The Musk Machine: How Tesla, SpaceX & Grok Plan to Take AI to Space

Kind: captions Language: en In January 2026, SP X [music] asked American regulators for permission to launch up to 1 million satellites. Not for internet, for artificial intelligence, giant solar powered computers flying above the Earth. It sounds like science fiction, but it is one piece of a much bigger plan. Rockets, satellites, chatbots, chips, cars, and robots. Elon Musk's companies are slowly connecting into [music] one machine and its fuel is AI. I am Evan. [music] Today we open the blueprint. Five layers. The rockets, the network, the brain, the chips, and the bodies. And then the end game, plus an honest reality check because Musk's timelines have a habit of slipping. Layer one, the rockets. [music] Everything starts with getting heavy things into space cheaply. SP X already launches more rockets than the rest of the world combined. In 2025, [music] it flew 165 orbital missions, more than half of all launches on the planet. And in September 2026, its giant Starship reached orbit for the first time. A fully reusable rocket is the key that unlocks everything else in this plan. In February 2026, SpaceX merged with Musk's AI company, XAI, [music] the maker of the Grock chatbot. The deal valued XAI at about $125 billion. Then in June, SP [music] X went public on the NASDAQ at a valuation of about $1.77 trillion in one of the biggest stock market debuts in history. Layer two, the network. Starlink is already the biggest satellite network ever built. In March 2026, it passed 10,000 satellites in orbit. That network does two jobs. [music] It brings in billions of dollars from internet customers, which helps pay for Starship, and it gives Spac X the experience of building and flying satellites like a factory by the thousands. Layer three, the brain. [music] The brain is Grock, built by Exaii. To train it, Xi built Colossus, [music] a supercomput in Memphis, Tennessee. In January 2026, Colossus 2 became what the company calls the [music] first gigawatt scale AI training cluster. That is roughly the power used by a large city. Together, the two Colossus sites run the equivalent of more than 1 million topend AI chips, and they are training the next model, Gro 5. [music] And here is the problem. AI is hungry for electricity. Musk says there is only about 20 gaw of AI computing in the whole world today. [music] Independent researchers estimate around 30. He says his companies will eventually need hundreds of gigawatts [music] on Earth and up to a terowatt in space. That is tens of times more than all the AI computing that exists today. And Grock is not just a chatbot on a website. It is built into the exocial network. Since 2025, it has also been rolling out inside Tesla cars where drivers can simply talk to it. [music] In the future, Musk wants the same kind of AI to see through car cameras, to speak through Optimus robots, and to think inside satellites. One brain, many bodies. Layer 4, the chips. You cannot run that much AI without chips. So, Musk wants to build them himself. In March 2026, he announced Terra Fab, a $20 billion chip factory in Austin, Texas, [music] a joint project between Tesla and SP X. It will make two kinds of chips, one for cars and robots, [music] and one hardened to survive radiation in space for orbital data centers. Tesla is also designing its own AI processors. i5 is being made by TSMC. I6 comes from a 16.5 billion dollar deal with Samsung. And in January 2026, Musk revived Tesla's Dojo project with a new purpose. He wrote that the next generation Dojo 3 will be for space-based AI computing. Layer five, the bodies. AI in a data center is only half the story. Musk wants AI to move through the physical world. On the road, that means robot taxes. Tesla says more than 1.7 million people now pay for its full self-driving software. [music] And then there is Optimus, Tesla's humanoid robot. [music] Initial production has started on a new line in Fremont, California. Deliveries [music] to outside companies are expected around the end of 2027. And the robots are meant for space, too. Musk has said Optimus could ride a Starship to Mars to prepare the ground before humans arrive. Robots do not need air, food, or a ticket home. [music] So, what is the end game? Put it all together and the blueprint looks like this. Terrafab builds the chips. Starship launches them cheaply into orbit inside solar powered satellites. Those satellites run Grock and other AI in near constant sunlight connected by Starlink lasers. And on the ground, the same AI drives Tesla cars and moves Optimus robots. [music] Why space? Because in orbit, the sun shines almost all the time. No night, no clouds, no fights over land or power lines. [music] And then the moon. In February 2026, Spac X shifted its focus from Mars to a sustained base on the moon with an uncrrewed lunar landing targeted for March 2027. Now, the reality check. [music] First, physics. In space, there is no air to carry heat away. [music] Cooling massive computers in a vacuum is one of the hardest problems in the entire plan. And radiation slowly damages electronics. [music] Second, the sky. A million more satellites would multiply the risk of collisions and space debris. Regulators are expected to cut that number. Third, the timelines. In 2025, Musk talked about sending Starships to Mars by the end of 2026. That mission has now been postponed. And money matters, too. [music] In the third quarter of 2026, Tesla's operating profit fell 56% [music] and its long-term sales target was cut to 5 million cars a year around 2030. Let me sum up the blueprint. $1 trillion plus rocket company now merged with an AI lab, a million satellite AI constellation on paper, a 20 billion chip factory, robots on a production line, and a moon base on the schedule. Visionary plan or the most expensive science fiction ever written? Tell me what you think in the comments. I am Evan. See you in the future.