Hello and welcome back to another episode of Sharp Tech presented to you by Stripe. I'm Andrew Sharp. Joining me in studio today, Ben Thompson. Ben, how you doing?
I'm doing okay. I do need to know what is more annoying.
I have a ranked choice list.
Number one, your... It's not over this week.
Here we go. That's right. Anyway, by the way, ranked choice, terrible idea. But anyhow, uh, number one, for another day. Flight delayed.
Wait, we got to get the list first. Number two, no Wi-Fi on the flight. Number three, this is take three because we've been having some technical challenges.
A good start. Also, so, uh, I'm going to go with number three there. That is by far the most annoying.
I could see it. You came in all chipper and cheery after no Wi-Fi, no flight delay, good rhythm, and now you're just getting madder and madder. I can see it.
Absolutely red as hell here on YouTube. Um, in any event, I'm ready for a great show now that we're finally off and running here. I've got two Diet Cokes by my side. Lot of ground to cover on this episode. So, we'll see what we get through. We will begin with the elections this week. You may have heard elections all over the country and all over the country, everyone was voting against AI data center construction. Out of 44 local ballot measures that oppose data center construction, 43 of those...
I'm glad you got that right. You said everybody. You sounded like everyone on Twitter was like every single one failed. No, we strive for accuracy. Uh, North Dakota was pro data center.
We're going to have to pause for a fact check. Um, maybe even more important than the... actually, can I add on that though? It wasn't just that all these measures against data centers failed, in several different places like mayors or like board commissioners got recalled.
Which I think is arguably more important than these measures failing because that is like a signal to everyone in office if you like your job on the line.
That's right. So it's like that it's not just that these measures passed. Mhm. It's that future measures that don't go to the ballot, but just go to the mayor's office or just go to the board.
Exactly. No. And look, there's been a groundswell of opposition even since we last podcasted about it. I think we talked at length about the data center opposition. Look, I... we've been... in April or May. Is there going to be a lot of like I said, like we talked about previously, but I do think this is one we like we were... we were early on and then the acceleration has just been unbelievable to watch and maybe even more important than the local ballot measures. Gina Hahosa was elected governor in Texas. She ran on a promise to impose a moratorium on new data center construction until the state legislature can...
Right. This state legislature that only meets every two... once every two years.
Oh, I wasn't aware of that.
Oh, you don't know. This is a great thing about Texas.
Part of the bottleneck. Okay.
Yeah. A great thing about Texas where one of the unique things about Texas is maybe a better thing to say. Yeah. They have their legislature meets once every two years.
So, uh, so there's this is more meaningful. And the other thing about Texas, uh, you are right to say more importantly Texas. Uh, back in the spring when we were talking about this, uh, I went to a conference and um I was talking to someone very senior in the in the data center game. Uh, he had been very heavily involved in the Wisconsin data center right up the road that we talked about in our previous podcast that got cancelled. But basically his take was it's fine. Everything's just going to get built because Texas, right? And and it's Texas. It's not just the...
Is that just because of available land and power?
Power? Yeah. There there is so much natural gas in Texas. There is there they set it aflame. So like they're pulling like oil out and there's just tons of natural gas coming out that like all that is there to be captured. There's a reason why there was a huge Bitcoin mining thing in Texas before because there's just all this excess energy and it's its own grid which is has pluses and minuses. But in this case it can like it's easier to...
For all this bring all this power online. You're not dealing with like in the US there's three grids. There's like the east grid, the west grid, and the Texas grid, right? It's like its own thing.
Oh wow. This a very educational episode. So many fun facts here. No, but that's why that is the one that really matters because the presumed backstop for all data center opposition has always been...
We'll just build it in Texas.
Everybody's going to Texas. Yeah. I mean, they were on track to be home to 25% of US data center capacity by 2030. In the long run, if you talk to anyone, the assumption is it was going to be way more than that. Like just going forward, yeah, it's fine. New York's going to pass a moratorium. Even Ohio, which is a a a very significant uh portion of it, that's, you know, yeah, might might might end up there.
We have 50 different states that can compete for these projects, but uh infrastructure-wise, Texas obviously has all kinds of advantages and may not be replaceable.
Well, and the assumption that it was going to have friendly government and so that will take care of it.
Well, that is out the window for now. Uh, we'll see what the state legislature says whenever they get around to meeting about any of this. So, what should the tech industry actually learn from the midterms? Can they actually change public opinion on AI, AI data centers? Is that a lost cause at this point?
Well, the question is is the lost cause is can the tech industry learn? We should probably start there.
Fair enough. Well, and looking at the market, it is a wipeout as we record this on Thursday afternoon. Thursday, um, we had a little dead cat bounce on Wednesday to throw everyone off the scent. Now, who knows? Um, is the American growth story just cooked in light of all these developments? Like, where's your head at as we process all this?
Yeah, the um I think we were right before. You're going to have to get a lot more transactional. Like, there there's the the degree of arrogance around this data center thing has been pretty stunning. And on multiple levels. Number one, just starting out and being like, these people don't get it, they're dumb, there's not a water issue, etc., etc.
Yeah. Missing like getting stuck on the details of what people were saying and not grasping the overall like people. Why is it that people are grabbing on to this? Go back the whole fake news debate.
Mhm. The everyone gets the whole fake news things wrong for ages and ages in that they get cause and effect backwards. They're like, 'People hear fake news, they believe it, therefore they support the wrong thing.'
And no, what the fake news thing is confirmation. It's defeating a confirmation bias. It's giving people ammunition for what they already want to believe.
And so you go back, you know, way back when, uh, to like, you know, 2016, the whole fake news thing was coined. You you have to sort of zoom out and understand why is this getting traction? There's like widespread resentment of...
That's right. So elites, right? So it's not like and well, it's just so infantilizing in...
Infantizing. Wow. I wasn't really sure about that one. Um, it's so in... paternalistic. Yeah.
Right. Where uh it just assumes people are stupid and if we give them the right information, they will change their minds.
That's right. Right. And um and but you understand why people come up with the people are stupid because some of the fake news stuff.
Well, some of the misinformation surrounding AI data centers is just easily disproven or refers to the way data centers operate.
You're operating on the wrong level, right? You're you're trying to refute facts when what you're actually dealing with is sentiment. And it's the sentiment that is attracting stories and that's why people will discard okay fine that's not true and they'll find another reason to oppose the data centers. They oppose the data centers and why do they oppose the data centers? Like there is just a fundamental distrust and unease I think with technology generally and certainly with AI specifically. That's why I do think this isn't even though it was all about data centers and people are talking about the environmental costs and electricity...
I think people are just voting against this version of the future.
Uh, and why not given the version of the future that's been sold and this is where you go back to globalization.
People didn't have a vote, right? The factories just closed. In this case, once people had a vote, they're not too happy about it.
Absolutely. I mean, if we go back to the spring, I said then, I still feel now it's an understandable human response given what AI portends, the way it's been marketed, and the ongoing centralization of wealth among a handful of big tech companies. People just want a chance to stick their finger in the eyes.
And why not? I mean, I can't stand most of big tech in the way they talk about AI. Like we the you know... I'm very proud of this one. I might repeat this one every podcast for the next several. If you love advertising, you're optimistic about AI.
Because why why do I like advertising? Because I advertising good advertising helps the customer realize they have a problem in their life and then it gives them the solution to it. Now, sometimes that's very cynical. You're like creating a problem that doesn't exist and then saying you'll answer it. We've seen lots of political ads the last few months. We know exactly how that works, right? But sometimes it's true. It's like I actually Yes, I could actually use that thing. There's a problem I didn't have and that's going to help solve it. This is very analogous to the AI question in two ways.
Number one, everyone trying to create AI for consumers fails to do this. Mhm. The the the the issue isn't that you're not showing people you're not showing people like what are problems you have in your life and how AI solves it. Like that's why the AI show and tell thing we do like we got tons of responses. I I appreciate it. I saw you moved it up in the outline today so we could actually get some into the show.
But like it's there's an aspect of just realizing there's a like... the humans are so adaptable and so we actually adapt to tons of problems in our lives. We just accept that's the way it is and actually it could be better.
So number one, AI has terrible marketing in the like it can actually make your life better in part because it feels like no one in this industry has any sympathy or understanding for normal people which why not? They're all like crazy rich all around the coast no conception of how normal people live.
Well, and I will also add that they're building a technology that is exceptionally expensive to build and doesn't have a clear application beyond saying, well, this can actually just replace wide swaths of the workforce.
No, exactly. Because they can't articulate it even though I I absolutely thinks it think it exists. And then number two where where they fail is in the again this is where I think there's a relation between not appreciating advertising as a medium is they see humans as utilitarian cogs in the wheel like that are that can be replaced because their desires and simples are all knowable and understandable and we can just design the perfect system. No, this is where you get the the alignment with the sort of Marxist view of the world, communist view of the world of look, we can design the system because we know all the inputs and humans are just reduced to an input in the system and we can recreate the input with AI and actually no the adaptability point is isn't just we can deal with discomfort, we can also generate new needs. So like like humans are like need generation, desire generation machines. We come up with all sorts of crazy stuff we want or want to do and serve and then people rise up to meet those needs or people figure out, oh, some number of people have this need. I'm going to figure out how to serve it. Wow, the capability I developed is scalable or other people could do this. We just need to let other people know they have this need. That is like what good advertising is. And and the whole industry, not the whole industry, but huge chunks of the industry are just they have such a narrow view of the world. They they they're humans fit in this column and the measurement marks on that column are like IQ points and that's how they rank the whole world and they're like, 'Oh well, AI is climbing up the column going to wipe you all out. Think how much money we're going to save, how efficient, right?' And so implicit isn't just the fact that the only message that is coming out of Silicon Valley is AI is coming for your jobs. Mhm.
Implied in that is all of you are worthless. You all of you like are are uninteresting to us. And this is why I go back to all of us uh I didn't vote against data centers, but all of Americans are raising their hand to the sky and flipping the bird to all these tech companies.
Absolutely. And why wouldn't they?
Yeah. And and so now now that's not gonna say they're not going to get built, but the just the it's one of those things like I'm in tech and there's aspects of this industry that are just so disgusting to me and have driven you crazy for years. Um well threshold question in all this because I do want to get into the implications for a a variety of players in this market but uh how durable and widespread do you think the backlash will ultimately be here because to me that remains something of an open question.
It is an open question. I think some things sort of come along and and then and then they and then they go away and then it's not not being... there's definitely been a mimetic quality to the backlash and the opposition and the misinformation around data centers and movements that have grown with that mimetic quality over the last 10 years or so generally tend to peter out over the course of two or three years but these are a pretty important two or three year stretch. Well, that's the question is is this is this a data center question or is this an anti-tech manifestation, right?
Because I think it's probably going to be pretty real for a while.
I don't think that's going away. That's why like I concluded the piece by saying, look, this is going to be on the ballot going forward. So, I do think this is going to be an ongoing an ongoing thing and it is going to your point, what does the industry learn? the industry. The sad fact of the matter is I maybe there was like some sort of fantastical past where corporate America did things out because it was the right thing to do, right? Multiple shareholders, our workers matter, our communities matter, like all like we're sort of all in this together. Mhm. And one of my big critiques of America generally, corporate America, I think this really started on Wall Street, even going back to the '80s, was the reduction of morality to legality.
If it is legal, it is it is not okay. That's right. And the re and where I've come at that point before is in the the problem with this of not doing things because it's the right thing to do is you're forcing more laws and regulations to be made. Because if if you don't like just naturally do the right thing and you're always finding the loophole and cutting corners, you have to regulate everything. Absolutely. And once you regulate everything, the problem is that is severely limiting on your upside because basically you're forcing regulation to curtail these downside outcomes.
But when you cut off the downside, you cut off the upside room for innovation and ingenuity. So that's why it's so important. There's another angle on this here where you know there was a world where the tech companies would have approached data centers not with the um it's just been very like what's been the big strategy go back to like say the remember the Amazon headquarter thing.
They actually have a building to that and that's all by you isn't it?
Uh I think they built some of it but not all of it. It was it was embarrassing for the tech industry that whole episode.
So this is the episode where Amazon basically like auctions off its alternate headquarters to various markets around the country.
Yeah. And like all these cities like who's going to give us the sweetest deal? Sure. It was like one of the most powerful companies in the world is just basically saying we have a lot of economic impact taxes. We want all kinds of subsidies. Sure. And it was shameless like that entire episode and that has characterized all of this stuff.
Everyone just assumed, oh, who wants to pay up for the data centers? Like, you know, it was baked into the business models of all these companies.
Absolutely. And the right thing to do would have been to not be so shameless, to realize this is like we to have some degree of awareness of how normal people might think in the full context of our messaging, people's overall increasing discomfort with tech, with like social media, with all the sort of things that we've talked about and to realize we need to get ahead of this, okay? And we need to be making the case about the industry as a whole and we need to be making the case to local communities about how the these benefit you instead of going to local communities and saying right up shaking them down or we'll go somewhere else. So that's exactly what we want.
And so what we're going to end up in is fine big tech. The community shake down the big tech companies. You want a transactional model, you're going to get a transactional model good and hard and you're going to end up paying people off. That's what's going to happen.
And the question is, will that work? Because there are still going to be community members who oppose these on philosophical grounds or because they just resent tech generally, resent what tech has done to modern society. I think it's interesting going forward because you mentioned there were bipartisan campaign ads that you saw.
Well, that was the crazy thing. There were so many data center ads I and both sides are trying to outflank one another and be more anti-data center.
It was not like the the like you would think zooming out like okay the Democrats are nailing the Republicans on this. Brooklyn's generally pro business etc etc. And I think in general that was the overall 6 months ago that's where we were and then the the Republicans were like holy crap people really don't like data centers we don't either.
There was one race that was in the Madison market for I don't even know who it was and every single they would go back to back of them attacking each other for data center support like they would just like manufacture means by which they somehow had supported a data center at some point in the past. It was crazy. And I like the whole time in the runup like I don't this is some life advice. I don't pay hardly any attention to the election.
Once the election happens, then I will look at the results, think about what it means, and analyze it because guess what? The election's going to happen no matter what. You don't have control of the election. I maybe the polls are predictive or maybe they're not. Maybe the prediction markets are on it. Maybe they're not. I'm just going to say we have some people in our lives who are living in the cross spent a lot of time that has basically no fam. But hey, maybe it's entertaining for them. Well, fair enough. But the one thing that was in my head early on is like, okay, this data center thing is unbelievable. I cannot believe how many ads they are. We talked about it before and it's just it's clearly way more potent. And again, the fact that there was just there was no affirmative message for data centers nowhere. Everyone is sprinting away from them as that was a losing message this cycle probably will be next cycle by the time we get to 2030 I don't know whether there's still going to be this like massive outrage about data centers and what interests me is there's a difference between campaign incentives and governing incentives and so over the past couple months it wasn't surprising to me to see both parties fall all over themselves to be more opposed to data centers than the uh opposition like even Greg Abbott who had been supportive of data centers was cowed into opposing data centers in Texas. Um but well that is a big I mentioned the globalization angle before at the end of the day call me a big tech shell a data center is great for a community.
That's the thing and like these are basically the cleanest and most profitable industrial projects in American history.
Well, so this but this is a good example. So people would, including I think I wrote this at some point because it is true. Point to Lord Lorden County.
Lowden County. You screwed this up the first time we talked about Lowden County. Lowden County in Virginia.
Len county of Virginia. Richest county in the in the country. Property taxes have gone down. I didn't even know that was possible.
I believe power prices have gone down. Like every part of already the rich county has gotten better because of data centers. Now why are data centers there? This actually goes back historically to uh there was a major like interconnect there from when the internet was built. It might have even been pre-internet but because DC is actually is downstream from DC being close and so AOL started there and just in general that was clear and if you were built there you were relatively close to Europe. You were the most populous part of the US was in the east coast. It's like the most natural and there was lots of power available. It was the most natural place to start building data centers and you got this virtuous cycle where you got more and more data centers. You got more and more power to serve the data centers. You got more and more connectivity and you end up with like this it's a great...
And you have an area that knows how to build data centers because of the industry that has sprung up.
There's just it's an entire ecosystem, right? It's it's like uh what you see in China around these manufacturing sites. Just everything sort of feeds on itself until there's nowhere else that's competitive. And it's still the largest place in the world. And by the way, it's phenomenal for everyone that lives closer to there. They don't pollute that. You have tons of power. Your power prices are lower. You have amazing like city parks and services and schools, right? And you don't pay anything in property taxes. It's like it's unbelievable.
It's not a bad deal. And so, relatively speaking, if you're governing one of these communities or states and watching these projects go somewhere else, there's going to be an interesting tension there because jobs are on the line. But so so big tech would look there's a tendency when you're looking at things at scale to look at it like at the end state.
But you have to get to the end state Louden County.
Uh the the data center started being built there 30 years ago.
So yes, you're seeing the outcome of a 30-year virtuous cycle. In the meantime, if you build a data center and you're not power neutral and everything and you're plugging into the grid, guess what? Electrical prices are going to go up. Now, electric prices go up, more demand. They're going to build new power generation. So, in the long run, it will pay off, but there needs to be more cognizance of you can't sell the long run if the downsides for people who can vote against your project, right? And again, not just not mitigating, but demanding to be paid for the privilege of running up people's electrical, right?
Correct. And I I would guess that the industry will be able to sort of accept that as a baseline for any of these projects going forward. Even if they do that, I don't know that that's a guarantee that the bottleneck will be removed in a place like Texas. We'll have to wait and see. For the sake of argument, though, let's assume that the bottleneck does remain in place over the next two or three years. And again pretty critical two to three years as far as the buildout is concerned and the compute demand...
Maybe so a lot of the stuff so even in Texas the moratorium is on new permitting for new there's a lot of stuff under construction there's like I believe 130 projects under construction right which so 20 2027 2028 all that stuff is still coming online I think where the concern is is more about 2029 going forward. So there's a few things that might happen. Uh number one, going abroad.
Like that is unfortunately something that is probably likely to happen. I think like there's just a huge big announcement of a new project in Malaysia by by Google. I think 15 billion for a data center in Malaysia.
Yeah. And and so you're going to get tech is shameless. Uh remember tech is the industry that started globalization getting Intel shipping chips to Hong Kong for for test and assembly and um and so I'm sorry you're not going to you're not going to get tech will do the right thing when it is no other choice to do it. They're going to...
Well these are multinational corporations. That's the way they operate. Sure.
And so yes for sure going abroad is going to be one thing and playing countries off each other. So the, you know, that is that should be a big concern. Uh, obviously number one. Number two, uh, this was the whole thesis behind the SpaceX thing. Like I'm like, we're going to outer space now. We have to...
Or might have to. I don't know. It's in play. I mean, this was always I think this was this was always a possibility like because even if you didn't have this backlash, well, number one, when I wrote about SpaceX and their S1, it was when this backlash is already building. So, I sort of built that in with from my analysis from the beginning. Number one, these are more viable than you think. It doesn't mean they're easy.
Can we go cliff notes on that in terms of the viability of space data centers?
Well, I think the the big thing is people have in their head you're going to have big warehouses floating in space. And basically what you will have is like individual wrecks, which are like an NVL72 is about the size of a Starlink satellite today, like the the uh the current ones, not even the big ones that are adding all the cellular capability.
So, a bunch of individual racks and redundancy built into the system.
Well, they well I mean the challenge is cooling. So you have to they have to have radiators um that you know that and the amount of cooling is like an exponential increase. They basically have these massive radiators going into space to dispel spell cooling. Like all this is like fairly insane and nuts and but you already have Starlink up there. You already have the the the laser-based communication for them to send data back and forth. You're going to have in the long run the shift to inference that happens independent of humans like and so the the just the...
So inference that's serving agents and time doesn't matter.
Right or ma matters less like you don't need sort of the immediate response basically the pieces are there it's just like why would you do this if you can just build a data center...
You have a lot of land in this country it's a fair question...
Or in the UAE or wherever.
I mean you had this escalation in bottlenecks and the bottlenecks get solved right so we were short on chips and make a bunch more chips oh now we're short on memory that will be resolved eventually uh then you're short on we're going to be short on power and by the way I do think we're far enough far enough along in power projects that when and if this bubble bursts if where the maybe the bursting's happening right now as we're sitting here who knows that that was always the question what's the long-term payoff of the bubble what stuff gets built that is like economically nuts but is actually worth it in the long run.
And we put our power generation if we end up like we've been living in power poverty for 50 years like since the oil crisis basically. You know it's funny people go back and they're like oh do you remember the 50s and 60s when America was dynamic and growing.
We treated energy as completely abundant. It wasn't a consideration for anything. Just build all the factories.
Build the beautiful long Cadillacs that you know that are like 15 or 20 feet long the boats. Absolutely. Right. Um build the the the crazy like overpowered engines like...
Why no one cared about efficiency, right? And you it's funny like that sounds scandalous to anyone today cuz we've all grown up in a world of efficiency and like you know...
Do more. Yes. And it is I think it's a it's been such a limiter in what is possible and I think it's been a limiter on imaginations like what does it actually mean to have infinite power? This is why the nuclear thing is such a tragedy that you know and we should have had infinite clean power at this point. Um, but if what happens is the bubble bursts, but we actually brought to bear tons and tons of new power, that is a long-term payoff that is going to be hard to calculate like what what what it was worth, right?
Um, and we're on the way to bringing on some of that surplus power. Correct.
For sure. There's been a ton I there's been a ton of new power projects and I think like more of this stuff, to your point, will get over the wine. This is the difference with globalization. Globalization, the factory leaves, there's no upside. Right. Right. Again, there was no veto, but there was also no upside. In this case, there is a veto, and that's what we're dealing with right now. There is an upside. And if the demand is there, there are enough people who will be incentivized to get some of these projects across the line that I don't really worry about.
If you want to get a data center built, put checks, put money in people's pockets.
Exactly. Like just the... and if that's what it takes that's what will happen.
On the other hand, uh the space example, I wanted you to give a cliffnotes of that because like when I first heard about that, I was envisioning like a data center in space and that's that concept sounds completely insane, would be completely in impossible to maintain at that level. But the more atomized version that you lay out there where you're talking about individual racks and the capacity to launch all those.
Yeah. Well, that's the thing that so like with Starling. So you we talked about the Craig Moffin interview where he's like why does Elon want to get into the cellular game which is a crappy business.
And that is correct if you're starting from like zero. Like we're going to build launch capacity that is gonna let us deploy satellites so we can be a cell phone company.
Crazy thing to attempt. Sure.
The launch capacity exists. It's it is coming regardless. And there is a lot about Starlink where there's you almost have to think uneconomically in that if you think economically, you should think about the cost of all the inputs and figure out does it make sense. From a Starling perspective, it's kind of like launch is free. Yeah. Because it's not really free. That's not correct. For anyone else who wanted to do this, it would be insane to attempt that capacity.
They built out this capacity and they need to use it for something, right? I mean, this is how you get uh this is how you get uh say like in the shipping industry, you'll have shipping companies that will run losses for years.
And it's like, why are like how do you operate at a loss? Well, it's an accounting loss because they're they are shipping at a they don't ship. If the marginal costs aren't covered, they don't ship. What are the marginal costs? Fuel, port fees, crew, all of which pale in comparison to the depreciation cost of like the ship itself. So maybe their marginal cost to ship a container is like $1,000. What that means is they will if someone will pay $1,100 to ship a container, they'll ship the container even though the cost on their accounting balance sheet is actually shipping that container cost $2,000.
But $1,000 of that cost is depreciation, which is like it's not fake. That's real money that went out the door, but the that money's been spent.
Yeah. So, as long as you're making anything incremental over your marginal costs, you will continue to operate even though from an accounting perspective, you operate a loss. This is a super important principle to understand because it it explains why companies will seem to operate at losses for ages. It's like why don't you just shut down or how you were surviving? Well, their cash flow is still positive. Um, now again, where they what happens is they end up not having enough money to service their debt and then they go bankrupt. That's how bankruptcy happens. Cuz to buy these assets up front, you take out loans cuz you're going to have a 30-year useful life. So ideally, you're paying it back over 30 years, right? Yeah. But so the just the way money moves around and and how bankruptcies happen, how companies operate at at paper losses for ages, but because what happens with the shipping industry, you get a period like COVID where containers are now like $17,000 a container. You're making so much money that like this is the crazy about commodity markets. This is what's...
happening in memory right now, right? Same same general sort of concept.
And so the reason why and by then it takes too long for anybody else to break.
Oh, yeah. No. Yeah. Buy online, right? And so people will will wow what an opportunity. Why you buy a bunch of ships and then it crashes and then like and then it's great for the incumbents because they crash, they go bankrupt, suddenly cheap ships are available.
[laughter] Yeah. Who wants to buy new ships?
So um the uh and that's why in commodity markets like I admire people in commodity markets because the way to succeed it is the ultimate sell the top buy the low sort of business. You need to be in the muck. Your business bleeding money. That's the time to invest because then you'll be ready when it comes back, right? Uh if you wait till it comes back, it's too late. But the the in these markets, the fixed costs are so high. Like buying a ship is really expensive. Building a fab is so expensive. It's like you and that's why you see like TSMC, we've talked about it. They don't want to expand too much because they're the one holding the risk. That's right. Uh and so where does this tie tie tie to SpaceX? It's not a perfect I'm not saying this is an accounting thing, but there is almost a mentality thing where look, we've built launch capacity. Starship is coming. Starship can launch much larger satellites.
It can launch a ton of them. What's this going to be used for?
Well, [laughter] no, but it's like, well, suddenly you start thinking about data centers in space from a marginal cost perspective. Like how much to but that marginal cost is like building the space rack, deploying it, operating it and you kind of treat the launch capacity not just the launch like not just the overall cost of developing the Starship but also the launching itself because why you want to launch more so you keep learning, keep improving like Musk companies in general are usually very good about their learning functions like and so continue going on you kind of treat that as a given it's free.
Now, it's not free to be clear. I'm not saying this is an accounting calculation, but the reason why you if you're Elon, you're not thinking about building a cellular service and fully internalizing all the costs. You're sort of this whole cost, which is actually the biggest cost. We're going to pretend it doesn't exist.
Wow, what a big opportunity we have, right? Because there's we just they they because there's a drive to use up this capacity. And I think there's a similar thing with data centers in space. Like would you start from scratch, build a new rocket, get launch capacity or even would you consider the price of launch capacity? I mean Starlink looks very profitable in part because they just pay the bare cost. There's no premium charge to them by so SpaceX itself is actually much more profitable and Starlink is much less profitable in a true arms-length sort of calculation of the cost. But all this is to say is like Starlink and SpaceX is the ultimate hammer looking for a nail company. Yeah, they've developed this hammer, incredibly valuable hammer, which is this launch capability, regular cadences, tons of stuff. They need stuff to put in space.
Well, that's why I said we'll see in terms of how exactly they monetize this. And look, data centers in space sounds less crazy today than it did 3 days ago. Well, [clears throat] wrote that a few.
Okay. Do you want to pat yourself on the back [laughter] here?
No, but that that is why like a Verizon or an AT&T or T-Mobile should be nervous.
Because the most terrifying thing as a business is going up against competitor that seems to be acting irrational.
Right. It's irrational for Starlink to build a cellular service. Yeah. But in a weird, which I was trying to explain here, I'm not sure if I fully explained it in a weird sort of Elon thinking way, it is rational. If your overall goal is increasing launch capacity, increasing like building on the learning function, I want to get to Mars.
I was going to say his overall goal is going to Mars at the end of all this.
That's right. And so, um, yeah. So, and I think that the stuff in space makes sense in that regard. And uh and by the way just from a purely strategic asset you can see the value of that. Uh absolutely. So yeah uh well I have more questions on specific that said most of the data are going to be built on land [laughter] especially for the foreseeable future.
Yeah I that's more of like a 5 to 10 to 15 year solution as opposed to the next couple years when the crunch is coming.
I don't know that's going to be like like there's still be a lot of data centers on Earth for now.
Today's show is brought to you by Stripe, the economic infrastructure for AI and for strategy itself and a company that processes digital payments for millions of businesses around the world. I learned that on strategy this week that the total revenue processed by Stripe is equivalent to 1.7% of global GDP. Just an astounding figure. Uh we're going to be spending more time in the next couple months going into some of the Stripe data. Ben, you spent a lot of time analyzing individual companies and industries. Have you spent any time with the Stripe data thus far?
I have not. Uh, full disclosure, the I mean, I think the Stripe blog that writes about some stripe economics. Yes. Well, it's kind of funny because one of the challenges in writing about the stock market, which you know, again, um, you do a little bit of that. No, I don't actually. Okay. It's actually it's pretty interesting. There's almost a challenge this week where my assumption was there was going to be a big drop uh because of these things again it was weird it didn't happen on Wednesday um then the Democrats decided to be like uh let's investigate all the corporate leaders and ban data [laughter] centers everywhere it's like it's almost like they're like I saw that as I was getting on the plane literally 24 hours of moderation and now here we are it's like yeah um it's like they weren't counting on the stock market drop so they're like it needs a little nudge. Let's give it a boot over the edge. Um, but so I assumed that would happen. But what I wanted to write about like what are the implica what does it mean if there's real friction in data center building like and so that was the 10 takes of of you know good for the hyperscalers I think moderately good for the frontier labs you we can sort of keep talking about it but also there can be such a disconnect between what's happening on the ground and what's happening in the market which is sort of anticipating things. So this week the little factoid was you know like what this spring everyone's dead. That's right. [laughter] Which by the way uh so that's happening. Meanwhile Stripe's like looking at their data all these small companies actually accelerating like the revenues are increasing including new companies which I thought was interesting in the Stripe notes right and so um that's not to say the stock market has it wrong. Like I actually think for the like long-term pressure on well there's a terminal value question like I think like does the stuff go to zero there's also like how much growth is there but it is interesting to see when you get a widespread view that you would think my assumption would be actually always stuff will be static growth's going to go away but the fact that it's actually been growing um the amount that that companies are spending on SaaS uh it just it gives you pause and it's like um because it's very easy to get in your head about theoretically what makes sense and is going to happen and then reality can sort of be like well some confounding factors here still doing fairly well uh so it's interesting so Stripe does get a good view of that um I am looking forward to getting more sort of I am too that's sort of interesting it offers kind of a complimentary view of what's going on in the economy beyond like earnings reports and for it's actually funny for me it's almost like it's a little like uncomfortable it's like [laughter] Um, it's more fun being in the theoretical data world. That's right. [laughter] This is what's actually here's some hardcore data. Uh, yes. Uh, economics. It's a cool Substack. It aggregates all the macro data in a bunch of interesting ways. So, uh, in any event, we'll put a link in the show notes if people want to check it out.
Uh, back to the show. As far as specific companies, it is sort of an ironic twist based on what you wrote this week. Considering much of the opposition to AI data centers is rooted in like deep abiding resentment for big tech companies, big tech companies may in fact be the beneficiaries of a lot of this opposition. Is that right?
I I think so. Well, you know, when you think about the uh they always win. They do always win. [laughter] It's amazing. Um, and like where does the where does value accrue, right? If you have the stack, you know, I think if there it comes from integration like like when you have different pieces you can put together and what things can you integrate in the AI stack. So you on one hand you could say like the frontier labs right now they want to be on the leading edge but you could say you know is that a sustainable business? What if you're easily substituted out? If another model come along that's better. That's a tough place to be. Yeah. So, one argument is they need to go up into more consumer-facing applications. So, you're integrating like the consumer-facing service with the model itself and then that's how you win is. So, you fuse that way. Another thing is maybe the frontier companies want to have the leading edge model and they're going to build their own data centers and design their own chips that are like meant for their model and they're going to be able to serve much more efficiently than anyone else. And so that's another point of integration that sort of could happen. Uh but one of the challenges is they need money to do this. Y number one. Uh number two they need to build new data centers. They don't have an existing footprint like Anthropic is talking about it. They have this deal with Broadcom to get chips. Those all need to be built.
Mhm. Meanwhile, you have the big tech companies which have a ton of data centers already. Uh they still have the best access to capital.
Can I ask you one question before you go on? The big tech companies with all their data centers that already exist, how many of those legacy data centers could be converted into AI data centers?
Okay. So, but the issue so because previous data centers were by and large air cooled um whereas modern GPUs need uh water cooling. I think modern GPU are much much denser. Uh whereas you have like more sort of spread out uh because you're using air cooling. You're trying to build it for the lowest cost possible. There's a lot of different assumptions and you kind of if you want to take an old data center and convert it's a significant you're basically rebuilding it.
Yeah. You just knock it down,
but you already have the land. You already have the power. You already have like or in just you already have the relationships, you already have the the the community is used to you being there. And so there's I think you're just starting further ahead than trying to get a net new data center somewhere else.
Yeah. Like I read uh a meta spokesperson this week said in lieu of building new data centers, one of the things we're going to be doing is expanding the data centers that we already have to meet some of this demand. Um all of which is much easier to do if you're a meta or you're an Amazon,
right? So they already have lots of assets. Number two, you have the financing ability, right? They have free cash flow. Now that free cash flow, they don't have free cash flow because it's all being consumed by the data center.
No, but the point is if you're spending what is Amazon, what was Amazon's thing like $200 billion on capex and you're Wow, that exceeds their free cash flow of $180 billion.
That's still $180 billion, right? [laughter] Like so we're talking about a bunch billion dollars ahead of anthropic or open AI who have no they have negative free like or in the fullness of time.
So the um so number one, they just have this huge base. Number two, they get better credit because they have other businesses that generate all that free cash flow. So their ability to borrow their their price to borrow is much lower. And that's why you end up with these deals where they're basically financing the like they get these great deals with Anthropic like Amazon does because they can finance them. Even if Amazon has to ultimately finance it themselves, it's like an arbitrage opportunity because my ability to finance is cheaper than yours. Sure. Um this ties into Nvidia and the Neo clouds and all that sort of thing as well. And then number three, they can issue equity and if you're an anthropic going needs to get to market again, they need to get to market. We'll see if that actually happens. They which I think it will, but might be a little hairy. Uh all their equity is 100% on will anthropic succeed or not.
When Amazon issues equity, that equity, some amount of it is will this capex be worth it?
But it's also there is an existing business that is very successful that you're getting a piece of. So of course it's going to be damaging to existing shareholders because you're diluting them and that money is going to this speculative thing that may or may not work out. But by and large the sort of the support for their equity is much more than it will be for anyone else.
Yeah. And that that's the funding of last resort, right? You just start giving away parts of the company to get more money, which is we're in that stage. Everyone is issuing equity at this point. So, but in all those stages, better free more free cash flow, cheaper credit, more ability to issue equity. The hyperscalers are just in a better spot than anyone else.
And they have more capacity build data centers. Number they're going to be better at lobbying. They're going to be better at like sort of getting these things over the finish line. And if you think about that, they are those aspects might end up mattering most. I think they're if data centers are going to be more expensive because you have to pay off the local population. Sure. Like they're in a better place to fund that to sort of to pay for it. Um, and yeah, big tech I think will be
I mean, we talked about it last week in terms of Anthropic and some of the pressures on them going forward, some of the equity deals that they did in the past in exchange for compute and what those could look like a couple years down the line if Anthropic can't pay for the compute all by itself. Um I wonder like in a compute constrained environment where you've got like Amazon and Google that have all this compute does anthropic do open AI do they just go hat in hand to some of these hyperscalers?
There's a good chance. I mean, I think one of the scenario opening eye and anthropic I don't think will ever go away.
The question is who are the equity holders? That's right. [laughter]
It's like does Google just own anthropic at some point?
Yeah. Amazon. Yeah. It's like a Google Amazon joint. I because there's
Yeah. the and it's funny because Amazon I mean Andy Jassy people were hating hard on Andy Jassy and Amazon uh after the earnings call a couple weeks ago and now like they are sitting pretty in a world where no one else can build data center capacity.
Well, they're also sitting pretty in a world where everyone has to kind of slow down.
That's the other bit like you've had this they will all tell you, oh don't worry, we have buyers lined up, etc., etc. That's always the case until I this is like when you have like all the crazy stock market things um you know the uh like 2008 everyone's like on paper don't worry there's a back stop there's a guaranteed the problem is what happens when the back stop goes away and you get this chain reaction you could see this risk still coming up right like where uh yes we have a guaranteed buyer that buyer's opening eye [laughter] you go down the chain all these sort of Um the uh it's so so yeah but they had to compete right what if Amazon doesn't invest then well Google's going to invest and take more and more and Microsoft sort of kind of briefly stepped on the brakes slightly and then suddenly we didn't have enough compute and it was a big problem and so everyone's like I can't slow down if they're not slowing down. You have the same we've talked about this dynamic with the labs themselves. This is the problem with Anthropics S1. The fact their inference looked so amazingly cheap should actually terrify you because the part they have to keep spending a lot of money.
That's right. Is and it's out of their control. If they don't spend to be on the leading edge, OpenAI will and they'll lose their position and uh or Google or Meta or sort of whoever whoever it might be. And in a world where everyone take a breath amidst this murderous capex buildout here, the shareholders love it because oh, we had like a bunch of projects canceled. What a bummer. And so our capex is going to be flat or you can you can imagine like an earnings call like this, right?
And everyone sort of announces this. Investors are like, "Thank goodness." [laughter] Yeah.
Suddenly their stock goes back up. Now they can issue more stock uh to get more money. They can like that that like that currency is more valuable. Maybe they're the credit like if the the big question is when does the revenue come?
If the once the revenue comes all these things unwind themselves in a very positive direction. And so if there is an external force that forces a slowdown such that revenue can get there in time to relieve all these capital pressures. Everyone could win. This could actually be a diffusing win win. Sure. But especially the hyperscalers and [laughter] so uh and so yeah uh it it's yeah everyone thinking about AI jumps to the final state where intelligence is a commodity.
And there that state might take a very that implies infinite compute.
Yeah. And it just might be the longer it takes to get there, the more analysis predicated on intelligent being a commodity isn't actually useful to understanding the world as it is.
In a world where there's infinite power, infinite compute, what is like what does Nvidia do? Like there's going to be other chip products, right? In a world where you're limited in power, you have to repurpose your data centers. Suddenly whatever margin you have to pay to get the most efficient most powerful chips it suddenly makes a lot more sense right and so it's this the analysis in a market defined by scarcity versus a market defined by abundance is fundamentally different completely flips right and and I think the the time defined by scarcity has probably been elongated
okay well it'll be interesting to see whether there is ever a capex slowdown because at this point I'll believe it when I see it. Whenever we get there, it will be fun to look back at some of the capex we've seen over the last couple years and just how consistently insane it was. Like I think there was a stat where a couple of these companies were spending more combined than the Pentagon spends every year and that's just where we've been living for quite some time now. So we'll see what it turns into. Uh Nvidia, you just mentioned them. How should we think about Nvidia in a world where US data center construction is constrained over the next several years?
Well, in the short term, it's definitely a problem because they're, you know, they're selling chips. There's nowhere to put them.
Um, you know, is their market being limited. On the other hand, overall, and this has always been a counterintuitive point about Nvidia, which is the one I just made, a world defined by insufficient power or insufficient data center capacity is the entity that delivers the most tokens per watt.
The best machine, the most efficient machine will be worth the upcharge. [clears throat]
Exactly. Because at the end of the day, the tokens are going to be more valuable. This is also a weird thing with the Neoclouds like all the existing NeoClouds out there are now if we reduce the rate that capacity comes online all those are more valuable right and but at the same time and so it's kind of weird to think through what this means what this means for the Neo clouds uh you know can they get they're going to have a harder time getting new projects started coming to market are they ever going to have sufficient scale but what they do have might be more valuable um and so I think in this world I think there's a lot of these things where do you mean better absolutely or better relatively okay right like uh on an absolute basis uh maybe this is worse for Nvidia they'll sell fewer chips on a relative basis as opposed [clears throat] chips like maybe this is better well and a future where compute is abundant is probably not great for Nvidia.
It's going to be challenging. Um like it's hard to sustain margins uh in a world and people inevitably come for their margins in that scenario,
right? Well, I mean you you think about you know why why was Nvidia so gung-ho about China because China is not a country that's defined by lack of power, right? And so I if they don't keep people locked in their ecosystem using their soft using CUDA defaulting to their chips over time it's like well I'll just build inefficient data centers more chips and not pay the Nvidia fee which is I think they've resigned themselves in the long run more coal power over there in China.
Uh well that was going to be my next question. So uh how should we think about the competition with China in a world where data center construction is constrained?
I mean it's a problem. I mean we in the we have a huge chip advantage. Obviously we've created the conditions where China has to figure out the chip thing on their side. When and if they do, they're going to have a lot more ability, I suspect, to build more data centers and more power than we will.
And in the fullness of time, how does this not look like the manufacturing world where they're just like have so much more scale and are so much cheaper that they actually end up owning all the all the compute? So they will have far more compute than the American side in the fullness of time. Even if they're using less efficient chips that are inferior in a lot of respects, well they'll move down the learning curve. Their chips will get better and better over time. Okay? Right? And and so but again, this is a huge timing question, right? Like the fullness of time. Everything's possible in the fullness of time. We're [laughter] going to move to the dorm room, right? We should have a dorm room theme when we want to have the dorm room.
Honestly, it's tricky for me because as the host of Sharp China who's reading news about China every week, I personally have never been less concerned about the competition with China between the United States and China. And so, China's always like invoked as a stalking horse in these arguments. And everybody this week is like, well, voters just handed the future to China. Um, and I don't know that I disagree with that because obviously China will have advantages that it can bring to bear over time here and we're gonna build data centers. I'm not that panicked about it. [laughter] So I think this is more this is a build another data center in the United States. Then we can have a conversation about
Right. Exactly. You know what? You know what though? This is why I don't like think about politics.
This is like I'm like sick of this. [laughter] Wrote an article. We've talked about it. Uh,
fair enough. Uh, well, final question then because we can move on to Google for you. Uh, Zach says, "When people discuss building data data,
by the way, maybe I'm the that that is evidence of tech disdain for normal people. [laughter] I'm the problem that I'm identifying.
