No One Realizes What Elon Musk Just Created

Kind: captions Language: en Elon Musk has announced his Terafab project, a 10 gigawatt facility with a footprint 10 times larger than the Tesla factory in Texas, which already stretches 1 mile in length. Yes, Musk is indeed outlining a facility 10 miles long to produce the annual equivalent in terms of energy consumption of 1 terawatt of semiconductor chips. To put the scale [music] into perspective, the entire US power grid operates in the range of 1.3 to 1.5 terawatts. Consequently, under the proposed framework, it would require the equivalent of the entire US electrical grid to power the Terafab's output in just 18 months. Naturally, you can see where this is headed. Most of the chips won't be powered on Earth as there simply isn't the necessary electrical capacity. No, instead, Musk wants to deploy them in space where they will be powered by orbital electrical generators in sun-synchronous orbit. Sun-synchronous means always facing the sun. Yes, you heard that right. He intends to commission the equivalent roughly of the entire US electrical output every single year. Now, this is all grandiose and would appear utterly preposterous to any rational person. Okay, Musk is the world's wealthiest individual and his industrial prowess is undisputed by anyone. But, however, that only makes this idea marginally more palatable because here we are talking about reshaping the global structure of energy production and computing as a whole. To put it even more bluntly, this is no longer at a corporate scale or it would represent a singular first in the history of humanity. At the same time, since this man has already revolutionized global aerospace, it's worth a closer look. To make this concept more digestible and more palatable, Musk employs an anchoring technique to normalize the scope, so to speak. He sets an even more distant horizon, even more wild that of conquering the solar system. He begins his conference by introducing the work of an author who established the classification principles of civilizations, terrestrial and extraterrestrial, via a scale that bears his name, the Kardashev scale. Please note the truly extraordinary nature of this move. A captain of industry who begins by introducing a completely disproportionate factory project within a context that is roughly a million times more disproportionate still. So, let's take a deep dive into the orders of magnitude Musk is proposing for the Terafab because they are literally extraordinary. But first, I'll take a few seconds to ask you if you like these tech topics. Subscribe to Macro Capitalist channel while remembering to hit the bell. As you may be discovering this video, our last video on Elon Musk's interview with Larry Fink. Before even touching on space, one must understand the problem Musk is attempting to solve. Today, the artificial intelligence industry relies on a fragmented architecture. Nvidia designs the chips, TSMC manufactures them, and the major technology companies deploy them. Even Nvidia, with over $4 trillion in capitalization, remains dependent on a foundry located in Taiwan. And that is something Musk cannot accept. Why? Because demand is already spiraling out of control. For the latest generations of chips below 3 nanometers, demand is three times higher than the available supply. We are talking about capacity already fully booked, and customers willing to pay 100% premium just to secure delivery. >> [music] >> In other words, the world is crucially lacking compute power. And all this before robots, autonomous vehicles, or AI agents have even been mass deployed. At this point, we are clearly facing a physical bottleneck rather than a software one. It's a matter of silicon, machinery, and above all, energy. Musk currently estimates that global chip production accounts for only about 2% [music] of his future requirements if Tesla, xAI, and Optimus meet their stated objectives. To be more specific, Musk envisions a scenario where Tesla produces tens of millions of vehicles per year, primarily driven by the robo-taxi. Whereas Tesla currently produces only 2 million cars annually. Along those same lines, Tesla only produces a few thousand Optimus robots, but when he speaks of a scenario where global chip production covers only 2% of demand, he assumes a world where there would be tens or even hundreds of millions of Optimus robots in active service around the world. So, here again, we are looking at a vision, to put it mildly. And this entire question of dependency Musk had already been seeking to address through Dojo, which was essentially Tesla's attempt to seize control of the computing value chain, with the objective of decoupling from Nvidia by developing an in-house chip. Optimized for training full self-driving models. But the problem is that even there, Dojo does not eliminate Tesla's dependency, it merely shifts it. Tesla would no longer rely on Nvidia to design and train its models, but remains tethered to TSMC for the actual chip fabrication. This is precisely the industrial chokehold that Terafab is designed to break. To truly grasp the ambition of the project, one must look at the orders of magnitude. On its own, Terafab aims to replicate roughly 70% of TSMC's entire global capacity at a single site in Texas. And yet, even then, while producing chips in Texas solves the dependency problem, it leaves another massive issue untouched. Heat. The more we compute, the more we heat, and the more we heat, the marginal cost of compute skyrockets. Consider that cooling alone typically accounts for 30 to 40% of a data center's consumption. The more massive data centers we build, the more this thermal management challenge becomes a blocker. This thermal wall, as Musk calls it, is what he seeks to bypass, and it is here that the project shifts into another dimension. Musk does not conceive of the Terafab as a single isolated factory, but rather a distributed system with a very clear separation. >> [music] >> About 20% of the compute remains on Earth using distributed processing, thus side-stepping heat dissipation issues. [music] And the remaining 80%, which would be centralized if produced on Earth in [music] conventional data centers, is sent into space. Why? Because in space, the laws of physics change completely. On Earth, we cool through convection with air or conduction with water. >> [music] >> In the vacuum of space, there is no air, but there is radiation. The hotter a system is, the more efficiently it radiates its energy. In other words, where Earth imposes diminishing returns, space offers increasing returns. Added to this is a second factor, energy. In orbit, solar irradiance is constant, without day-night cycles, [music] without clouds. Musk is talking about deploying massive solar capture capabilities [music] directly in space to power these infrastructures. So, we have a system where energy is abundant, cooling [music] is passive, and compute density can skyrocket. To truly understand, one must go one step further into the system's logic. What Musk is building is not just massive compute capacity. It's an architecture, an architecture where compute is separated by function. On one side, what could be called reaction, on the other, reflection. In practice, this translates into two types of chips. On one side, onboard inference chips directly in Optimus robots or Tesla vehicles. These must be fast, local, and capable of making decisions in real time without relying on an external connection. And on the other side, chips optimized for space, designed to operate at high temperatures, resist radiation, and maximize compute density. This is where model training, planning, and global optimization takes place. It is this separation that explains the 80 to 20 ratio. The 20% on the ground are not there for massive compute. They exist to minimize latency to allow physical systems to interact instantly [music] without relying on a round trip to orbit. This organization has a very tangible consequence. Today, a chip is designed in one place, manufactured elsewhere, then assembled with its memory in a third place. Each state adds latency to the production process. Complexity and energy consumption. With Terafab, Musk seeks to integrate this entire chain in the same place as he has done for SpaceX rockets. Design, fabrication, assembly. Result, fewer transfers, fewer losses, and an iteration speed that is much higher to widen the gap with competitors. But attention, Terafab is not an isolated project. What Musk is currently building is a complete industrial chain. From this perspective, Tesla is no longer an automaker, it is the commercial outlet for Terafab. With a target of 100 million Optimus units, Tesla provides Terafab with a guaranteed internal order volume that helps to amortize, at least in part, the $25 billion investment. Where Intel must search for clients, Terafab already has part of its order book filled for years to come. This may seem far-fetched, but in reality, he is replicating the history of SpaceX and Starlink. Understand that a major barrier to competition to develop 100% reusable rockets is that no other space company in the world had a sufficient volume of orders to justify 165 launches per year as was the case for SpaceX in 2025. SpaceX can pull off this feat among others for there are many feats in SpaceX's success because Starlink, its own telecommunications satellite subsidiary, accounted for 75% of those launches. When Tesla becomes the primary outlet for Terrafab, Musk is trying the same SpaceX Starlink play again, which proved immensely successful. It may seem crazy, but at the same time, this man pulls off truly insane things. So, well, uh >> [laughter] >> And here, once again with Terrafab, SpaceX is again the linchpin of the system. With its new rocket in development, Starship, Musk is seeking to industrialize space transport. Version 3, whose inaugural flight is expected in the coming weeks, aims for a theoretical goal of launch cost below $10 per kilogram at full maturity. As a reminder, the cheapest competitors on the market are at roughly $6,000 per kilogram. Now, you might say he isn't at $10 yet because many steps remain before Starship is finalized. But still, we see Starship fly, we see it re-land, and little by little, it traces its path. And here, the gap with the competition becomes vertiginous. At this cost level, $10 per kilo nonetheless, perspective shift radically because sending hardware into space becomes a truly rational option relative to the billions in investment, the electrical consumption, and the physical constraints linked. But SpaceX does not provide only the transport. Let us not forget the other major asset Musk has in his pocket. Once again, Starlink. Starlink would provide the network infrastructure required to link these orbital brains to ground systems with latency low enough to make the whole viable. Without Starlink as well, Terrafab would remain science fiction. It is precisely through Starlink that this compute power becomes a global resource accessible everywhere independent of local infrastructure. This is what allows, for example, to imagine an Optimus robot functioning with the same intelligence in a factory in Texas or in the middle of nowhere without depending on a terrestrial network. And finally, there is X.AI that should not be forgotten. With the development of its models like Grok, X.AI is engaged in a compute race at an absolutely unprecedented scale. We are talking about training models that are more and more massive, but above all, real-time interference at a large scale. This is precisely from there that the demand for Terrafab comes. X.AI is not a simple user. It is one of these structural motors of compute consumption. X.AI brings the layer that was missing until now, the reasoning layer. Where Tesla provides the body, the robots, the cars, X.AI provides the brain, a decision level capable of planning, interpreting, and acting in complex environments. And it is this layer that allows the transformation of an autonomous machine into an intelligent agent. With projects like digital Optimus, it is clearly understood that what Musk wishes is to create a complete loop. Compute is produced by Terrafab, it is leveraged by X.AI, and it is embodied in the real world by Tesla, propelled by SpaceX, and linked to Earth with Starlink. And I will even tell you, as surprising as everything I've just told you is, we are likely still underestimating the true scope of this project. In Musk's rationale, Terrafab is not the end goal. It is an intermediary milestone. The next step is to industrialize space from the moon. Why the moon? It always comes back to the same point to break through ferrous. Gravity [snorts] there is six times weaker than on Earth. There is no atmosphere, so no friction, and solar energy is abundant and continuous. In these conditions, it becomes possible to imagine a totally different infrastructure. Robots, accompanied by humans, capable of operating directly on the lunar surface, to build industrial systems. For instance, electromagnetic mass drivers capable of launching payloads into space without rockets this time, directly from the surface of the moon. You launch an object hard enough, and since there's no friction in the atmosphere, it will place itself directly into orbit. Or in deep space, and from there, the cost of space access doesn't just drop, it mechanically trends toward zero. But all this brings us to two deeper reflections. The first is that Elon Musk's company seem to be heading toward a single entity, and a merger of the whole is more and more in mind. However, in such a case, we are talking about a highly diversified conglomerate spanning automotive to aerospace, including artificial intelligence, robotics, and a social network all for several trillion dollars in valuation. This represents incredible firepower, but I still think that what truly distinguishes such a group from what the competition can produce is SpaceX. Whose vocation is to crush by several orders of magnitude the cost of access to space, something the Chinese themselves lack. My second reflection is that it seems Elon Musk is unveiling more and more a vision of the future, where consciousness will be projected across the solar system, but where the human being, as we know it today, will have been marginalized, or will have even disappeared from view. Let me explain. Space conquest, by its nature, strongly excludes a human presence, for a human being is totally ill-suited for a prolonged stays in space. So, yes, we have a good chance of reaching the moon and other planets, but this we in question will more likely be something conscious than something human. The statement that best supports this idea is the response he made to Larry Fink during an interview at the Davos Forum, where he discussed abundance for all. Larry Fink essentially asked the following question, "In a world of total abundance, what would be the remaining role or meaning of life for the human being?" And Musk's answer was very clear, "You cannot have both." But but I mean I mean it is it it is a a necessary um Like you can't have both. You can't have work that has to be done uh um and uh amazing abundance for all. Yes, one cannot have total abundance and find a meaning to life, because what gives meaning to life is to confront our limits. Yet, in total abundance, of limits, there are none left. When Musk announces that AI and robotics will saturate the needs of every human being and that consequently everyone will have everything they desire, this means that a more complex system and thus more intelligent will sit above and that it will be this system at the helm of the world, just as your Labrador is in the care of its master, who guarantees that all its needs are satisfied. Another confirmation of this idea is when Musk announces that in a world of total abundance, currency will no longer exist. >> In Banks and His Culture Books has it pretty much right, where there there actually isn't money in the future and there is abundance for for everyone. If you can think of it, you can have it. That's it. This may be a detail for you, but for me, as you can imagine, it means a great deal, because the very essence of money and human existence is to arbitrate uncertainty. Since we are always confronted with a host of variables we do not understand, we invented money. To say [music] that whoever has wealth disposes of the capacity to allocate resources where they believe they're most needed. Therefore, a world without money, and I am persuaded that Musk is well aware of this, is a world where all choices of resource allocation are arbitrated a level above us. In consequence, it is a world in which we can be happy like a Labrador well treated by its master, but in no way free. Because to be free, it is to be able to make choices and thus be confronted with uncertainty. Put differently, Musk is perhaps building what comes closest in science fiction to Cyberdyne Systems in the movie Terminator. The film Terminator presents a dystopian vision of the arrival of AI and robots, but if we go up one level of abstraction further, Cyberdyne in reality is the company that gives birth to an intelligent superior to that of humanity, Cyberdyne. It is the end of human supremacy. And these questions, I feel they weigh heavily on Musk, for he does not fail to explain that the essence of his nightmares, the most terrible ones where he wakes at night, concern dystopian scenarios about AI. Moreover, in its day, the Open AI project that he co-founded was intended to promote a collective development open source model. And then with what happened at Open AI, and with the turn that the sector has taken in the world, Musk resigned himself to passing from a collective whistleblower approach, since at the time that was the case, to that of a competitor who wants to develop AI first. For as he says so well, if he cannot moderate this movement of AI development, then he does not intend to remain in the position of spectator. I I previously advised that we slow it down. I've said that many years, and and I was like, okay, then I finally came to the conclusion I can either be a spectator or a participant, but I can't stop it. My bet, if I had to make one, is that Musk has resigned or is progressively resigning himself to taking the transhumanist turn. We shall see if I am right by following upcoming developments of his company Neuralink in particular. And if you want to learn more about the global vision of the man who could also become the first trillion-dollar man, go watch this video where I explain the secrets behind a 1,000 billion-dollar bet.