# SpaceX has no competition — Saurav Shroff, Starpath

Markdown copy of https://realalpha.app/v/QEACSjIYPrg/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

Saurav Shroff argues that launch competition is essential to make space-based AI compute accessible, and says Starpath may build a launch company if supply stays closed.

- Show: Relentless
- Hosts: [Ti Morse](https://realalpha.app/people/ti-morse/)
- Guest: [Saurav Shroff](https://realalpha.app/people/saurav-shroff/) is founder and CEO of Starpath.
- Published: Oct 28, 2026
- Length: 58 min

## The best minutes

- [8:56](https://realalpha.app/v/QEACSjIYPrg/?t=536) [AI data satellites need cheaper launch and much greater scale](https://realalpha.app/s/ai-data-satellites-need-cheaper-launch-and-much-greater/) (6 min)
- [21:23](https://realalpha.app/v/QEACSjIYPrg/?t=1283) [Starpath rebuilt space solar panels from a clean sheet](https://realalpha.app/s/starpath-rebuilt-space-solar-panels-from-a-clean-sheet/) (8 min)
- [30:36](https://realalpha.app/v/QEACSjIYPrg/?t=1836) [Starpath builds factory capacity ahead of customer demand](https://realalpha.app/s/starpath-builds-factory-capacity-ahead-of-customer-demand/) (6 min)

## Also worth watching

- [0:05](https://realalpha.app/v/QEACSjIYPrg/?t=5) [A launch monopoly has little incentive to sell capacity](https://realalpha.app/s/a-launch-monopoly-has-little-incentive-to-sell-capacity/) (7 min)
- [15:18](https://realalpha.app/v/QEACSjIYPrg/?t=918) [Starpath may build its own launch company](https://realalpha.app/s/starpath-may-build-its-own-launch-company/) (6 min)
- [48:39](https://realalpha.app/v/QEACSjIYPrg/?t=2919) [AI revenue can make rocket development economically rational](https://realalpha.app/s/ai-revenue-can-make-rocket-development-economically-rational/) (5 min)
- [53:56](https://realalpha.app/v/QEACSjIYPrg/?t=3236) [Mars may need far fewer flights than AI data satellites](https://realalpha.app/s/mars-may-need-far-fewer-flights-than-ai-data-satellites/) (4 min)
- [39:55](https://realalpha.app/v/QEACSjIYPrg/?t=2395) [Starpath is filling supply gaps for AI satellites](https://realalpha.app/s/starpath-is-filling-supply-gaps-for-ai-satellites/) (6 min)

Skip the Austin factory-scale-up and sleeping-at-the-factory discussion, plus the moon-tourism price estimates.

## Full analysis: Saurav Shroff says Falcon 9 misses AI's cost target fivefold

- Shroff estimates Falcon 9 costs SpaceX about $1,000 per kilogram against a roughly $200 target.
- He puts the cost gap at fivefold and the scale shortfall at tenfold, with launch access still unresolved.
- Founders should separate factory forecasts from orders and verify launch offers.

Its 6 sections, what it means for you and what to watch are on [the page](https://realalpha.app/v/QEACSjIYPrg/#analysis), free with an account:

- [11:47](https://realalpha.app/v/QEACSjIYPrg/?t=707) Falcon 9 misses the stated AI-satellite hurdle
- [6:02](https://realalpha.app/v/QEACSjIYPrg/?t=362) Launch access may stay closed even when rockets work
- [15:27](https://realalpha.app/v/QEACSjIYPrg/?t=927) Shroff may build a launch company if supply stays closed
- [27:20](https://realalpha.app/v/QEACSjIYPrg/?t=1640) Starpath rebuilt its panel around lower costs
- [33:31](https://realalpha.app/v/QEACSjIYPrg/?t=2011) Capacity reassures customers but raises capital risk
- [52:35](https://realalpha.app/v/QEACSjIYPrg/?t=3155) Shroff says AI demand could justify rocket investment

## Summary

Saurav Shroff is the founder and CEO of Starpath, a Los Angeles-based startup focused on space solar and AI data satellites. Starpath supplies mass-produced satellite components, including its ultra-thin space solar panel, and is building manufacturing capacity for future demand.

Shroff argues that launch access, not merely rocket performance, may constrain space-based AI compute: providers can have capacity but no incentive to sell it, and Starpath may build a launch company if the market does not open. He says AI satellites will require fully reusable rockets because Falcon 9 falls short on both cost and launch scale. Shroff says Starpath redesigned its space solar panels from first principles and is expanding factory capacity and component manufacturing ahead of customer demand. He sees AI infrastructure as a potential economic driver for wider space development, including lunar tourism and a Mars settlement.

### Minute by minute

- [0:05](https://realalpha.app/v/QEACSjIYPrg/?t=5) **Launch supply remains scarce.** Shroff says SpaceX is the only launch company currently serving the market at meaningful scale, with Falcon 9 supporting Starlink and Starship aimed at future spacecraft. He says other providers are still catching up to Falcon 9 or lack a rocket, leaving launch services unavailable to many customers.
- [4:50](https://realalpha.app/v/QEACSjIYPrg/?t=290) **Shroff sees little incentive to sell.** Shroff argues SpaceX has little economic reason to sell launch if it can use Starship to deploy its own AI satellites, given his view that demand for superintelligence will exceed supply. He sees two possible outcomes: new launch competition, or one company controlling superintelligence capacity without pressure to lower prices or improve quality.
- [6:58](https://realalpha.app/v/QEACSjIYPrg/?t=418) **AI satellites will reshape launch demand.** Shroff divides the market into a near-term phase, through roughly 2029 or 2030, when telecommunications satellites remain dominant, and a later phase when AI satellites dominate by size or power. He says Starpath can use Falcon 9 and other providers catching up to it in the near term, but sees the later phase as more uncertain.
- [8:56](https://realalpha.app/v/QEACSjIYPrg/?t=536) **AI satellites require cheaper launch.** Shroff describes an AI data satellite as a simple powered shell carrying compute, memory, solar panels, networking and control systems. To compete with ground data centers, he says it must cost about $5 per watt all-in, with roughly $1 per watt for launch; his 200-kilowatt example totals $1 million. He estimates Falcon 9 is 5x short on cost and 10x short on scale.
- [13:00](https://realalpha.app/v/QEACSjIYPrg/?t=780) **Reusable rockets face a selling dilemma.** Shroff says AI-satellite launch requires reusable first and second stages because expendable upper stages limit cost and flight rate. He identifies SpaceX and Stoke Space as the only two companies he sees seriously developing fully reusable, two-stage orbital rockets, but argues neither has an incentive to sell launch if it can deploy its own AI satellites.
- [15:18](https://realalpha.app/v/QEACSjIYPrg/?t=918) **Starpath may build a launch company.** Shroff says that if launch access does not improve—through SpaceX entering the market or multiple providers bidding on large contracts at roughly $200, $300 or $500 per kilogram—Starpath may build a launch company. He says the goal would be to help customers deploy space-based compute rather than leave superintelligence capacity with a few firms.
- [18:47](https://realalpha.app/v/QEACSjIYPrg/?t=1127) **Shroff revises his launch assumptions.** Shroff says he had assumed three months earlier that Starship would create near-infinite launch supply. A friend's questions led him to examine provider incentives; after asking rocket companies what they would sell and at what price, he concluded that supply could exist yet remain unavailable, and began sketching a rocket.
- [21:23](https://realalpha.app/v/QEACSjIYPrg/?t=1283) **Starpath rebuilds space solar.** Shroff says Starpath rebuilt its space solar panel from first principles rather than trying to cheapen legacy designs, which relied on rare materials and could cost $300 per watt, perhaps as much as $1,000. He describes Starpath's product as an ultra-thin, OLED-like panel made from common materials and affirms the host's claim that its cost is down 50 times.
- [30:36](https://realalpha.app/v/QEACSjIYPrg/?t=1836) **Factory capacity leads current demand.** Shroff says Starpath's 130,000-square-foot LA factory is expected to become a 50-megawatt-per-year production line in about five months. He says the total power of launched satellites is around 50 megawatts including Starlink and around 5 megawatts without it, while demand forecasts from companies that spend billions on data centers justify building ahead.
- [36:19](https://realalpha.app/v/QEACSjIYPrg/?t=2179) **Customers want capacity ahead of need.** Shroff says Starpath's next planned factory, Starpath Austin Giga, is designed for 1.2 gigawatts. He says customers want to see future capacity before they need it, and expects factory builds to remain difficult: hundreds of millions of dollars of equipment must work reliably.
- [39:55](https://realalpha.app/v/QEACSjIYPrg/?t=2395) **Starpath plans to supply more parts.** Shroff says roughly half the factory will make solar modules and half will make other, not-yet-public satellite products. He says existing suppliers may not be ready to deliver components such as thrusters at the low costs and high volumes AI satellite constellations require, so Starpath plans to expand beyond panels to help customers deliver.
- [46:56](https://realalpha.app/v/QEACSjIYPrg/?t=2816) **Tourism could help seed a moon base.** Shroff sees lunar tourism as an early source of demand, estimating that at $10 million per trip at least 100 people might pay, and at $1 million per trip thousands could. He says a long-term price could settle around half a million dollars per trip, potentially enough to support a base.
- [48:39](https://realalpha.app/v/QEACSjIYPrg/?t=2919) **AI demand could finance space infrastructure.** Shroff says AI compute demand could drive rapid rocket development, much as passenger and cargo demand drove aviation. He argues that the side benefits could be greater in space: a mature launch and satellite supply chain could lower the cost of lunar and Martian projects, while AI firms could justify rocket development through compute revenue.
- [53:56](https://realalpha.app/v/QEACSjIYPrg/?t=3236) **AI infrastructure could lower Mars costs.** Shroff contrasts the scale of the two efforts: he says an AI satellite constellation could require 100,000 Starship flights, while 10 Starships sent to Mars could have a profound historical impact. He argues AI demand could pay for rockets and build a mature supply chain, lowering the cost of a Mars city, though operations would still have to be paid for.

## Transcript

Relentless, hosted by Ti Morse, with Saurav Shroff.

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

[0:00](https://realalpha.app/v/QEACSjIYPrg/?t=0) Today, I'm sitting down with Saurav Shroff, the founder and CEO of StarPath. StarPath is going to be putting a lot of solar panels in space in a few short years here. And so I wanted to start this off on the state of the launch industry as it stands today with like SpaceX, Blue Origin, Rocket Lab, and a few others. So can you just walk me through that? So the state of the launch industry is pretty simple. There's one launch company. It's called SpaceX. They know how to launch rockets. that work for the present day. And they also know and are making a rocket that works for the application of tomorrow. So break that into pieces. Rockets today are used for basically one thing, which is deploying telecommunication satellites. The most common telecommunication satellite, and by extension, the most common satellite is SpaceX's Starlink, which services many millions of customers, maybe even tens of millions. I don't actually know a lot and is basically the only product in space that is used by humans broadly. There are some niche products that are used by a very small number of humans, but Starlink is the only product that is serviced from space that is used by humans broadly and paid for by humans broadly today. And the application

[1:20](https://realalpha.app/v/QEACSjIYPrg/?t=80) of uh the the servicing of uh telecommunication satellites requires on the order of a hundred rockets per year roughly it you know you could argue it requires a little bit more than that or maybe like a few hundred but on the order of 100 rockets here that are the size of spacex's falcon 9 and so spacex has been building for three or four years now hundreds of falcon 9 rockets here where the majority of them are for starlink um for those same five years there have been rocket companies that have seen that TAM, TAM of telecommunications, and said, wow, it's a big TAM. It's a $40 billion market, $100 billion market, and then tried to build rockets that are similar to Falcon 9. And most of them, basically none of them have succeeded so far, and some of them will probably succeed the next few years. But more importantly, the future applications of rockets will service telecommunication satellites and AI data satellites. And the ratio between the quantities of these two will be something like 0.1% or 1% and 99.9% or basically 99 and 1. Okay. And the type of rocket that can service the second category of demand, AI data satellites in space, by definition must be cheaper and of a much bigger

[2:44](https://realalpha.app/v/QEACSjIYPrg/?t=164) scale, which means that it has to be completely reusable. So basically to achieve the cheapness and the scale that is required to create a meaningfully sized constellation of AI data sats requires a fully reusable rocket. So, I mean, if you think about the launch market, zero competition, one player that can service the present slash old market, and that is steamrolling the world and racing towards a rocket that can service the next generation of spacecraft. Meanwhile, you have like 10-ish competitors, quote unquote, that are trying to catch up to the yesterday rocket and basically just one competitor that's also trying to build at the same time a rocket that can achieve the economics and scale of the future to service the future market. So why is the market bad? It sounds decent. Well, um actually nobody there are there are zero vendors um so you know to call it a market is actually a misnomer um right now if you try to buy launch from spacex they won't sell to you basically under any circumstances um they may sell you like sort of obscure products like they'll sell you launch to the moon or they'll sell you They have a new product called Starfall, which are great new products, but they will not sell you traditional launch services that are useful for either of these two applications, which collectively, one plus 99, account for 100% of satellites that humans will launch.

[4:15](https://realalpha.app/v/QEACSjIYPrg/?t=255) Meanwhile, the other companies are also not in the market. Some want to be, but have no rocket, and so are sort of a selling rocket launch service that doesn't exist. most are not even trying to match uh the capabilities of starship they're trying to match the capabilities of falcon um and so it might actually be the worst market of any kind because um it looks like there there used to be a market there used to be supply it looked like the supply was increasing but actually there is no supply um and no matter how much money you have you can't buy supply so um very bad so with this sort of thing if spacex is going to be mostly putting up AI data center sats in a few short years. I think Elon said that the cost of putting compute in space in like three years from the date of the Dwarkech podcast, it would actually be cheaper than here on Earth. Sounds about right. And if that's his view, and he believes that roughly there's like no, you know, demand limit on intelligence, why would he like ever sell launch to anyone else, even if it scales, you know, scales launch like crazy with Starship?

[5:23](https://realalpha.app/v/QEACSjIYPrg/?t=323) Yeah. And that's the perfect question to ask. If I owned a monopoly on a service and that service unlocked the scaling of what I personally and maybe others believe to be the most useful and valuable technology ever made. And the supply and demand relationship of that technology is that there is drastically more demand than supply, which is to say the prices are extremely high, the margins are extremely high, because there is more demand than supply.

[6:02](https://realalpha.app/v/QEACSjIYPrg/?t=362) **Saurav Shroff:** Any rational economist would say that it's dumb to sell launch services, the bottleneck, to anybody else.

[6:13](https://realalpha.app/v/QEACSjIYPrg/?t=373) And so actually the future that I see is there's really two possible futures. Future one is that we create competition. And I think this is the probable future. Actually, some very good reasons to believe that that's the case. And then future two is we continue to fail to create competition. And the result is that all of superintelligence capacity is owned by one company that isn't motivated to reduce price. And isn't motivated to increase quality. And, you know, if we just ask ourselves, this is like not some kind of like revolutionary question, which future is better? Just about every single person in the room, except for the holder of Monopoly, would say the one with competition.

[7:02](https://realalpha.app/v/QEACSjIYPrg/?t=422) If you had to think through your own business, if your entire business relies on the ability to launch your product to orbit hopefully in a few years right um how do you think through like basically solving that problem or working with one of the other providers so it kind of makes sense where you can scale at a reasonable rate i guess it's best to answer your question by breaking up the future into two relatively distinct time periods we're gonna call the first time period which is in the immediate future the period where ai data sets are not the majority and then we're going to call and we i think that's like roughly now till like 2029 or something or 2030 then we're going to call the period after that the period where ai sats are the majority and when i say the majority i mean the majority of satellites sent to space measured by their size or power okay the aggregate size of power of all the satellites during the first phase our business really has no problems as in SpaceX, before exiting the launch market, which they recently did, sold lots of launch using their current rocket, Falcon 9, to many companies, most of whom are our customers.

[8:11](https://realalpha.app/v/QEACSjIYPrg/?t=491) And there are other rocket launch companies who will be moderately successful in catching up to Falcon 9's capability, which is to say 10 to 100, well, I guess Falcon's higher than 100, but 10 to 100 flights a year, 10 tons per launch, and the economics of a rocket that is not fully reusable. and all of those things will be true over the next two to three years and during that period of time our products will continue to sell and continue to provide immense value to customers so we're not worried about this phase phase one phase two is where there's some reason to be concerned let's look at what an ai data set is why it's better than a terrestrial data center aka a data center on the ground and then what's required to make it good on the rocket side so ai data satellite very simple take a chip take a memory basically take a compute module not a memory take memory sorry take a chip and memory and put it inside of a satellite that's quite a bare-bones satellite that's extremely large and generates a huge amount of power but is very simple in that it doesn't have you know great pointing accuracy it doesn't have the ability it doesn't have very many unique capabilities. It's basically just a powered shell. So just a box that has a huge

[9:26](https://realalpha.app/v/QEACSjIYPrg/?t=566) solar panel, a tiny battery, some interconnection, you know, networking devices, some attitude control devices, a small thruster, but basically a very bare bone satellite. That's what an AI data satellite is. Roughly speaking, an AI data satellite needs to be about $5 per watt for it to be competitive with a data center on the ground, because the data center on the ground costs between $10 and $15 a watt and lasts a little bit longer. Okay, so if you want to make it better, it has to be about $5 a watt. And that's all in. So that's building the satellite. That's every single component that's out the door and into space.

[10:04](https://realalpha.app/v/QEACSjIYPrg/?t=604) So roughly speaking, to make a $5 a watt satellite, something like $1 per watt needs to be launched. It needs to be the launch cost. So if we break this down and we look at the numbers and look at, let's say, a 200 kilowatt class data satellite, This satellite should cost about a million dollars in total, of which something like 800,000 will be the cost to build the satellite, and 200,000 should be launch. If 200,000 is a launch, and the satellite weighs something like 1,000 kilograms, which is quite optimistic, actually, for the weight of the satellite, it's actually very optimistic for the weight of the satellite, then launch needs to cost $200 a kilogram.

[10:43](https://realalpha.app/v/QEACSjIYPrg/?t=643) Right now, SpaceX, well, when they were in the market, would sell you launch on Falcon 9 for single-digit thousands of dollars a kilogram, and it would cost them something like $1,000. So you can see how Falcon 9 is not a sufficient vehicle to make- It's like 5X off. Yeah, it's 5X off, exactly. Additionally, if you look at any of the companies that are currently deploying data centers at large. Okay, look at OpenAI, Anthropic, any NeoCloud, SpaceX themselves actually with XAI. They're deploying data centers on the order of sort of one to 10 gigawatts a year. So let's do some more math. If a 200 kilowatt satellite is one ton, and a rocket like Falcon 9 can carry about 10 tons to sun-synchronous orbit, which is where AI data sets need to be in order to be useful, each Falcon 9 launch can carry two megawatts. What that means is that to get to two gigawatts, okay, which is a medium number in comparison to one to 10 gigawatts, it's a thousand

[11:47](https://realalpha.app/v/QEACSjIYPrg/?t=707) **Saurav Shroff:** launches. So Falcon 9 is 10x short on scale and 5x short on cost. And so what this means is that

[11:56](https://realalpha.app/v/QEACSjIYPrg/?t=716) you need a roughly one order of magnitude improvement in both scale and cost. cost just to get to the baseline where the AI data sat is both better at cost-wise and even modest in scale, right? People talk about how AI data sats are the solution because, you know, solution to needing lots of AI compute because they have infinite scale. And this is true. More or less, they have infinite scale once you have infinite, you know, pseudo-infinite launch capacity. But to get there from Falcon 9, you need 5x cheaper cost and 10x higher scale just to get to the baseline. And then you need 100x scale to get to appreciable capacity and 1000x scale to get to really serious capacity. So a rocket like Falcon 9 is limiting. And it's limiting because it's too expensive and it's too slow to launch. The thing that makes its cost high and the rate of launch low is the fact that the second stage of the rocket is not reusable, which is to say it gets thrown away on every launch. So to launch it twice, you have to build two second stages, which is time consuming. And to launch it twice, you have to build two second stages, which costs money. So the solution to this problem that is known and has been known for

[13:05](https://realalpha.app/v/QEACSjIYPrg/?t=785) decades is to make both the first and second stage of the rocket reusable. So in simple English, rockets that are useful for deploying AI data sets are defined as rockets whose first and second stage are reusable. Right now, there are exactly two companies on earth that are seriously developing a two-stage orbit fully reusable rocket one of them is spacex who we already described has roughly no incentive to sell services and the second is a company called stokespace which i think is an amazing company that is also developing a rocket of similar categorically similar performance although it's a bit smaller which doesn't really matter and so it is for this second stage of the future where we are immensely scared because um if you look at the optimal sort the game theory optimal logic for each participant in the market it's as follows first spacex has a rocket that works let's say they have stars let's say starship works in about a year so let's say spacex has a rocket that works they have no incentive to sell they deploy their own data sats because super intelligence is valuable great if another company makes a rocket and that rocket works and the other people who are also building rockets and by rockets here i'm talking about

[14:17](https://realalpha.app/v/QEACSjIYPrg/?t=857) Useful rockets for data sets, so reusable ones. The other reusable rocket makers are not selling services. Then the game theory optimal choice for them is the same. My rocket works. I have no incentive to sell it. Scale AI data sets because superintelligence is valuable. Then if another company builds rockets that works, their logic is the same. So it's for this future that we're worried. So basically it's never optimal for any of these companies to sell launch once. On an instantaneous basis, this is true. And as soon as multiple companies start selling launch, this dynamic changes. So really you just need two companies that are selling launch.

[15:01](https://realalpha.app/v/QEACSjIYPrg/?t=901) And for a long time, we thought the launch market is gonna be solved because SpaceX sells launch. So it just takes one other person or company to come along and also sell launch. So if you sold launch, we'd be in good shape. But we went from one hoping for two down to zero hoping for one but when we're at zero nobody wants to be the first um and so uh our solution to this problem is it's very simple and this will be my first time saying this on record which is that we have a date in the back of our heads where if the launch market doesn't fix

[15:33](https://realalpha.app/v/QEACSjIYPrg/?t=933) **Saurav Shroff:** itself um we are going to go build a launch company and um the reason that we have to do it is because we want our customers to be successful.

[15:42](https://realalpha.app/v/QEACSjIYPrg/?t=942) So for those who are listening who haven't had, who don't know about us, we supply mass-produced components, including the world's lightest space solar panel to companies building satellites, including those building AI data centers in space. And because we want our customers to be successful and because they have no way of... successful in the present world, there's a date in the back of our head and there is a clock that is ticking towards that date. And if something doesn't change, like SpaceX entering the market or multiple players competitively bidding large volume launch contracts and offering reasonable prices like $200, $300, or heck, $500 a kilogram, we will go build a launch company and sell that product. And the reason is for the betterment of the world, because we think superintelligence is something that everybody should have not just a couple of companies when you're thinking through the decision to like this business alone is hard enough like it's very very difficult to build one space company um how did you kind of math the logic in your head for even thinking about doing a launch company as well so first of all uh i don't want to understate the difficulty um like there's a phrase it's not rocket science and that phrase exists for a reason because rocket science

[17:02](https://realalpha.app/v/QEACSjIYPrg/?t=1022) well rocket science isn't actually that hard but rocket engineering and manufacturing is extremely challenging i would say it's like it's less math and more logic i mean you said how did you math the logic there wasn't a lot of math there was mostly logic i mean a little bit of math like okay what would rocket look like how big would it be where would you launch it from uh these types of questions but it's mostly logic like the logic is like okay super intelligence is being made um So you could make a bet that it doesn't need to scale, in which case we have no problems, right? We don't need to scale. The ground is sufficient. Deploy all your data centers on the ground, every sappy. Okay, that would be one bet. That doesn't really seem very likely.

[17:41](https://realalpha.app/v/QEACSjIYPrg/?t=1061) um in the more likely case where we do need to scale uh the ground doesn't work space becomes better um and then if you just go through the logic of what happens in that scenario which seems to me like the likely scenario um it's quite grim it's like um you have this like sort of standoffish market with like massive companies and i'm not against big companies okay i'm not socialist but i am against like um futures where technologies are don't are like important technologies aren't competed right like like uh i mean you could get a little political and make some comments about like medicare and stuff like i don't need to like if super intelligence is good enough to do most people's jobs and you know society looks like a society that's like supported by ubi like do you want that super intelligence to be owned by one company that is like effectively the government like i'm not saying that will happen or that that's anybody's intention i'm saying i don't think we want that um and uh i looked around i asked every launch company how much launch they would sell me and for how much discovered that the market is completely fucked and um wanted to do something about it and and it's not like i have um i'm just uh sounding

[18:52](https://realalpha.app/v/QEACSjIYPrg/?t=1132) the alarm that it's a problem there's this uh funny line everyone has a plan until they get punched in the face and i got to imagine that the ai data center punch is effectively both like a great blessing and a great curse which is there's like infinite demand in space for solar panels and ironically you will not be able to put any of your solar panels up in space um so how did you kind of go through the period where you realized this was kind of a fact and how did you kind of changed direction in that period um well i mean i basically told you everything that like i basically stepped through a hundred percent of the logic already um and uh the more graphic answer is um a friend of mine who probably will know i'm referring to him but i can't name him by name asked me some questions but that led me to reconsider my opinion okay so so just for some context three months ago i didn't think this was a problem i was like yeah of course starship is the best rocket ever made um it's so amazing like it's going to create infinite supply and i still believe i literally still believe that um that Starship is the best rocket ever made and that it will create pseudo infinite supply of launch services. SpaceX has proven themselves as a ruthlessly competent executor. So three months

[20:12](https://realalpha.app/v/QEACSjIYPrg/?t=1212) ago, I thought, fine, you need to build for the biggest possible scale, which is what we're doing here, which we haven't stopped, by the way. It's not like we changed our plans. And we need to build for that big scale because superintelligence is coming. That's going to drive insane demand. and you need to be able to service it. And my friend asked me a few questions that led me to reconsider this belief, led me to consider for the first time, hey, what if actually the incentives work out such that despite having pseudo-infinite supply, none is available? Okay, what a concept, right? It's like imagine like a banana plantation, but you're starving, it's confusing.

[20:55](https://realalpha.app/v/QEACSjIYPrg/?t=1255) And I stepped through the logic and I was like, you know it might actually be the case that we are starving on a banana plantation um and then I asked the banana plantation owners so to speak okay the companies the rocket companies um you know what they would sell me and at what price and discovered that in fact we will soon be starving on the banana plantation proverbially um and so graphically what happened is took a deep breath opened a spreadsheet and started designing a rocket you've been able to drop the cost of a space solar panel by like at least an order of magnitude 50 times yeah 50 times can you walk me through why space solar panels used to cost so much and what you were able to change yeah um so i'll answer the question directly first because i tend to like give like lots of roundabout answers um which is basically just we we we rebuilt it from first principles so long time ago 10 years ago space solar panels used to be made out of these like super rare earth materials built into wafers it would cost hundreds of dollars a watt so we talked about a five dollar a watt data set okay a space solar panel from 2010 would cost you 300 bucks a watt maybe even a thousand okay so

[22:12](https://realalpha.app/v/QEACSjIYPrg/?t=1332) like ridiculous sums of money and they're made this way because i don't know like some scientists discovered that if you made it this way it'd be super super reliable and the volumes were low enough that nobody cared how much they cost this this went on for a long time uh where the volumes were were were very low and people tolerated high cost because their end customer might have been the military or maybe they were doing you know a demo mission for a satellite that they planned to scale but hadn't scaled and so didn't care about the unit cost of the demo and this went on 2010 through 2020 maybe even 2023 2024 um and we basically said like okay what are the materials like the minimum set of materials that you need to make a space solar panel and uh worked completely from first principles up to try to produce a solar panel out of those materials so like we didn't say like how is the old panel made and what about it is expensive because the answer is everything like every single component like you can't even buy one you know you can't even buy like the plugs okay uh for for a price that makes sense and we thought about every single material so so didn't look at the old one and say how we could make it

[23:20](https://realalpha.app/v/QEACSjIYPrg/?t=1400) cheaper but said okay given that we have this cost target and scale target that is driven by the need from data set can we invent a solar panel from first principles around material a material set and a manufacturing process that is both cheap and scalable so the short answer is we started from a complete clean clean sheet and it turns out that actually a lot of things are solar panels like believe it or not the led in your phone like like when you turn your phone flashlight on if you shine a bright enough light at that thing it's a solar panel too it's just not a very efficient one so a lot of materials that make solar panels and the ones that people were using for space were just not the right ones. Now, it's not like we didn't run into challenges along the way. There is an immensely long list of reasons why solving a problem like this wasn't and isn't trivial. The only thing that's true about now that wasn't true about whatever, 2010 or 2020, is that in 2010, it looked like you needed to ship the size of a small schoolroom, you know, like a school classroom's worth of solar panels. Like imagine a school classroom and, you know, laying a bunch of solar panels on the floor and filling it up. Like that

[24:33](https://realalpha.app/v/QEACSjIYPrg/?t=1473) was the annual volume. I mean, maybe getting the math like a little bit wrong here, but that was like roughly speaking, the annual volume of solar panels for space in 2010 or something like that. And now it looks like the annual demand for solar panels in space will be like, you know 50 000 football fields a year um and so because the demand is so much bigger the incentive to make the cost lower is so much higher and the incentive to develop new processes and invent new materials and test those materials and um is extremely high so before we go into the actual solar panel we need to talk about the two types of cost a solar panel for space has and it's quite simple the first is the actual cost so how much it costs to make and plus markup how much we sell it for which is a really really compelling price a price that works for a data set that costs five dollars a watt okay so we list it for fifteen dollars but we sell it for a lot less than that if you buy a lot of volume way way like an order of magnitude less the second is the cost of launching it okay so the heavier it is the more it costs to launch in fact earlier we were talking about how a hypothetical 200 kilowatt class data sat might weigh just one metric ton

[25:54](https://realalpha.app/v/QEACSjIYPrg/?t=1554) um if you take the solar panels off of for example a starlink satellite the solar panels alone will weigh about one metric ton for a 200 kilowatt class satellite maybe a little bit more i actually don't know because i don't work at spacex and i don't have their data, but roughly speaking, that's about true. So to make the platonic ideal best and cheapest solar panel that delivers the lowest levelized cost of electricity in space, you need both to be extremely cheap to manufacture. So it can't be made out of rare earth materials. It can't be made out of things that are in short supply. It has to be made out of common materials.

[26:31](https://realalpha.app/v/QEACSjIYPrg/?t=1591) And it can't be made out of any processes that are exquisite or challenging or extremely high temperature or require like very very precise instrumentation without revealing too much because i can't i can't you know i can't i'd love to i'd love to just be like okay step one step two um i can't talk about the exact recipe but um we when we started we asked ourselves like okay are there other optoelectronic devices so that the broad umbrella term for for things that are solar panels and led lights and stuff are optoelectronics are there other optoelectronic devices that meet the requirements we have, which are that it's extremely like dirt cheap to manufacture, like negligibly cheap to manufacture, meaning it's made out of super commonplace materials, and it's made out of materials that are either printed or deposited or in some capacity ultra, ultra thin. It turns out that there is, and it's called an OLED. An OLED is an organic light-emitting diode or a light-emitting diode display. And everybody maybe hasn't heard the term but has certainly used one because just about every iPhone uses an OLED display as a screen on the front. So the screen of an iPhone is an OLED display. And an OLED display is made using most

[27:43](https://realalpha.app/v/QEACSjIYPrg/?t=1663) of the same processes as our solar panel. So one of the funny terms that I like to use is to describe the solar panel as an OLED, an organic light-absorbing diode, because it is a diode, But it is light absorbing and it is manufactured very much like an OLED display, which is very, very cheap.

[28:00](https://realalpha.app/v/QEACSjIYPrg/?t=1680) **Saurav Shroff:** So, the solar panel is really like a cheaper and light-absorbing optimized version of an OLED display. And we know that it is roughly as cheap as it can get because the materials that it uses are absolutely commonplace materials.

[28:15](https://realalpha.app/v/QEACSjIYPrg/?t=1695) Like there are materials that if I told you the names of all the materials, you could go online and buy them all in one day for a few hundred bucks and have them tomorrow. Like they're not like we're not nothing from like, you know, like Africa or, you know, some weird like every single ingredient can be bought in every single country can be made in like hundreds of countries in the world. Super commonplace materials. the processes are also very simple and we know that it is the absolute thinnest that it can be because if it were any thinner it would reach a limit of thinness where it would be physically impossible for it to be anything but transparent in other words it is so thin that if it were thinner it would be see-through and something that is see-through is not a solar panel by definition um so if you guys want to take a look at it i have one here um this is what it looks like and um you can see it's like you know roughly speaking as thin as a piece of paper this one is a little bit thicker than a piece of paper um by about 20 percent and um it's unbelievable i this is a is a piece of paper that you can roll up um and it generates electricity so i guess this may be a slightly crazy question but how does the thickness of

[29:33](https://realalpha.app/v/QEACSjIYPrg/?t=1773) solar panel because most solar panels that i've seen are incredibly thick like an inch yeah why does this work if i told you everything about how it works um i wouldn't be doing my job very well so i can't um i will correct one thing about what you said which is that um a the average solar panel on the ground looks like it's an inch thick from the outside because it's got this thick aluminum frame around it but if you took the aluminum frame off it would only be a few millimeters thick um so we're talking about this solar panel being like the thinness of a piece of paper and the average space solar panel being the thinness of like 10 or 20 or maybe 30 pieces of paper.

[30:07](https://realalpha.app/v/QEACSjIYPrg/?t=1807) So like a small stack, you know, not even the height of a pad of sticky notes, okay? Like thinner than that. So it's not like we're talking about like an inch to a paper. We're talking about like a small stack of paper to a piece of paper, which at the scale of an AI data set really, really matters. Like it is, I mean, it is probably the case that a satellite with this product at a 200 kilowatt class scale will be like one third of the weight. than a satellite without it, and oh, by the way, way cheaper. Right now, we're sitting in a massive new warehouse that's like 130,000 square feet that you're gonna basically turn on your first major production line.

[30:44](https://realalpha.app/v/QEACSjIYPrg/?t=1844) And forgetting that who knows how the hell you actually even secure launch, you are building ahead of- Which we do know, by the way. Right, right. How are you thinking about basically building ahead of capacity? You said something like there's roughly five megawatts worth of demand for solar panels in space today and you're going to build this for 50 megawatts of capacity so how do you think about that yeah okay uh great question so we are sitting inside uh starpath megafactory la uh this building is uh near long beach that's 130 000 square feet you said um and in about five months this will turn into a 50 million watts per year fab for starlight air which is the name the brand name of our ultra-thin low-cost solar panel and a few other products that we can talk about a little bit later if you add up the wattage meaning the power of every single satellite launched including starlink it sums to something like 50 megawatts here if you don't include starlink it's like five megawatts here so the question is you know you might ask okay why are we building a 50 megawatt fab that's 10 times bigger than demand like why would we Why would we do this? It seems like we spent 10 times too much CapEx. It seems like a big mistake. Why would we do this? And the answer is quite simple, which is that the demand that is being forecasted to us by our customers, who, by the way, are not all, you know, like sometimes people will be like, the demand I'm seeing for my customers, and then the customer is like a startup, okay? And we have customers that are startups too, and that's fantastic. We sell to them, absolutely.

[32:27](https://realalpha.app/v/QEACSjIYPrg/?t=1947) lead. But when I'm talking about demand that's being forecasted, I'm talking about from companies who currently spend billions of dollars on data centers, okay? So the demand that is being forecasted to us by our customers is so immense that, and it grows on such a steep exponential year over year, quarter over quarter, whatever you want to call it, that it leaves us with two options you know so your demand demand is growing on a smooth exponential roughly and it's very very steep and your supply as in your capacity grows in steps so your factory comes online and then you have the same capacity for a little while and then another factory comes online and so you're building basically this like staircase shape graph it's like step shaped graph of capacity and you have you have roughly two choices you can either build the steps above the exponential so here's the exponential and here are the steps so factory one factory two factory three etc or you can build it below and there is one governing word that comes out of every conversation with every customer and it is not quality which is important it is not weight which is important

[33:37](https://realalpha.app/v/QEACSjIYPrg/?t=2017) **Saurav Shroff:** it's not cost which is important it is capacity the questions that our customers are asking us do include those first few things quality weight cost but mostly include how much capacity do you have in year X, in year X plus one, in year X plus two. And we're talking about numbers that

[33:59](https://realalpha.app/v/QEACSjIYPrg/?t=2039) are so immense, they're almost hard for me to comprehend where I am working on building one of the factories right now. Like the numbers are like, okay, this is a 50 megawatt factory. That's 50 million watts. Our customers are saying, how fast can you get to 75 gigawatts a year? that is 1,500 times the scale of this factory, which is the largest building I've ever leased in my life by an order of magnitude. And so if we want to tell a compelling story to our customers that we actually will have the thing that they want, which is capacity, we need to have more capacity than our customers want today.

[34:41](https://realalpha.app/v/QEACSjIYPrg/?t=2081) I mean, it's like, it's as simple as that. You want to sell a big amount of stuff. You better show a big amount of machines that can make a big amount of stuff, even if it means you're spending an extra $50 million on CapEx in a year where you could have gotten away with spending like five, right? So the operating word of the next 10 years for us, not for the whole world, but for us, infrastructure and industrials providers for data centers in space is capacity. And the capacity required by our customers is so gargantuanly immense that I cannot even understand what they mean. Which we won't. If we wanted to do it in LA, which we won't, we would have to lease every building between Long Beach and the 105. Every single one in order to fulfill the demand that our customers want to see in five years now obviously we're not going to do in la we're going to do it in texas where you know like it's easier to build the regulation is more friendly but um so your question was why are you building 10x the capacity of what your customers want and it's because the work the thing they want is capacity even if they're not buying it next year and between building capacity ahead of the curve and building capacity behind the curve.

[36:03](https://realalpha.app/v/QEACSjIYPrg/?t=2163) It is the smarter business decision to build it ahead. And for those saying, which nobody was really saying, but if you said like, you know, what about the cost of capital, et cetera, et cetera, I would say, well, what about the cost of losing? So if you think through basically what your biggest bottlenecks are, you get the lease here, you start filling it out with a bunch of different machines. You have the lineup in, I don't know, six, eight months. March 1st, 2027. great uh so it's about six months what does the scaling process look like from going from one factory with 130 000 square feet and like 50 megawatt production annual output to like a gigawatt um so incidentally 1.2 gigawatts is the size of our next fab which is called starpath austin giga um austin why did you choose like that amount versus a smaller amount like what what's the logic um and we need to climb the orders of magnitude and so um you choose a smaller amount but uh it's the same answer as before which is like in 2027 the things that customers are going to want to see is capacity um more so than what they need and then because the next year they're going to need it right and then in 2028 uh or whatever the year after the things that customers are going to want

[37:19](https://realalpha.app/v/QEACSjIYPrg/?t=2239) to see is capacity more so than what they need because they're going to need it the year after and then in 2029 what are they going to want to see capacity capacity exactly right um customers want to see capacity because they have huge plans and those plans if not done block billions if not tens of billions in revenue for them and that's proven they know that so what they want to see is capacity in simple english they're not fucking around the customers want to see capacity so why one gigawatt um it's just like roughly an order of magnitude higher than this facility this facility yes is technically 50 megawatts so it's 20 times as opposed to 10 times roughly it's an order of magnitude higher what do you expect to get easier as you scale from like a factory that roughly what is that 24 times the size of this one um no uh i don't expect anything well yeah i mean i think some things will be easier like you know we will know tons of things going into austin giga that we knew nothing about going into uh this building this is our this is our first rodeo um but you know like i don't think that there's anything easy about building a fab um and i think if somebody does think that i would say where's your fab um so i don't think it's going to be easy it's hundreds and hundreds of

[38:44](https://realalpha.app/v/QEACSjIYPrg/?t=2324) millions of dollars of equipment almost all of it needs to work almost all of the time it's not quite all but like 90 of it needs to work 90 of the time there's nothing easy about that and it's hard for me to foresee what the challenges are but they'll be immense and with each fab i will move into the corner of the building so at this building we're just in the middle of buying an airstream airstream will be installed in that corner over there and i will sleep in the airstream until the fab is producing product that ships to our customers um and then when we go build austin giga likely the same or maybe a different airstream to be honest i haven't thought about it um will be used for the same purpose um in a similar position the corner why is it important to sleep at the factory if you're sleeping in the factory and you know there's not too much sound deadening and you can hear machines and forklifts then okay actually that specifically that might be a little annoying but um it'll be hard to take your mind off of it um which is exactly what i want one of the things that you mentioned earlier today when we were basically walking around and you were just like telling me the layout and showing me the layout of everything

[39:55](https://realalpha.app/v/QEACSjIYPrg/?t=2395) is you want roughly half the factory to be producing solar panels and you want other half of the factory producing components yeah for like ai data center stats yeah um why are you even doing that in the first place instead of just focusing like really focusing on this one objective um yeah so um first of all i should just clarify one thing quickly which is that we are focused on bringing up production of and shipping to customers the best space solar panel ever made because they've placed orders they've given us down payments we've given them a schedule and we need to uphold to the promises we make to anybody, especially our customers, especially.

[40:36](https://realalpha.app/v/QEACSjIYPrg/?t=2436) So we are focused. You are right that roughly half the factory will be for production of solar modules and the other half will be for the scale production of other products, most of which, all of which actually have not been publicly released. But I'll give you my rationale for building them. So first of all, a data sat or really a satellite bus of any kind has about, I mean, you can slice it up in various ways, but it has about 10 major components. Solar panel, I won't list all of them. So solar panel, it has an attitude system that points the satellite, has a thruster that propels the satellite and allows it to change its orbital altitude or inclination slightly, and a few other things. And when we entered the solar business, the satellite solar business with starlight air um we quickly learned from our customers that it was pretty fucked up market right like they're very they're effectively no vendors there was no there were a lot of people doing sort of like half-ass not that good quality production in-house for their own relatively small volumes and there were no great products on the market as defined by the things that we talked about made products great made it made us make the things we talked about make a solar panel great

[41:51](https://realalpha.app/v/QEACSjIYPrg/?t=2511) for space. Low-cost manufacturer, lightweight, aka low-cost launch. So we learned that pretty quickly. And just through simple curiosity, we also learned that the situation for most of the other things that make a satellite bus, and especially in the case of an AI data set, were in an equally not good position. In other words, if you wanted to make an AI data set, and you needed, you know, 100,000 thrusters, and you needed them to cost, you know, we said the data set should be $800,000, you want the thruster to cost, you know, 50,000, 80,000. That's not a product you can buy. Okay, nobody will sell it to you. And it's not just like that nobody will sell it to you today, which is sort of obvious, like, okay, it's a very cheap satellite. Okay, obviously, nobody's going to sell you, you know, parts to make a dirt cheap satellite today. But it's also the case that if you went to the vendors who were selling those components who are selling those components um and you asked them hey i want to buy a hundred thousand and i want to put money down today what unit economics what unit economics can you achieve they're like not even up to up to the challenge like you could argue like the dev cost for a satellite is in the hundreds of millions

[43:07](https://realalpha.app/v/QEACSjIYPrg/?t=2587) of dollars for the first time you make it and this is like a well-supported fact by the fact that like you know starlink v1 costs around that much and so all these satellite startups that

[43:16](https://realalpha.app/v/QEACSjIYPrg/?t=2596) **Saurav Shroff:** spend a few hundreds of millions of dollars building satellite. And if you approach a satellite components vendor and say, hey, I want to buy 100,000 thrusters, I'll give you 100 million up front, which is certainly enough money to develop a great thruster. They're not even up

[43:29](https://realalpha.app/v/QEACSjIYPrg/?t=2609) for the challenge. And by up for the challenge, I mean are willing to sell components, in this case thrusters, but this applies to all the components at a price that actually works for data side like idiot index is just retarded yeah the idiot index is off the charts um in the case of space solar i've seen idiot indices in the hundreds which idiot index is like the ratio between the cost of the materials and the cost of the finished good The idiot index on this product here is, you know, like six at our highest volume price. So if you buy a huge volume from us, we will sell it to you for six times the cost of materials.

[44:13](https://realalpha.app/v/QEACSjIYPrg/?t=2653) And that doesn't mean it's a six times markup. That means it's the materials cost us, you know, let's just say a dollar. They don't, but it's a dollar. And then, you know, process and rent and labor costs like a couple more bucks or three or four more bucks. And we'll sell it to you for six. um that's like an unheard of factor in the space industry just like this doesn't doesn't happen so why do we decide and you know another products that we're going to be building here i know you've seen some but none of them are public um but the the basic gist is there are about 10 components that make up a data set and we are making more than and right now we make one of them very well extremely well um the best in the world we will be making more of them too um and the reason why is because we want our customers to be successful and so if we're selling our customers you know if we have a if we have a eight billion dollar contract with our customers to deliver them a huge quantity of solar for a data set constellation but they can't be successful in making that data set constellation because there's nobody who can supply them with stuff that they need um we want them to be successful and we don't want them to go out of business or

[45:21](https://realalpha.app/v/QEACSjIYPrg/?t=2721) drop the contract because they can't do it so it's as simple as like it's the same conversation as a rocket it's like all like all of our motivations come from you know the feeling that number one we need to earn our customers business by providing them with a great product and number two once we've earned their business we need to make sure that they're successful in every way that we can and part of that is the most obvious thing just delivering on time a good quality product which we're very quite anal about and then the second thing is looking around and wondering you know what else can i do for you um so it's like a hospitality mindset almost which is like i know like almost sounds dumb it's like what do you mean it's business isn't it supposed to be like super cut and dry but it's like not for me not really it's very it's a from a hot it feels like feels the same as the reason why like you know to offer to make you tea after dinner it's like oh what else can i do for you um and uh you know the first thing i can do for you is is give you great solar panel and then if i if i notice that i think there's something well it's not very good at cooking okay um the first thing i can i can do

[46:29](https://realalpha.app/v/QEACSjIYPrg/?t=2789) for you is give you a great solar panel and then if i notice that there's something else that i need to do for you my thought process is well what else can i do for you and um that's the impetus between behind a lot of these products um now past the hospitality phase and in the execution phase it's very different the mindset is more like okay we've decided to do this thing what's a schedule and let's get the damn thing done um and so it's not it's not a very hospitality like mindset when we're executing but when we're thinking about what to do it totally is i know that the biggest push for building moon base is going to i think be the mass driver on the moon which is just basically pumping out am i completely off on that um i think you could totally be right it's it's not something that strikes me as like completely obvious like the most obvious like first valuable thing that the moon is able to drive i would say it's probably not the first valuable thing i would say it's probably the first very valuable thing right the first valuable thing on the moon is a product you could call novelty so like if you can bring humans there you will certainly find a lot of humans that are willing to pay to go you know if it's safe i mean obviously

[47:42](https://realalpha.app/v/QEACSjIYPrg/?t=2862) if it's safe and if the price is reasonable like probably in the 10 million dollars per trip price you'll probably find like at least 100 people who want to go um and probably in the 1 million dollars per trip price you'll probably found find you know thousands of people who want to go and then if you can get the price to 100k, which is like, it's kind of pushing it in terms of like, what's physically possible. Like, I mean, if you just assume the Starship is perfectly free, SpaceX sells at no margin, it only costs fuel, even then it'll cost around 100k. And there's no demand pricing of people can, you know, only a small number of people can go.

[48:19](https://realalpha.app/v/QEACSjIYPrg/?t=2899) Yeah, I'd say the price probably lands in like sort of half million dollar range in the long term. And at that price, there is certainly like a pretty high amount of demand, like enough demand to create a base, enough demand to fund a lot of flights. It will not look like United Airlines, but it will look like probably how United Airlines looked when airplanes were brand new. One of the really interesting things that I'm thinking through in this conversation is you mentioned that there's effectively or there's effectively going to be zero limit on the amount of intelligence, I believe, that people want to use, which means that if people, you know, SpaceX can put satellites up in orbit that generate intelligence, it creates this amazing demand pull forward where it's like effectively what is the maximum scaling rate?

[49:13](https://realalpha.app/v/QEACSjIYPrg/?t=2953) that you can build these starships, that is the thing that you should do. Or fly the starships. Yeah, fly the starships, build the starships, whatever. Build the launch pads, all the stuff. Yeah. Whatever that maximum possible rate is, we should, you know, march down that rate. Yeah. Because there's always going to be demand. Yeah. And this is, like, not a guarantee with anything other than just GPUs and space. Yeah. And so, ironically, it's, like, both bad short-term because, you know, there's so much demand for it and great long-term because... How is that bad? but bad short term because you aren't able to use it right like they're not building for the customers they're just because it creates this like selfish dynamic where you don't want to be cooperative then long term they have such a ridiculous amount of starships built because of this intelligence explosion and ability to fly them and they've got like a million tests on one starship yeah it's actually it's actually like um so i actually like wasn't um uh i mean i'm like clearly ai pilled but i i wasn't i didn't enter like the space industry for obviously for that reason i mean and i didn't have that foresight um but the thing that excites me

[50:24](https://realalpha.app/v/QEACSjIYPrg/?t=3024) probably the second most with super intelligence the thing that excites me the most is like what it will do for humans obviously like okay like if you maybe work will become optional which i know like socialists think is horrible but like it's very easy to solve like just add ubi um so you can do work if you want like that i think is amazing okay like maybe art will have a resurgence maybe like renaissance 2 or something not that like that sounds particularly interesting to me but um i think it's like incredible but the thing that excites me the second most actually about the space industry excuse me about the super intelligence is that that was what you just said which is that it is becoming and it will become like the ultimate driver for scaling space technology, like cargo and passenger transport was the ultimate driver for scaling passenger and cargo air transport technology, with the caveat that scaling passenger and cargo air transport technology came with barely any benefits, right?

[51:24](https://realalpha.app/v/QEACSjIYPrg/?t=3084) It was just like, we have passenger transport, we have cargo transport, and we have like, i don't know general aviation where you can like screw around in an airplane as a result so like the side benefits of united airlines scaling their business are small and the main benefits which is that they have passenger transport you know are big but in the case of superintelligence the side benefits of scaling a rocket are in my opinion immense because like for so many years you know nerds like me i mean i'm pretty young but like there are older nerds right for a lot for like decades It's nerds have been saying, you know, why don't we build base on the moon? Why don't we build base on Mars? And the answer is like basically some combination of like, we can't afford the CapEx to develop the core piece of technology, which is the rocket. Okay, undeniable that the rocket is the main piece of technology. And on top of that, you know, like, even if we did, it would take too long. But with superintelligence, it's super different because the motivation to put the CapEx into the rocket is completely different.

[52:30](https://realalpha.app/v/QEACSjIYPrg/?t=3150) completely legitimate economically. Like it's totally a rational decision because it's like, I spent $5 billion developing a new rocket. I earned $5 billion in, you know, in Anthropik's case in two fucking weeks. Like literally, like literally people for decades were like,

[52:46](https://realalpha.app/v/QEACSjIYPrg/?t=3166) **Saurav Shroff:** why don't we make a rocket and go to Mars? And now it's like the value of a rocket is two weeks of Anthropik's annualized revenue run rate. And so like they can quote unquote pay for a rocket

[52:58](https://realalpha.app/v/QEACSjIYPrg/?t=3178) in two weeks now obviously you can't design and like you can't design a rocket that works in two weeks you can't spend the money that fast but it's totally rational decision and then the the side benefits are now immense which is that you have like this sort of like base infrastructure base technology layer you've like unlocked this level in the tech tree that enables all these other lower cost levels of the tech tree with respect to the moon and mars and so like super intelligence lets you have working rockets a huge amount of revenue associated with them so they're basically free in the context of other industries that obviously are not free in absolute and then those other possible industries that could form tourism on the moon mass driver on the moon city on mars heck god knows what on mars are now now basically have like a free rocket that all they have to do is pay the operating expensive um so there actually may be some airline analogies here that make sense but i'm not really one for analogies like this is the second most exciting thing to me about super intelligence which is that like it will almost certainly be true that we will build a city on mars after you make um you know the terawatt or even even a two gigawatt

[54:01](https://realalpha.app/v/QEACSjIYPrg/?t=3241) data set constellation in space will get you to a working rocket i mean there it's necessary how fast do you think the mars base is going to get built because if you use a starship for mars you get it to use it once every couple years whereas if you use it for satellites in orbit you don't need that many it's like multiple times a day yeah pretty fast also like and it's worth noting that a mars base is built on on the back of like a thousand starships and um a a ai data sat constellation is built on the back of like a hundred thousand starship flights so while the re while it is true that the reuse factor for satellite deployment is drastically higher or well two orders of magnitude or more higher it is also true that the the sort of goal density, meaning how much of your goal you achieve per flight is much higher in the case of

[54:54](https://realalpha.app/v/QEACSjIYPrg/?t=3294) **Saurav Shroff:** Mars. I mean, if I just put 10 starships on Mars, just, if I put 10 starships on Mars, then that will have a profoundly large impact on, like a categorical impact on history.

[55:09](https://realalpha.app/v/QEACSjIYPrg/?t=3309) There'll be a city, a small one, a city. But if I put 10 starships in sun-synchronous orbit and deployed data sats you know that's like that's like not even that's like zero at some point along the line of building a thousand starships it just makes sense to take 10 of those and then just send them off yeah yeah exactly like the thing about the thing that's true about super intelligence is that it gives you the technology for free like if you are a mars enthusiast which i am the technology for going to mars becomes free with super intelligence because you get this excuse to build a rocket and get it paid for by the end user, the end main user of the rocket, which is the AI customer. And you get to build this extremely large and mature supply chain, automotive-like supply chain for data sats.

[56:00](https://realalpha.app/v/QEACSjIYPrg/?t=3360) Where importantly, like if you want to build something that's like an offshoot of an automobile, it's doable with automotive components, right? Like the automotive industry feeds many other minor offshoot things that are automotive-like too. And so like the infrastructure to build a city on Mars is satellite-like. In fact, this is such a full circle irony because the way that we got into the business of selling solar panels to satellite companies is by first making them for Mars. like we were first a mars company and are now both and we're like okay let's make a great solar panel that works on mars and then we're like oh it's pretty good let's sell it um and then people people loved it um in fact the original uh sort of like i guess like marketing line or whatever like the thing that i had listed on the website for starlight was engineered for mars flies anywhere um so the thing that you get from super intelligence is the free is the free tech tree unlocks you don't get the free opex so you still have to buy the rocket you still have to pay to use it obviously i mean it's like it's not a charity but um the you know if you asked somebody from like let's say you know some enthusiast person on twitter 10 years ago like hey what

[57:09](https://realalpha.app/v/QEACSjIYPrg/?t=3429) does it cost to build a city on mars to be like okay well you have to develop a rocket and then add up all these costs and then the biggest item would be developing the rocket you ask that same person you know not today but three or four years from today hey what does it cost to build a city on mars they'd be like well we have a rocket we have all the solar panels we need we have all the power electronics we need we have all of the you know whatever components that can be reused from data set so we just need to build these other things which aren't necessarily simple but that total cost would be you know 10 times or 100 times less than what it would be before a superintelligence sort of gave you the rocket for free it's not to understate the difficulty of making rocket it's just to say the intent the incentive of doing it for super intelligence is incredible and very exciting because it gets done