of work, right? And if you look at everybody else that's building on top of that, they'll start building what the most expensive thing, but then they need to figure out how to make their application work with something cheaper.
So there's this lag between the most intelligent piece of model that comes out and then ultimately what gets built by enterprises that gets sold out to everybody else and actually drives revenue. I think what people are seeing right now is you saw these amazing models get dropped by anthropic open AAI and you're seeing everyone now go back into their R&D and build on top of these and releasing products and start to release that products right
we just haven't seen the revenue from those come out from these enterprise SAS companies you see it from the the startups which are moving very quickly and capturing we see companies that started with 5 million of AR at the beginning of the year they'll end close to 150 200 million of error at the end of the year.
If you saw that for any of these enterprise SAS companies on a new product, people would go would go nuts over trying to buy it, right? So like that's the the action the price action we see in the private markets and this lag that you see between you know releasing the latest model and then ultimately I think a lot of the revenue that gets generated later is people building on top of it using it because they want that cheaper price for the intelligence.
Right. And so I yeah I I I hear what you're saying. I think one way that I am wrong on that analogy is people do kind of just say chat G TBT no matter what like they they're saying it you know in lie of AI um and so they do get that but you know if I'm at a restaurant and I order a Coke and they say I have a Pepsi like I say okay like but um I do have a question um a month ago uh when all of the CEOs were sort sort of, you know, sort of hitting the alarm vials.
Not all of them, but there was it was a little bit of a divide. U and some people suspected that they were trying to create a moat. Um, is it possible that the Dems sort of sweeping the midterms have maybe made this come true in a way because it's it's going to be a lot more difficult for people to enter the market.
100%. Uh, first off, like AR AI has a PR issue because of Daario and maybe because of Sam as well. And guess what? The politicians use that leverage to win votes for their re regions. And that was like one of the easiest levers they can pull on slowing down AI progress because the fear got created from Daario.
And why did Daario create this fear? Oh well, open models are gaining a lot of momentum. We're a $2 trillion company growing 50% sequentially. Only way to maintain this momentum is having guard rails as a duopoly with open AI. How do we do that? We we yell from the mountain tops, "AI is happening too quickly.
We can't control it." And now the government's going to intervene like, "Okay, we trust you, Dario. We trust you, Sam, to be our saviors of AI." And that's going to create a massive barrier of entry for all open AI models, all tier 2. And those That's fascinating. Do you do you do you think they were that far ahead?
And and do you think it was a setup for the midterms? I think they were planning this IPO for over a year and they have strategy sessions by the weekend and this came up and that I think it's that's brilliant. It just it's makes sense but I was shocked the fact that they didn't calculate today's performance in the market
but I don't think they're looking at one day's performance either. They're looking at their competitive pricing power over the long term. Right. Right. So, I just really wonder I wonder if Dario was fearful about the technology or was he more fearful about the potential litigation and fearful of competition and he's looking to create more barriers to entry as Shay is suggesting here, which I don't think that's uh entirely implausible.
I mean, listen, what's going to erode a $2 trillion company? Everyone can just build on top of open AI models and all of a sudden like why am I paying as an enterprise? Why am I paying this much for you anthropics for you open AI? It's natural because they're growing at a rapid pace. Like I'm I agree like this the private markets like this growth if it was a software company it's unheard of.
It's it's an N of one. But the biggest question now is going forward for Anthropic everyone knows the growth is going to decelerate. What's the pace of that deceleration? That's the biggest question especially for Anthropic operating at this scale and that's going to become the really important price discovery for how to value Anthropic in 2028 is how much of this growth is going to decelerate because everyone in 2026 they're experimenting.
Yeah. It's just throwing something at the board and like let's see what sticks the churn in next year two years is going to be the entropic story. Who's staying? Who's actually making money off it? And that's the biggest unknown and that's kind of like uh am I buying a $2 trillion company? It doesn't matter what the growth numbers are cuz in 2028 like I don't know because only a handful of companies are actually making real money on AI right now and Palanteer is the number one company doing that for everyone.
Yeah, I think the analogy Sorry, go ahead. I used to call the uh open AI and and anthropic the fountain head. You know, that's the where the knowledge lives. Um and you know, the value really is in that channel to the end user, whether that end user be an enterprise or an individual user and we really haven't developed that yet, right?
Um and frankly like you know Anthropic you know goes live in IPO and then the question really is is it going to be another snowflake? Is it going to be another cerebras? Is it going to fall because these software uh companies are coming up right and and that channel is being built. So there is actually like a long-term value proposition question about anthropic and open AAI.
I know they they probably need to strike while the iron is hot and they are sort of the only game in town at the moment. Uh but you know, years long, I'm I'm really not entirely sure um how they're going to develop as a company and what their c customer base is actually going to look like in terms of profile.
I think your analogy of a fountain head is exactly where I would say is the analogy that I want to make is that open AI feels like getting offered a glass of water versus anthropic is ask asking if you want sparkling water. Are you willing to pay that premium for sparkling water rather than say call it Coke or Pepsi?
Like I think that's the better analogy. And then it's a question of can then open AAI then offer other things with a base of water versus anthropic offering a base of sparkling water. Can they offer soda or all these other different products on top of that? Or then can you just say you know water do you want to offer juice or all these other things that OpenAI can do?
I think that's really the the the bifurcation that you're seeing the types of products that are going to be offered by these companies and whether or not they can do that sell to those next products. And can I just since we keep bringing up Coke and Pepsi here, my wife will only drink Diet Coke. I will only drink Diet Pepsi.
They taste incredibly different and their stocks have performed incredibly differently. So, I just want to put throw that hat into the ring as well since you keep using that as analogy here. Now, to add color to the fountain head and the sparkling water, natural water, I agree. But guess what? Like, I used to spend $100 probably on sparkling water and now I got bought a machine at home and I make it for free with a fixed price every month.
That's what's going to happen with these frontier models like there isn't a moat except for I don't know the connection of just having your memory in this specific model which is great for right now and there but eventually capitalism is brutal companies enterprises will create solutions to save money on their end on top of these open models that's natural it's going to happen but again they do they have something under the hood that nobody is aware of that's just this like RSI type of model that is self-reflecting on AI on AI I maybe, but that's a hail Mary to bet on a company at $2 trillion at the IPO.
Is that exactly why you're trying to create this moat by throwing up the need for guard rails and regulation, do you think? Oh, if if someone threw a rumor that I might get cancer from making my own sparkling water at home, I I'll buy the can. Like, I'll trust that more. That's kind of the fear that he's creating a bit.
I know this is like a ridiculous analogy we're creating on two once in a-lifetime companies aka open anthropic, but it's kind of what's happening right now in a really dumbified way, but it's kind of that's what's happening. Is so is is it possible that this is kind of going to be a reset and maybe the PR problem has improved?
Is that is that possible? Well, I think that's why they're all going for enterprises right now. And I also think that's why OpenAI um hired that coder from Enthropic. I'm blanking on her name. Uh, but Catwoo, no, Catwoo. Yeah, Catwoo is a big hire. I think Catwoo is a probably a perfect example because the labs right now are fighting so aggressively for product and engineering talents.
Uh, where you're seeing like companies like Aster Labs, for example, they just reported this week like they're growing headcount and semiconductors are hiring for skills that they don't even need 5 years ago. Palanteer is growing like 100% growth with a really small sales organization because the product and AI tools from make these existing teams so much more productive.
So tech hiring is becoming in increasingly more polarized where the best AI talent can essentially name its price while generalist entry- level hiring becomes a lot slower. But man coding talent is probably the scarce input right now. Scarcest input right now for these developer companies. Yeah, the bottlenecks I see in the AI revolution are obviously compute, power, memory, and now, as we're finding out, talent.
And talent is appearing to be the most expensive. And these companies are going to need to pay up to get the talent that they need and marry that talent with the compute, the power, and the memory that they need to continue going forward. And it's a very expensive proposition. And investors need to realize that and be patient.
Also, it's a huge because look at Apple the last four years. They can't solve Siri because they have no AI brains in their organization because they didn't buy the AI talent. And luckily they have the every they own eyeballs. It doesn't matter. But everyone else doesn't have the luxury. So like AI brains matter a ton.
Just look at Apple's development and R&D the past three years. Yep. Willie, it looks like you were going to say something or No, I mean I was just going to say because you were talking about Catwoo and the importance of hiring for product. I think what you're seeing develop for a lot of these AI labs is look I think the intelligence actually is a pretty strong moat being the leading intelligence there even if there's open source that is you know slightly that's close that's like 80 90% of intelligence I I think there's lots of organizations that are willing to pay that extra premium for the leading intelligence model for any whether it's open AI or anthropic they're willing to pay that extra amount what I think is important and why product is becoming very important for these companies And why that hire is important is because yes, it's important to build just the intelligence.
But when you look at cognition on the private market side, they're effectively replacing what is cloud code for a lot of organizations. It's not just the intelligence, but the harness and the tooling around that. How you build like a vertical product, AI, a vertical AI product on top of that. And so you have all these AI labs that are now trying to figure out how you can capture not just the intelligence because that's great revenue, but also being able to charge additional for the full vertical application.
And so you're seeing OpenAI do it, you're seeing Anthropic do it. You're seeing all these intelligence players not just focus on just intelligence. That is just that is a huge fountain to be able to tap. But how can you then own the entire workflow and then be able to have that stickiness of you're not going to rip out this entire workflow rather than just the intelligence?
Well, I also think that stickiness is agentic AI and I think that's like we're all talking about training AI for enterprises, but a gentic AI is going to be that stickiness for frontiers. And I I I just don't think OpenAI and Anthropic have really done much on that front compared to Grock and Muse and even Gemini.
So that's a big question as well is like what are they going to do for AI shopping assistance? Wait, you gonna say something Anna? No, I'm going to build my own agent so that it can actually interact with other shopping agents to find the best price. Well, also like I I think as I say say that like can Anthropic and OpenAI really conquer AI agents because they're going after enterprises.
Like the real winners of Agentic AI, they're going to be the applications that can buy stuff for you like groceries, household supplies, I don't know, replacing the same product that you've already bought um or a cheaper version of that that's really easy to return if the AI agent messes up. You can't do that in the enterprise space.
you you just don't have that luxury of like having an AI oopsies because an enentic AI went somewhere like went bye-bye like you don't. So I think that consumer mentality like maybe open AI has that because everyone says chat GPT this chat TV that they have that mind share already but I don't know I'm not giving an agent complete control over my enterprise data or choosing a house or buying a car or any kind of emotionally important gift like
you say that you say that today but it could be different. I feel like we've been being and this maybe this is the the wrong terminology, but I feel like we've all been being groomed for the last 10 or 20 years. Like we all jump at the chance to make our lives easier, right? Like we all like we we're giving them everything almost as it is.
Um and so I just think it's a matter of time. We [snorts] are a society of convenience for sure, but that convenience comes at a price and it co comes at a a cost of risk, right? So I think some of us want the convenience but at what cost and how much of our personal data do we need to expose to get that level of convenience and I think that's what the consumer is going back and forth with right now and enterprises clearly are not there from a security standoint.
Let me ask you this. Let me ask you this Kevin or everyone actually. Um so I 100% agree with that like there's a choice that you have to make in terms of like giving your data and whatnot. What happens when it's a competitive disadvantage to not be going allin with this stuff? Like if your competitor [snorts] has 87 AI agents working for them, like you're going to lose.
You you Yeah, you could argue that you're at a competitive disadvantage now if you haven't started to spend in and invest in AI capabilities because it's going to take quite some time to catch up. But I think you do need to do it thoughtfully and from a portfolio manager that's done this for a better part of three decades right now.
That's why cyber security is so important to safeguard that data and that's why I feel that's a tremendous opportunity within the AI ecosystem that we very rarely talk about. I think that's the nuance here. You know it's not a binary thing of you either use you know 90 agents or you don't use anything at all.
There's something in between. And it's also not, you know, I'm gonna hand over all of my sensitive information about my business [clears throat] to AI agents versus not. I mean, there is this middle ground where you're actually interacting with the agents because the agents will keep coming back and say, "Do you want me to do this?
Do you want me to send the email or not?" I think you can actually have some control over the process of that workflow and allow the agents to come in uh you know and draw the boundary. Um, I think you can have that and so that interaction I think is where the workflow is going to live. Not necessarily handing over everything to the agents.
Compromise. Do you trust the agents? Do I do you trust a do I have um I have uh claude agents u that do a lot of the administrative stuff for me because I'm an entrepreneur, right? So you know one of the benefits that I have is you know I started my business right as AI was becoming really interesting.
So it allowed me to really experiment on using agents for my business and it really has been transformative. Yeah I mean for me yeah go ahead. No no I was just say it's a conversational tool for me. I don't trust it at all. And I think that it's not even close to a phase where you can uh I also think that trust is actually a premium price you can offer to people that they were willing to pay.
I mean, Meta went that down that route. They didn't go for the fastest. They went for like as crazy as it sounds. Zuckerberg went for the more thoughtful approach. I didn't think he was capable of that, but he did. He didn't fall into the prisoners dilemma that a lot of these CEOs face right now in in AI.
And I think trust is going to be a big factor. Just look at Palanteer. They're creating their whole 2027 strategy is sovereign AI. And if you want to like follow what AI is going towards, follow the leader, Palanteer is the AI application leader of this generation. And what they're doing is what's going to be important for everyone else down the road.
Sovereign AI is huge. Trust is going to be the number one thing for everyone. Yep. People pay for trust. I actually think right I I think that's where OpenAI is very interesting because if you ask people when OpenAI and OpenAI chatbt first came out it's like would you trust this with your personal information like your taxes for example like can it help you with your taxes?
I think the first year everyone was very hesitant. They're like, I don't know if I want to put all that information inside of chat GBT. And then they hear their friends did it and it was really easy for them. And so then you decide to do it and now it has your social security and all this other information, right?
But you still trust it because it's not it's not like it's blasting it across the internet. It's now in the memory with open AI. It's similar to how like people got comfortable trusting plat like one of our portfolio companies plat. So like people trust that with your banking information like that's that was a pretty big jump but people got comfortable with that and be able to connect those rails.
And so there's like these steps of like people trusting different partners whether that's open AI or cloud or whoever with these specific types of recommendation. And it's a question of if agents can then connect all these these pieces and give you an experience where you're comfortable that your information is safe and that it can still execute a lot of this stuff.
I I think that is really going to be sort of the flow of this rather than a hey like trust giving muse your credit card information, your banking information and all this other stuff. It is these players that have already gained this trust or already earned this trust with their customers and then being able to connect these different pieces together into actual experience that that works.
Well, let me let me ask you this. Do you think Oh, I'm sorry. It's okay. No, I'm um I'm gonna have to say goodbye to you all very very shortly. So I speaking of trust, I trust all of you. You guys were wonderful. Thank you for joining. Uh hopefully you all come back. Absolutely. And uh I learned a lot.
Yeah. So thank you. Great market hang everyone. Yeah. Heat. Heat. Hey, hey, hey. Heat. Heat.
댓글 0