Elon Musk is sandbagging SpaceX - Open AI, Anthropic, META, Goggle will all stall on compute @mxux

Okay, we're going to um take a look at Terapab. I um I did a presentation a while back and I'm going to try to intercut it. Uh this is Elon Musk's presentation on Terraab with that sound. I did a presentation a while back on Terraab and on the whole orbital AI um SpaceX Tesla joint partnership and the Terraab and uh we may go over that in a second, but basically what it came down to as far as the success or even feasibility of this project uh to meet the timeline that uh Elon Musk has set out which is uh to have orbital AI established in 2028 in a meaningful level to uh garner market share. There's two things that have to happen. Uh, one is they have to have a a launch vehicle that they can turn around basically and launch hundreds of times. A quick reuse launch module. They have that. Uh, it just flew. It's going to orbit. Uh, it's Starship 3 or whatever. The point is the technical I believe uh the technical aspects of this project the launching uh the launching of the placement of the satellites in space the design of the orbital AI satellites I think that's a math problem and a money problem and um SpaceX just did a oh almost a billion dollar bond offering and that's on top of their uh you know two trill IPO. So I do not believe there is a problem with the space aspect of this project. The other thing they have to have, and this is this is the whole idea of Terraab, they have to be able to have a supply, a supply chain, an uninterruptible supply because they're going to be launching hundreds of satellites, okay? And in quick succession and they have to have chips. They have to have chips uh for the satellites. They have to have uh chips for um the uh Tesla robot Optimus. And they have to have chips for the robo taxi. Um as you know, the chip production is sold out and and reserved for years. Uh I don't know what the workarounds are on that but even the memory micron is has sold out memory for two years. So this this is the key stumbling block is the chip supply that is the critical factor. I think the engineering is there. I think the talent is there. I think Musk has proved he's a really good CEO. He knows how to uh manage money and manage large organizations and technical organizations and you know he's the new Howard use or he's one of the the best uh you know hands-on technical CEOs you know probably ever. I you know I don't think Bill Gates can compare but anyway the point is so they have to have these chips and they're planning on building Terapab which is going to be a monstrous uh location in Texas and uh it's going to be like 10 million square feet. You can look up the stats. It's going to be the largest of its type and it's going to be built in stages. one one building after another and I think that SpaceX I mean Tesla and Musk certainly can do this construction. Okay. The question is and this is the question I raised in the video I made on this topic. the Lamb uh Lamb uh research uh UV lithography machines that make the chips are on back order. They only make 50 of them a year or whatever. It's uh there's a bottleneck there. And they are partnering with I with Intel and I don't know if Intel has access to pre-made orders on those, but that is a serious problem. There is no way they're going to fill these fabs up with these lithograph machines because there aren't any. And uh you know um there are other constraints as well. They they say that conventional metrics say that if they're going to build and let's say a 2 nanometer plant, I don't know what they're going to do, 3 nanometer, whatever. I think IBM just did below one nanometer. U let's say a 2 nanometer plant, uh chip foundry like TSMC. Um they're saying five years. five years uh to construct that TSMC. Uh they just built one in Arizona. They used a method called copy exact. Almost drove them crazy. They made a exact down to the millimeter nanometer copy of their facilities in uh Taiwan including the water supply and everything else. They're saying that uh you know this is this is a uh you know it's a five-year project anyway and at the at the scale Musk wants to do it's the cost and I some reliable estimates put a conventional build cost at $3 trillion. Okay, that's $3 trillion. Okay. So that's I don't know they can probably raise that but the time frame of 5 years is going to put them out of Musk wants to be launching in in uh you know a year and a half. He feels that he will not he will lose market share to competitors if he does not have a working system up by 2028 and he wants to start launching uh basically next year. Um so I said you know I you know is this another one of Musk's uh you know high flulutin ideas that is going to be delayed and delayed and delayed and I if you do the conventional research on the building of these fabs and fab chip fab foundaries that's all very true. However, this is the breakthroughs uh breakthrough throughs multiple that is going to make this a successful SpaceX is going to be able to do this. Tesla is going to be able to do this. They're going to own the full stack on everything. And uh they're going to use uh two different methods uh to do this. One is they're going to make AS6 chips application specific integrated circuits. These are not going to be Swiss Army nice circuits. Even even the GPUs really they're they are a sick but not really. Okay. So for each application the uh orbital AI uh Optimus robot robo taxi Tesla etc. Um they're they're going to make specific chips. These chips are going to be simpler uh than a full functioning you know Intel PC processor. They're going to be optimized for production. I would be willing to bet they're even going to go with a three or five nanometer scale because if they're going to be specific, they can make them bigger. Musk has mentioned two things that are critical. The reason, one of the reasons I think it's going to be a larger nanometer scale is because um Musk mentioned that we don't need clean rooms. We're going to have dirty rooms. And uh you know when you're at higher nanometer tolerances you know you can tolerate more dust. That's just my opinion. But the second thing is having to do so they you know they can build the factories. Uh you know Musk can engineer the factories and build. Nobody builds factories quicker or better than Tesla right? Uh you know they can design the A6 chips. Okay. uh they probably already have designs. The question is um how are they going to come up with the you know $3 billion or $3 trillion I believe was the quote to to pay for all this and the answer is they're going to simplify everything. And the main thing in my mind and the clue to this is Mus said we are experimenting with new physics right now. uh to accomplish this task. And they have a prototype system working off in a corner at the uh Texas plant there. They've built a a little experimental lab. They have a thing called X-ray lithography. Right now they use uh ultraviolet light lethography with the LAN machines. And what they do is they have a giant laser light apparatus inside the machine. It's actually the biggest component inside those lithograph machines. They put a drop of tin in a chamber. They hit it with a laser. It explodes. It creates this ultraviolet light and they use that because it's such a small wavelength uh to etch the integrated circuits on the silicon discs. Okay. The the it's a lithography thing. Okay. And that's very complicated and very expensive and uh this is the state-of-the-art at the moment. Now there is a new way of doing this that's faster, cheaper, better, more efficient. Uh the Russians are doing it. I'm sure China has something going on. There's some experimental projects. There is one company that's working on this that is not public yet. Uh but they want to build a prototype plant. I don't know. Musk may be working with them. Who knows? Maybe they will buy them. I don't know. But this is what I suspect is how Musk is going to solve this terap problem is they're going to use X-raybased lithography. Rather than having the uh ultraviolet light, they're going to use X-rays to generate these uh beams to uh do the lithography on these chips. And the way this works, and this is why I think Mus is going to use it, cuz he said, "We're playing around with physics right now. One, there are several different types of this going on. The one I think that he's most likely to use actually uses a small linear accelerator, you know, like CERN, okay? And it accelerates atoms uh at the speed of light and u it splits them and it turns them into X-rays and that is the X-ray that they use to etch uh the chip. It's much simpler than the current uh lamb research method. And the thing about it is this is what really struck me. um they can have one linear accelerator and this is a small scale. This isn't certain level. This is probably fit in the warehouse or whatever. They can have one particle accelerator running and it it can it's like a a steam line or something. They can port that output to many different locations. Okay. And it will work. Uh so in other words uh and I'm not saying they're going to do this but like I said the lamb machines are sold in in advance there's no leeway in the ordering unless Intel has some and maybe Tesla has pull but if you get the lamb machine without the ultraviolet light generator which is the largest most expensive part of that machine if you just have the optics and then instead of using that ultraviolet light, you simply port this X-ray beam uh into the machine and use that. It cuts the cost of the the LAN machine or some similar device with the LAN machine, let's say, down to the optics. It's like 50% maybe more than 50% of the cost is to in those machines is to have this device to generate this ultraviolet light. You don't need it. You poured in the electron beam with the X-rays and you can you can have uh 10 lamb machines working in 10 different spots on the plant and they can all be fed by one linear accelerator and uh they're much simpler. Um, they they're not competing with anyone cuz nobody else is using this. If Lamb chose I'm not saying Lamb is going to do this, but I'm sure they could come up with another way of doing this. But the point is they just order the lamb machines without the light source, okay, which is the most difficult thing and most precise part of that whole apparatus and most expensive. And they just have the lens work, okay? Uh Zeiss makes these lenses, okay? And uh and they port this electron beam, this X-ray uh beam into these different uh foundry machines, lithograph machines. This is how they can do it. This is how they can scale production quickly. And by the way, these X-ray uh lithography machines, Kitty, what's the matter, baby? Baby, come here, baby. Come on. These X-ray lithography machines are they can do all kind of stuff. Multiple layers. They're much they work much quicker. This is one workaround. I think that this is what Mus is going to do. Some derivation of this. So, in other words, a hybrid lithography machine. And um this is this is going to do it for them. This is going to enable them to uh at a more reasonable cost scale uh this effort quickly. So in other words um yes SpaceX and Tesla can do this and the terapab uh is uh the chip supply is the key factor here in all of these Tesla and SpaceX projects. Um once Musk gets this up and running and I'm sure they're going to do it a piece at a time. Uh Musk is going to have uh you know he's going to own the full stack uh you know from the wafer to the to the IC chip and so forth. And by the way going to make memory too. And by the way I'm of the opinion dojo has been is resurrected in the background and I and this is like the serabus chip. It's all on one chip. This is perfect for a space application. memory, ICU and so on and so forth on one chip. So in other words, um this is all doable. This is not pie in the sky. It's not impossible. It's not I don't even think it's improbable. And I think that it's achievable financially. And I think at that point I did another uh video um which not a lot of people saw about the advancing S-curve. Um we're at a a a pivot point in the whole AI race right now. What people are missing, you know, there's an S adoption curve and they have worked this out for years for technology and so on and so forth. The thing is AI is moving so quickly it's compressed that um adoption curve and we are now far ahead of uh where we should be uh where we we were supposed to be and we're at the point where there's according to the S-curve theory with this new advanced uh speed it up calculation we're at the point where there there should be a a gigantic change in the whole thing. Now uh Liupold Alultz uh he has he has mentioned uh two things but he's short every chip manufacturer um and he's long power and facilities um cuz he feels that that is going to be the limiting factor and uh nimi and so forth by putting these uh AI nodes in space. It solves all these problems, the power, the real estate, everything. And by the way, also is Leopold is short Nvidia at least as of his last last F13. So the point is this is all coming together. Um this Scurve is going to uh accelerate. you know, we're we're simply not and there are workarounds being undertaken now, but we in the United States are not going to have uh the electricity that we need to power all of the output that we need to have. And I do believe Oracle is at like 98% of all the capacity they have is under contract. They have no room to grow. We're moving past the LLM model and we're moving into the inference stage. And you know, Liupold uh just did a um a short I saw on the data wall. And he is wondering if we're going to hit the data wall. And by that he means um you know, they've already copied basically everything on the planet and put it into the LLMs. And according to him, he said you can do that about 16 times. And each inference, each iteration uh gets a little better, a little better, a little better, and then at some point it doesn't get any better. You hit diminishing returns. It's like, uh, you know, copying a VHS tape over and over again. At some point, you got nothing. It's mush. and he says there's optimizations and and different models that can be run and everything, but he's thinking out loud that this is going to happen in which case inference is going to be uh you know the uh the go-to the LLMs uh are uh you know they're done for right now anyway, but anyway, you can look into that yourself. The point of this video is that SpaceX and Tesla are going to be able to do this. They are they're going to be able to do this orbital AI. It is going to happen. And as I said, I went through every variable and the the bottleneck, okay, is the chips uh that are necessary to accomplish all these projects. With the Terra Fab, they're going to solve those problems and own the stack and keep costs down. Okay? And you know, can they can they do it? Can they get this up and running within 2 years? I think they can with these new technologies and simplified specialized chips. I think they can do it and I think they have the money to do it at a re, you know, that the cost is not going to be the three trillion that the experts are saying because they're going to use this simplified fab method and simplified chips and likely a larger nanometer. They're going to have the money. They're going to have the talent. They got uh Intel there on board. Um they've got this new X-ray lithography technology. There are other versions of that as well. And they have the money, they have the talent, they have the time. Um so I'm bullish. I'm bullish on SpaceX. I'm bullish on Tesla and I'm bullish on Terraab. I think that uh as as crazy as it sounds, I think this is all a I don't know. I mean, I've looked into this pretty carefully. If you can you can look at my video catalog. There's several videos I've done on different aspects of this. And with this newest information, I could see this, you know, coming to fruition. I do believe this is a done deal. See Mus snickering there, grinning. You know why he's grinning? Cuz he knows he's going to pull it off. He's going to pull it off. And um moving the AI into space is going to it's just going to lower cost and solve problems and uh really make it more practical. But anyway, I don't know. Uh as I said, Liupold is or is his is last F13 or whatever the filing is is is shorting all the chip makers. my video I did the analysis of the S-curve has accelerated. There is going to be a you know a ch it's time to advance this whole AI thing and move it into a new scurve to start a new scurve and that's early. uh people are thinking that uh you know it's not it is it's early uh it's earlier than people think I think um uh this reality is the new the new scurve I think is terapab is orbital AI I think is Musk and uh who knows maybe quantum in space as well at some point but uh uh the point is uh very bullish um these terrestrial models are going to be like an old dried up calendar. Musk said he wants um this system up. He said if I don't get it by 2028, if I don't get a get a system up, we're going to we're not going to be able to capture market share. I think they're going to be able to do it. I think they're going to, you know, we'll see. I think they're already advancing the construction of empty buildings for the terapab. Now they do have the lab running but as I said um this company one in particular they feel they've pro project u perfected this and they want a demonstration project they want to go into production they're ready I think this is what is doing all right this is MXUX good luck in the Good job.