Why Does Musk Need Chips, Robots and the Moon?

Kind: captions Language: en A billion [music] robots, computers in orbit, factories on the moon. All of that appears in the plans of Elon Musk's companies. And in Texas, [music] Tesla has already started building a research facility for making chips. But what is the moon doing in a story about chips and robots? [music] How do we get from a factory here on Earth to Mars with an entire space industry in between? A billion robots is still a stated goal with no deadline. But behind that huge number is something more interesting. [music] the connections between these projects. Let's follow them from a Texas construction site to Musk's proposal to make satellites out of lunar materials and see what this could mean for the work you and I do here on Earth. Let's start in Texas. It's easy to confuse two different [music] sites and come away thinking an enormous factory is already finished. In April, construction began on a research facility on the north side of Tesla's factory near Austin. [music] The idea is to develop chip manufacturing processes there. In its July update, Tesla described early construction work and equipment procurement. [music] Then in August, a second location entered the picture, Grimes County. That's where the future Terraab industrial complex was announced. [music] The first phase comes with a proposed investment of roughly $17 billion. That's an enormous amount of money. But the main question is still very practical. Can they first master the manufacturing process, [music] then produce the chips they need in huge numbers? The next links in the chain depend on it. [music] So why take on something this difficult? Why not just order more chips from companies that already know how to make them? SpaceX points to future demand. By the company's calculations, it and Tesla will need far more computing power than is available today. [music] They still expect to work with suppliers, but they also want to build additional manufacturing capacity of their own. And that's our first connection to the robots. You can design an impressive machine. You can teach it a spectacular trick, but if you plan to build enormous numbers of those machines, [music] you need enormous numbers of electronic components that perform consistently. And designing a chip and manufacturing it are two separate challenges. Picture a thin silicon wafer. [music] Tiny structures and connections are built onto it layer by layer. Extremely sophisticated optics transfer the patterns. [music] Even contamination we wouldn't notice can ruin the result. Then you have to test what you've made. connect the computing components to memory and put everything into a working package. Terrafab is meant to bring those stages together at one site. [music] The companies say that should speed up the cycle. Design, manufacture, test, improve. Intel is a partner in the project, but even an experienced partner can't make a new factory operational at the push of a button. Once the building is up, it still has to produce highquality products again and again. Now, imagine the chips they need are finally ready. [music] Where do they go? One intended destination is Optimus, Tesla's humanoid robot. Take an ordinary box. [music] Picking it up off a table is almost effortless for you. A robot has to work out where the table ends, where the box is, [music] which side to grab, and how tightly to close its fingers. Then, it has to shift its weight, keep its balance, [music] and avoid hitting the person beside it. All of that happens in a single movement a viewer could easily overlook. Tesla is developing vision, movement control, navigation, and object handling for Optimus. Its onboard computing hardware is supposed to let those systems run inside the robot itself. It helps to separate training from use. Large computing systems help train the models. The onboard computer then uses a trained model to make sense of what's happening around it. A factory of their own could supply more of the right hardware. The skills would still have to be developed separately. More chips won't suddenly teach a robot to unload a dishwasher without breaking anything. [music] And Tesla included a revealing detail in its July update. The first robots off the new production lines are meant to go to its own Optimus Academy to collect data and practice skills. [music] In other words, before sending the machines out to work, they still plan to teach them. Now, let's [music] zoom out. Why does Musk want so many machines that can work with their hands? In part four of Tesla's master plan, labor, transportation, and energy sit alongside one another. [music] The company describes a future where automation makes goods and services more affordable. Robots are supposed to do physical [music] work. Autonomous vehicles move people and cargo. Energy systems power that [music] infrastructure. Looking at these plans together, here's how I read the idea. bring as much as possible of what the other projects will need inside the companies themselves. Chips provide computing power. Robots are supposed to do physical work. Transportation moves people and cargo. And all of it needs energy. Which brings us to the next question. Where does the electricity for all that new computing [music] capacity come from? On Earth, you also have the cost of infrastructure and cooling. One of SpaceX's [music] answers takes an unexpected turn. send the computers themselves into orbit along with solar panels. [music] In February, SpaceX acquired the artificial intelligence company XAI, an AI developer and a company that sends cargo into space are now under the same roof. The connection between computing and rockets is becoming quite literal. SpaceX intends to use some of the future chips in space-based data centers. Picture a satellite that draws energy from the sun, processes data, and sends back the results. The project's website describes laser links between the satellites with data sent down to Earth through Starlink. To massproduce them, the company is building a separate factory in Bastrop, [music] Texas. It hopes to begin volume production as early as late 2027. SpaceX expects solar power in carefully selected orbits to give it an advantage. But space has rules of its own. In a vacuum, you can't just blow air over hot equipment. Heat has to be carried away and radiated through specially [music] designed surfaces. Those surfaces, the electronics, and the panels all have to be built, delivered, deployed, and made to work for a long time. And that's where the rocket comes back into the story. Starship is being developed as a system with a reusable spacecraft and booster. SpaceX hopes frequent flights will make it cheaper to move enormous amounts of cargo. Now, the connection is clear. Chips need a ride into orbit. The rocket needs payloads and frequent flights. A computing network could [music] become a major customer for those launches. But even that isn't the most ambitious part of the idea. In its published prospectus, SpaceX describes a further step, building satellites on the moon. It proposes sourcing a substantial share of their mass from local materials while bringing chips and other relatively lightweight components from Earth. To launch the finished satellites, it's considering an electromagnetic launcher. Think of an enormous machine that accelerates a [music] payload to the speed needed to leave the lunar surface. This is still a concept for future manufacturing, but it explains how the moon ended up in a story about a Texas factory. Picture the chain. Texas makes the electronics. Rockets deliver equipment. Lunar industry uses local materials. The new satellites expand the computing network in space. And in SpaceX's February announcement, Musk links the growth of space-based data centers to funding future bases on the moon and a civilization on Mars. So space has two roles in this plan. A place to do business and a direction for further expansion. At least that's how Musk himself [music] describes it. Back on Earth, an uncomfortable question remains. If machines do a large share of the work, who gets the income from that work? The robot's owner? Its manufacturer? Whoever sells the computing power that makes the machine useful for our future, that could matter more than the date of the first city on Mars. It's a fascinating picture, but it has a weak point. Every link in the chain needs the one before it to actually work. The factory has to make good chips at the right price. The robot has to be useful without constant intervention. The rocket has to fly often enough. And the space-based computing network has to make financial sense once manufacturing launches and failures are all accounted for. On paper, you can draw an arrow between two projects in a second. [music] In reality, that one arrow may represent years of work. So, this plan doesn't yet tell us when robots will be everywhere. Here's a more useful test. Look at how long a machine works on its own, how often a person has to help. What happens when its surroundings change, and what a completed task really costs, including maintenance? Say a robot spends all day moving identical boxes. Then the handle on one box is torn. Does it keep working, stop, [music] or drop the load? These seemingly random situations are what determine whether the technology is ready for an ordinary business because a business needs results every day. and impressive movement in front of a camera isn't enough. So, where do people fit into this system? And how can we benefit from any of it if we don't own a factory or a rocket? Tesla already has a job posting for an Optimus data collection operator. [music] It's a small but telling detail amid all the talk of a billion robots. Machines need to be trained, tested, [music] maintained, and integrated into real work. If you like working with hardware, look into equipment maintenance. If you're more interested in software, look at how it's used in real production. But start with demand, a problem someone actually wants solved. Imagine a small warehouse. Every day, an employee checks a delivery against its paperwork. [music] Sometimes they make a mistake, then have to track down the discrepancy. Before offering a humanoid assistant, understand the task itself. How long does it take? Where do errors happen? [music] And what could a camera, a scanner, or software already make easier? Pick one task like that for the coming week. With the owner's permission, run a [music] small test. Measure how much time you saved and how many errors remain. A consistent result gives you something concrete to discuss with a potential client. And the experience stays useful even if Musk's plans [music] change. Now, we can return to the question at the beginning. The moon enters this story as a potential manufacturing site. In SpaceX's vision, local materials would help produce new satellites while factories on Earth supply their electronics. That's how a Texas construction site becomes part of a much longer plan. Computing, physical work, energy, and transportation are meant to support one another. In its most ambitious form, that system extends beyond Earth. There's an enormous distance between here and that future. Some construction is real. Individual technologies exist. Many of the connections still have to be made. But the next time you see an optimist walking, remember this. Musk wants to build an entire industry around machines like it. And if that industry works, we'll have a very downto-earth question to answer. What work will we do in that