Tag: AI Infrastructure

Elon Musk explains IP (Intellectual Property), the fabs today all basically use machines from like five companies. Yeah, you know, so you’ve got ASML (ASML Holding), Tokyo Electron, KLA, Lam Research, you know, et cetera.

Elon Musk with Dwarkesh Patel & John Collison – The Future of AI is in Space – Part 8: Terafab, ASML & Chip Manufacturing (Full Transcript)

Gail AlfarLeave a comment

In Part 8, Dwarkesh Patel points out that even with abundant power from space-based solar, the ultimate bottleneck for scaling AI will likely be chip production. The conversation dives into Elon’s plans for building “Terafabs,” the difficulties of scaling semiconductor manufacturing, and why memory may be an even bigger constraint than logic chips. It then shifts to the broader philosophical mission behind xAI and SpaceX — propagating consciousness and intelligence into the universe.

Transcript:

Dwarkesh Patel noted that even with more efficient solar panels in space, the chips themselves would still be the ultimate limiter long before reaching terawatt scale. He asked how the world would produce a terawatt of logic compute by 2030 when today the entire planet only has about 20–25 gigawatts.

Elon Musk: “You need to build a lot more chips and make them much cheaper.”

Elon Musk: “I guess we’re going to need some very big chip fabs.”

Elon Musk: “I’ve mentioned publicly that the idea of doing sort of a terafab, terabytng the new Giga.”

Dwarkesh asked for details on the plan: what level of the stack they would build themselves versus partnering with an existing fab for process technology.

Elon Musk: “Well, you can’t partner with existing fabs because they can’t output enough. The chip volume is too low.”

Elon Musk: “IP (Intellectual Property), the fabs today all basically use machines from like five companies. Yeah, you know, so you’ve got ASML (ASML Holding), Tokyo Electron, KLA, Lam Research, you know, et cetera. So at first I think you’d have to get equipment from them and then modify it or work with them to increase the volume. But I think you’d have to build perhaps in a different way. So I think the logical thing to do is to use conventional equipment in an unconventional way to get to scale and then start modifying the equipment to increase the rate.”

On November 24, 2025, on a rainy day in Bastrop, Texas, Prufrock-5 left the Boring Company factory.
On November 24, 2025, on a rainy day in Bastrop, Texas, Prufrock-5 left the Boring Company factory.

John Collison drew the parallel to how The Boring Company started.

Elon Musk: “Yeah, kind of like. Yeah, you sort of buy an existing boring machine and then figure out how to dig tunnels in the first place and then design a much better machine that’s, I don’t know, some orders of magnitude faster.”

John Collison asked whether the fact that China has not duplicated TSMC gave Elon pause about the difficulty of building advanced chip production.

Elon Musk: “It’s not that they have not replicated TSMC, they have not replicated ASML. That’s the limiting factor.”

John Collison asked if Elon thought it was simply the sanctions preventing China from advancing.

Elon Musk: “Yeah. China would be outputting vast numbers of chips at leading edge if not for the sanctions.”

John Collison followed up, noting that China had been able to buy 2 nm or 3 nm chips until relatively recently.

Elon Musk: “No. The ASML bans have been in place for a while, but I think China’s going to start making pretty compelling chips in three or four years.”

Elon explained the massive manufacturing requirements needed to match orbital AI ambitions, noting that memory was actually his biggest near-term concern.

Elon Musk: “I don’t know yet is the right answer. So it’s just that to produce at high volume and to reach large volume in say 36 months to match the rocket payload to orbit… You need 100 gigawatts worth of chips. You’ve got to match these things. The master orbit, the power generation and the chips. And I’d say my biggest concern actually is memory… That’s why you see DDR prices going ballistic.”

Elon then shared his current plans and constraints around chip production.

Elon Musk: “I don’t know how to build a fab yet. I will figure it out. Obviously I’ve never built a fab.”

Elon Musk: “I don’t think it’s PhDs. It’s mostly people who are not PhDs. Most engineering is done with people who don’t have PhDs.”

Elon Musk: “Right now, like Tesla’s pedal to the metal max production of going as fast as possible to get AI5 chip design into production… That’ll probably happen around the second quarter ish of next year, hopefully. And then AI6 would hopefully follow less than a year later.”

Elon Musk: “Yeah, and we’ll be using TSMC Taiwan, Samsung Korea, TSMC Arizona, Samsung Texas and we still booked out all the capacity we can.”

Elon Musk: “The point is you’ve got to build the fab and you’ve got to start production, then you’ve got to climb the yield curve and reach volume production at high yield. That from start to finish is a five year period. And so the limiting factor is chips. Limiting factor once you can get to space is chips. But the limiting factor before you can get to space will be power.”

Elon explained that while launching at massive scale from Earth would be extremely difficult, the moon offered a much better long-term path using mass drivers.

Elon Musk: “I don’t see any way that you could do 500 to 1,000 terawatts per year launch from Earth. But you could do that from the moon.”

Dwarkesh then zoomed out to the bigger philosophical picture behind SpaceX and xAI.

Dwarkesh Patel asked whether, by the time humans are sending ships to Mars, Grok would be on board with them, and how that relates to the main risks people worry about with AI.

Elon Musk: “Well, I’m not sure AI is the main risk I’m worried about. I mean the important thing is that consciousness, which I think arguably most consciousness or most intelligence, certainly consciousness is more of a debatable thing. The vast majority of intelligence in the future will be AI… So you want to take the set of actions that maximize the probable light cone of consciousness and intelligence.”

Elon Musk: “Yeah, I mean to be clear, I’m very pro human, so I want to make sure we take sort of actions that ensure that humans are along for the ride. We’re at least there. But I’m just saying the total amount of intelligence, I think maybe in five or six years AI will exceed the sum of all human intelligence. And then if that continues, at some point human intelligence will be less than 1% of all intelligence.”

Elon discusses the need to build massive Terafabs, the challenges of scaling chip and memory production, and the deeper mission of propagating consciousness and intelligence into the future. In Part 9, the conversation continues with more on the long-term vision for humanity and AI in space.

Elon discusses SpaceX potentially becoming a hyperscaler for orbital AI, the realities of raising massive capital, and the long-term physics required to scale significantly up the Kardashev scale.

Elon Musk with Dwarkesh Patel & John Collison – The Future of AI is in Space – Part 7: SpaceX as Hyperscaler, Capital Markets, and the Kardashev Scale (Full Transcript)

Gail AlfarLeave a comment

In Part 7, the conversation turns to whether SpaceX could evolve into a hyperscaler — building and operating vast orbital AI infrastructure and potentially providing compute power to others. John Collison and Dwarkesh Patel explore the capital requirements, the possibility of going public, and the deeper physics of long-term energy scaling. Elon shares his thoughts on speed as the ultimate constraint and what it would actually take to move significantly up the Kardashev scale.

Transcript:

Dwarkesh Patel asked whether the long-term vision was for SpaceX to become a hyperscaler — launching and operating vast orbital AI capacity and then providing (or lending) that compute power to other companies.

Elon Musk: “Hyper. Hyper, yeah. I mean, if some of my predictions come true, SpaceX will launch more AI than the cumulative amount on Earth of everything else combined.”

Dwarkesh followed up on whether this capacity would mostly be used for inference or training.

Elon Musk: “Will be inference already? Inference for the purpose of training is most training.”

John Collison then explored the business implications, noting the shifting narrative around a possible SpaceX IPO. He pointed out that SpaceX had long been extremely capital efficient, but the scale of building orbital AI infrastructure would require capital raises far beyond what private markets had demonstrated they could comfortably provide — even as AI labs were already raising tens of billions. He asked if going public was the logical next step and more broadly about the difference in capital availability between public and private markets, as well as whether debt financing could suffice.

Elon Musk: “Yeah, I have to be careful about saying things about companies that might go public.”

Elon Musk: “There’s a price to pay for these things.”

Elon Musk: “Yeah, there’s a lot more capital in the very general. There’s obviously a lot more capital available in the public markets than private. I mean, it might be, it’s at least, at least, it might be 100 times more capital, but it’s at least way more than 10.”

John Collison noted that highly capital-intensive sectors like real estate are typically debt-financed once they have predictable near-term revenue.

Elon Musk: “A clear revenue stream.”

John Collison agreed.

Elon Musk: “Speed is important. So I’m generally going to do the thing that, I mean, I just repeatedly tackle the limiting factor, whatever the limiting factor is on speed, I’m going to tackle that. So there’s, if capital is the only factor, then I’ll solve for capital. If it’s not limiting factor, I’ll solve for something else.”

Dwarkesh Patel observed that, based on Elon’s past comments about Tesla being public, he would not have expected Elon to see going public as the way to move fastest.

Elon Musk: “Normally I would say yeah, that’s true. Like I said, I mean, I’d love to talk about this in more detail, but the problem is like if you talk about public companies where they become public, you get into trouble and then you have to delay your offering and then you.”

John Collison noted that this was again about speed.

Elon Musk: “Yes, exactly. So you can’t hype companies that might go public. So that’s why we have to be a little careful here.”

Elon then pivoted to the fundamental long-term physics of scaling.

Elon Musk: “But we can talk about physics. So the way you think about scaling long term is that Earth only receives about half a billionth of the sun’s energy. And the sun is essentially all the energy. And this is a very important point to appreciate because sometimes people will talk about marginal nuclear reactors or any various fusion on Earth, but you have to step back a second and say if you’re going to climb the Kardashev scale and have some non trivial and harness some non trivial percentage of the sun’s energy, like let’s say you wanted to harness a millionth or a millionth of the sun’s energy, which sounds pretty small, that would be about, call it roughly 100,000 times more electricity than we currently generate on Earth for all of civilization, give or take an order of magnitude. So it obviously the only way to scale is to go to space. With solar, from launching from Earth you can get to about a terawatt per year. Beyond that you want to launch from the moon, you want to have a mass driver on the moon, and that mass driver on the moon you could do probably a petawatt per year.”

Elon discusses SpaceX potentially becoming a hyperscaler for orbital AI, the realities of raising massive capital, and the long-term physics required to scale significantly up the Kardashev scale. In Part 8, the conversation continues with more on the engineering and strategic path forward.

Elon predicts that within five years, more AI will be operating in space than currently exists on Earth, and discusses the Starship fleet size and launch cadence needed to support it.

Elon Musk with Dwarkesh Patel & John Collison – The Future of AI is in Space – Part 6: AI Capacity in Five Years and Starship Launch Rates (Full Transcript)

Gail AlfarLeave a comment

In Part 6, John Collison asks Elon to project what AI compute capacity might look like in five years — both on Earth and in space. The conversation shifts to the enormous number of Starship launches that would be needed to support large-scale orbital AI infrastructure. Elon shares his prediction that AI in space will surpass all terrestrial AI within five years and discusses the practical realities of achieving very high launch rates.

Elon Musk: “My prediction is that we will launch and be operating more AI in space every year than the cumulative total on Earth, which I would expect to take at least five years to reach. So we’re talking about a few hundred gigawatts per year of AI in space, and rising.”

Transcript:

John Collison shifted the conversation to a concrete five-year horizon. He asked what installed AI compute capacity would look like on Earth versus in space by then.

Elon Musk: Five years? I think probably if you say five years from now, we’re probably going to be launching every year in space the sum total of all AI on Earth, and then some. My prediction is we will launch and be operating every year more AI in space than the cumulative total on Earth, which I would expect to be at least sort of five years from now. A few hundred gigawatts per year of AI in space and rising. So you can get to, I think on Earth you can get to around a terawatt a year of AI in space before you start having fuel supply challenges for the rocket.

John Collison pressed for confirmation on the hundreds-of-gigawatts-per-year figure.

Elon Musk: “Yes.”

Dwarkesh Patel highlighted the launch cadence implied by those numbers.

Elon Musk: “Yes.”

Dwarkesh Patel continued, noting that delivering 100 gigawatts in a single year would require roughly 10,000 Starship launches annually — the equivalent of one launch every single hour, nonstop, from this city.

Elon Musk: “Yeah, I mean that’s actually a lower rate compared to airlines like aircraft.”

Dwarkesh Patel pointed out that there are a lot of airports around the world.

Elon Musk: “A lot of airports.”

Dwarkesh Patel noted the additional complexity of launching into polar or sun-synchronous orbits.

Elon Musk: “No, it doesn’t have to be polar, but there’s some value to sun synchronous. But I think actually you just go high enough, you start getting out of Earth’s shadow.”

Dwarkesh Patel asked how many physical Starships would be needed to sustain 10,000 launches per year.

Elon Musk: “I don’t think we’ll need more than. I mean, you could probably do it with as few as like 20 or 30. It really depends on how quickly the ship has to go around the Earth and the ground track before the ship has to come back over the launch pad. So if you can use a ship every, say 30 hours, you could do it with 30 ships, but we’ll make more ships than that. But SpaceX is gearing up to 10,000 launches a year and maybe even 20 or 30,000 launches a year.”

Elon predicts that within five years, more AI will be operating in space than currently exists on Earth, and discusses the Starship fleet size and launch cadence needed to support it. In Part 7, the conversation continues with more on the technical and operational realities of building large-scale AI infrastructure in orbit.

Elon Musk explains why SpaceX and Tesla may have to start manufacturing turbine components themselves and shares their aggressive plans to scale solar production.

Elon Musk with Dwarkesh Patel & John Collison – The Future of AI is in Space – Part 5: Turbine Shortages, Casting Bottlenecks & Scaling Solar Production (Full Transcript)

Gail AlfarLeave a comment

In Part 5, Dwarkesh Patel raises the question of whether the engineering challenges of building large-scale AI infrastructure might actually be easier in space than on Earth. The conversation then turns to the very real bottlenecks on the ground. Particularly the massive shortage of turbines and specialized casting capacity. Elon Musk explains why SpaceX and Tesla may have to start manufacturing turbine components themselves and shares their aggressive plans to scale solar production.

Transcript:

Dwarkesh Patel asked a central question: while Earth-based power challenges are already enormous, wouldn’t building in space introduce entirely new and unprecedented engineering difficulties — such as radiation hardening, orbital lasers, and other issues that haven’t been solved before? He wondered why anyone would choose these novel challenges over simply building more turbines on Earth, where established companies already know how to manufacture them.

Elon Musk: “I invite again, try doing it and then you’ll see. So like, the turbines are sold out through 2030.”

John Collison asked whether they had considered manufacturing their own turbines.

Elon Musk: “I think in order to bring enough power online, I think SpaceX and Tesla will probably have to make the turbine blades, the vanes and blades internally.”

John Collison asked if Elon meant just the blades or the entire turbines.

Elon Musk: “The limiting factor, you can get everything except the blades. They call the blades and vanes. You can get that 12 to 18 months before the vanes and blades. The limiting factor of the vanes and blades, and there are only three casting companies in the world that make these and they’re massively backlogged, it is Siemens.”

John Collison asked whether it was GE and the big names or subcontractors.

Elon Musk: “No, it’s other companies. I mean sometimes they have a little bit of casting capability in house. But I’m just saying you can just call any of the turbine makers and they will tell you it’s not top secret. They’re probably on the, it’s probably on the internet right now.”

Dwarkesh Patel asked whether, if it weren’t for the tariffs, Colossus would be running on solar power.

Elon Musk: “It would be much easier to make it solar powered. Yeah, the tariffs are nuts, so several hundred percent.”

John Collison began to suggest that Elon surely knew some people who could help.

Elon Musk: “We also need speed. Yeah, no, you know, President has his, you know, we don’t agree on everything and this demonstration is not the biggest fan of solar. We also need the land, the permits and everything. So if you’re trying to move very fast, I do think scaling solar on Earth is a good way to go. But you do need some amount of time to find the land, get the permits, get the solar, pair that with batteries.”

John Collison pressed further, asking why not simply stand up their own massive solar production, noting there is plenty of private land in Texas and Nevada.

Elon Musk: “As I said, we are scaling solar production. There’s a rate at which you can scale physical production of solar cells where we’re going as fast as possible.”

John Collison confirmed they were building the solar cells domestically at Tesla.

Elon Musk: “Both Tesla and SpaceX have a mandate to get to 100 gigawatts a year of solar.”

Elon explains the severe turbine and specialized casting bottlenecks and why SpaceX and Tesla are aggressively scaling their own solar production to 100 gigawatts per year. In Part 6, the conversation continues with more on the engineering and infrastructure challenges of building AI at planetary scale.

Elon Musk: “We are going to make solar. Okay, great. Both SpaceX and Tesla are building towards 100 gigawatts here of solar cell production.”

Elon Musk with Dwarkesh Patel & John Collison – The Future of AI is in Space – Part 4: Turbine Bottleneck & Space Solar (Full Transcript)

Gail AlfarLeave a comment

In Part 4, John Collison asks whether Elon would try to solve the turbine shortage himself or go straight to manufacturing solar at enormous scale. Elon reveals that SpaceX and Tesla are already moving toward 100 gigawatts of solar cell production and explains why solar cells destined for space are dramatically cheaper and easier to produce than those on Earth. He also gives a detailed breakdown of why most people severely underestimate how much power a real AI data center actually requires.

Transcript:

John Collison suggested that the turbine blade bottleneck sounded like the kind of problem Elon would want to attack directly, and proposed that making solar themselves might be the smarter long-term path.

Elon Musk: “We are going to make solar. Okay, great. Both SpaceX and Tesla are building towards 100 gigawatts here of solar cell production.”

Dwarkesh Patel asked how deep into the supply chain they planned to go — from raw polysilicon all the way to finished solar panels.

Elon Musk: “I think you got to do the whole thing from raw materials to the finished cell. Now, if it’s going to space, it actually costs less. And it’s easier to make solar cells that go to space because they don’t need glass or they don’t need much glass and they don’t need heavy framing because they don’t have to survive weather events. There’s no weather in space. So it’s actually a cheaper solar cell that goes to space than the one on the ground.”

Elon emphasized that solar is already extremely cheap on Earth, but moving it to space changes the economics dramatically.

Elon Musk: “Solar cells are already very cheap. They’re like farcically cheap. And if you say, I think solar cells in China are around like 25, 30 cents a watt or something like that, it’s absurdly cheap. And when you take into account now put it in space and it’s five times cheaper because it’s five times — in fact, no, it’s 10 times cheaper because you don’t need any batteries. So the moment your cost of access to space becomes low, by far the cheapest and most scalable way to generate tokens is space. It’s not even close. It’ll be an order of magnitude easier to scale.”

He then shared the real-world difficulties his team faced just getting one gigawatt of power online for the Colossus supercluster.

Elon Musk: “And chips aside, an order of magnitude. The point is you won’t be able to scale on the ground. You just won’t. People are going to hit the wall big time on power generation. There already are. So the number of miracles in series that the xAI team had to accomplish in order to get a gigawatt of power online was crazy. We had to gang together a whole bunch of turbines. And then we had permit issues in Tennessee and had to go across the border to Mississippi, which is fortunately only a few miles away. But then we still had to run the high power lines a few miles and build a power plant in Mississippi. And it was very difficult to build that.”

Elon then explained why most people dramatically underestimate how much electricity is actually needed at the generation level to run a real AI data center.

Elon Musk: “And people don’t understand how much electricity do you actually need at the generator level, at the generation level in order to power a data center? Because they look at the specs, will look at the power consumption of say a GB 300 and multiply that by the number and then think that’s the amount of power you need.”

John Collison noted that even those calculations miss major additional loads like cooling and supporting systems.

Elon Musk: “Wake up. Yeah, that’s a total noob. You’ve never done any hardware in your life before. Besides the GB 300, you’ve got to power all of the networking hardware. There’s a whole bunch of CPU and storage stuff that’s happening. You’ve got to size for your peak cooling requirements. So that means can you cool even on the worst hours, the worst day of the year? Well, it gets pretty freaking hot in Memphis, so you’re going to have like a 40% increase on your power just for cooling.”

He continued breaking down the additional multipliers that are almost always overlooked.

Elon Musk: “Assuming you don’t want your data center to turn off on hot days and you want it to keep going, then you’ve got to say, well, there’s another multiplicative element on top of that, which is are you assuming that you never have any hiccups in your power generation? Like, oh, well, actually sometimes we have to take the generators, some of the power offline in order to service it. Oh, okay, now you add another 20, 25% multiplier on that because you’ve got to assume that you’ve got to take power offline to service it. So the actual — roughly every 110,000 GB 300s inclusive of networking, CPU, storage, cooling, margin for servicing power is roughly 300 megawatts.”

John Collison asked him to repeat the number for clarity.

Elon Musk: “It’s roughly — or think about it like a way to think about it is like 330,000. What you need at the generation level to service, probably service 330,000 GB 300s, including all of the associated support, networking and everything else, and the peak cooling and to have some power margin reserve is roughly a gigawatt.”

Elon breaks down why naive power calculations for AI data centers fall far short of reality and why space-based solar could be the only way to scale at the required speed. I

n Part 5, the conversation continues with more on the challenges and opportunities of building at this scale.

Elon Musk on The Massive Scale of Power Requirements and Utility Bottlenecks

Elon Musk with Dwarkesh Patel & John Collison – The Future of AI is in Space – Part 3: The Massive Scale of Power Requirements and Utility Bottlenecks (Full Transcript)

Gail AlfarLeave a comment

In Part 3, the conversation turns to the enormous scale of power required to run advanced AI at the level Elon envisions. Dwarkesh Patel and John Collison press Elon on the real-world challenges of building terawatts of electricity generation and why the utility industry is such a major bottleneck. Elon explains why private power plants co-located with data centers may be the only practical solution.

Transcript:

Dwarkesh Patel sought clarification on the scale Elon was describing, confirming that he was talking about terawatts of power. The discussion then moved to the extreme difficulty of actually building that much electricity generation at the speed AI development requires.

Elon Musk: “Yeah, well, all of the United States currently uses only half a terawatt per hour on average. Right. So if you say a terawatt, that would be twice as much electricity as the United States currently consumes. So that’s quite a lot. And can you imagine building that many data centers, that many power plants? It’s like those who have lived in software land don’t realize that they’re about to have a hard lesson in hardware, that it’s actually very difficult to build power plants. And then you don’t just need the power plants, you need all of the electrical equipment, you need the electrical transformers to run the transformers, the AI transformers.”

Elon pointed out that the utility industry moves extremely slowly because it is heavily regulated and “impedance matched to the government.”

Elon Musk: “Now, the utility industry is a very slow industry. They impedance match to the government, to the public utility commission. So they’re very slow because their past has been very slow. So trying to get them to move fast is just like, you know, if you’re trying to do an interconnect agreement… have you ever tried to do an interconnect agreement with a utility at scale? Like with a lot of power?”

Dwarkesh Patel laughed and admitted that, as a podcaster, he had never tried to do an interconnect agreement with a utility.

“Now, the utility industry is a very slow industry. They impedance match to the government, to the public utility commission. So they’re very slow because their past has been very slow. So trying to get them to move fast is just like, you know, if you’re trying to do an interconnect agreement… have you ever tried to do an interconnect agreement with a utility at scale? Like with a lot of power?” – Elon Musk

Elon Musk: “In fact, yeah, they have to do a study for a year. Okay. Like a year later they’ll come back to you with their interconnect study.”

John Collison asked whether companies could simply bypass the utility bottleneck by building their own private power plants right next to the data centers.

Elon Musk: “You can build power plants. Yeah, that’s what we did at xAI for Colossus.”

John Collison followed up, asking why this private-power approach wasn’t being treated as the obvious solution to the utility problems they had been discussing.

Elon Musk: “Right. But it begs the question of where do you get the power plants? Where do you get the power plants from? I mean the power plant makers.”

John Collison summed up the deeper issue: there is currently a massive backlog for gas turbines and power plant equipment in general.

Elon highlights that even if companies build their own power plants, they still face major constraints in actually obtaining the equipment. In Part 4, the discussion continues with more on the practical challenges of scaling AI infrastructure at this level.