AI Bubble, Robotics, and Real Estate — Full Breakdown & Transcript

[00:00] You are not going to want to go to sleep until you watch this whole video. Actually, we got a lot to cover in this video. We are going to talk about what just happened with Opening Eye.

[00:12] Is it bullish or bearish, especially in context with what happened with Anthropic earlier. Then we're going to talk about model compression, which is a really big deal, especially as it relates to high bandwidth memory and bandwidth.

[00:25] That's going to be very particular for who the winners and losers are for artificial intelligence. Then we're going to talk about figure robotics and what they just showed us this week and my initial reactions to it,

[00:37] along with some hints around where I'm looking in the supply chain when it comes to robotics. Then we're going to talk about Buffett. Then we're going to talk about real estate. That is the scope for this video, so there's a lot to get off my chest.

[00:50] it's Friday night and you might ask yourself why? why are we doing this on a Friday night? why do you have to stay up for this? well in exchange for all the information in this video I will take the opportunity

[01:03] to say we are literally raising the price on the Meet Kevin Alpha membership $900 as soon as this coupon expires so you go to meetkevin.com meetreinvest.com it doesn't matter you join

[01:15] right now you will get graciously rewarded you've got some Really cool things coming for course members. Get all nine courses, every trailer. You already know the pitch. You can read it on the website. But I'm going to do you one better.

[01:27] I'm actually going to show you what we said in the alpha report this week every single day. And after this, we'll get into the actual goodies. First of all, remember how we went into the week and we're like,

[01:39] Okay, you got to buy on the QQQ. QQQ, which is an index, is at a 721 close. We literally sent a buy alert within a fraction of a point of $700 on the bounce.

[01:58] That's the lowest point you see right there. We crushed it on the Q's. But I'm going to show you the wins and the misses this week. This was Monday. Okay? Monday. We went into the week saying we expect software to run today and chips not to.

[02:14] and from open, some of these stocks went up 6%, some of these stocks went up 14% from the levels where they were at open. So in other words, like, we sent the alpha,

[02:28] the market opened, and then these things zoomed. That was a fantastic call on Monday. We called for a top support at 702.87, the low of the day was 702.74.

[02:40] We said we were in peak fear days, Monday and Tuesday, The Alpha Report nailed it on Monday. We didn't get anything wrong on Monday. Sundays we get stuff wrong. You'll see that.

[02:52] Tuesday, we got something wrong. I called in the Alpha Report, which why did you have this work? And I'm going to show you the wins and the misses for this week. This is what you get, okay? Every day, right around 5.30 a.m., I send the Alpha Report before the market's open.

[03:05] Levels, tests, short-term, medium-term, long-term plans, trades, ideas, investments. buy sell alerts we sent two $100,000 buys this week

[03:17] we sent we sent a course member live stream out and we add some detail about here's what we think for the day and add some more context for the alpha short and long term analysis medium term analysis you know that's where you'll see things like

[03:29] don't buy open door it's going to be a bag holder and the thing's gone from like you know $6 to $10 in that range down to $2.50 all the courses plus the app alright that's what you get

[03:41] You already know that, though. Tuesday. Here's what happened Tuesday. We called for no rocket, no collapse. We were mixed on the day, because it was the day before the Fed.

[03:53] And so what happened that day? The Q's were actually down 0.7. We stated there was a 95% chance of a hike, and do not fear it. Be bullish on it. Mind you, the market was up 3% two days later after that hike came.

[04:07] Kind of important. Last dip by days. Now. Right? And we said that would be today or tomorrow. We ended up buying on Wednesday. Wednesday was the low. Hardware moves up slowly with the call.

[04:20] I did think Cyborg would give back some of its enthusiasm on Tuesday. It was actually still up a little bit on Tuesday. So, we were right about this. I was wrong about Cyborg. I thought after a 14% day we were surely going to get something up.

[04:32] I was wrong. Fed Day. Now, Fed Day was interesting. because we conditioned specific language on we could be 80% bullish if we get an idea of how far away from neutral we are.

[04:46] We got a lot of mixed commentary from Fed. Well, Chairworsh. We knew we were going to get a hike. And we literally called $700 would be hard support. And it literally bounced to the penny on $700 for QQQ.

[05:01] And we bought it as soon as we saw it. Buy alert $100K. Buy alert $100K. here are the times when we sent those. Money where the mouth is, baby.

[05:13] Oh. Alright. Almost done. Thursday. And I'm serious. We are jacking the price up massively because we're changing a little bit of what we're doing. And course members are going to win bigly

[05:27] who get in before this expiration. Thursday. Said PeekFear was open. Markets relaxed and go and up we go, I said. NASDAQ ended up 1.7%, best day of the week.

[05:39] Hardware day, I called. 7.15 to hold, and that hardware would go up. We had a 6% day on AMD. Micron was up 5.5. Marvell was up 5%. NVIDIA was up 2.5. We said watch 246 for a reject on Marvell.

[05:53] We actually ended up hitting 247. We rejected, and we came back down. But we did get there. AVGO did lose 3.55, though. And we've been pretty bullish on Broadcom. It just took a couple of extra days for it to really hit.

[06:07] Broadcom ended up closing at $357 on Friday. So we got back to that line. We got a nice push on Broadcom. We've been looking for it. However, so far, we have not gotten $227 on NVIDIA yet.

[06:22] I thought we would test it. We did end up getting up to $222. We did not get to $227. I was hoping for that breakout test by Friday. That was wrong. We did not get there. We only got to $222.

[06:34] Meta, my goal is we get to 700 by next week. Okay, we close at 666. The devil's number. So, we'll see. I'm optimistic about that one. Friday. Today was a good outflow for us.

[06:48] The test was holding 715 all day. We closed at 721. I did say this morning we were less likely going to get to 227. So, I try to follow up and say, okay, this is looking like it's starting to weaken or whatever.

[07:02] We still got to 222. but we were right to downgrade it. Regency is what we really called for today. And the Nasdaq popped mostly at the end of the day. If you tried to day trade today,

[07:14] it would have been messy because we only got the buy surge at the end of the day. So that's sort of in alignment with our medium-term bullishness. But on short-term, this is going to be a little tough one today. You have triple witching.

[07:27] People are fearful about the weekend. That's the kind of stuff we're talking about. Retail fear makes me bullish. There are a lot of people who are fearful right now. We're going to talk about some of that fear in a moment. Marvell, we've talked about once again, the reject call.

[07:40] And then we were not wrong today as far as my summary. I ran, just to make sure, I just ran all the alpha reports through AI. Am I missing anything? Are we being fair here?

[07:53] I think we're being fair. Okay. So that's it. The Alcoa report goes up $900 if you don't use coupon code WASHEDBALLS. Now we're done with that.

[08:06] Now let's talk about all the crap that we actually have to talk about. OpenAI. First thing, OpenAI. OpenAI just announced that they expect to burn $280 approximately billion dollars by 2030.

[08:19] That's insane. That's a lot of money burning. And it's not because they don't think revenue is going to be up. They think revenue is going to be up to $350 billion between now and 2030.

[08:34] Now, this is a little bit complicated how they're saying it, so, you know, I'll draw it out with a whiteboard. But basically, the bottom line is, they're expecting to spend a whole lot more money than they're expecting to make.

[08:49] This shouldn't really be a surprise, and that's, in my opinion, partly why we're seeing the in-propic IPO continuously get delayed, because this is not a time of euphoria right now. You may have seen my tweet.

[09:01] Maybe you don't follow me there. You can always follow me at RealMeCutton. There you will find the good old by-the-dip-face ringing the New York Stock Exchange bell. And what I'd like to point out is this right here.

[09:14] This is strange, right? But it is the MO these days. We talked about this in the Anthropic Delay video. I'm not going to belabor all the details of this. But when you get a company like N-Scale IPO-ing on basically a hope and a prayer,

[09:29] it's kind of weird that they literally say we have no plans to finance the $44 billion, or we have no, like, commitment as to how we're going to end up financing $44.6 billion to commit to our Anthropic deal.

[09:41] But Anthropic is basically saying, we're going to assume you're going to get that. We're going to market as if we already have that compute. And N-Scale is going to market the anthropic deal as if we already have the deal. But they don't have the money yet.

[09:54] So the financing has to come from somewhere. And when you condition it on an IPO, I do get nervous that N-Scale IPOs and then the puppy tanks. And then they don't get the financing and the bubble's over.

[10:07] Now, I want to be clear. I don't think the bubble is over. But I also don't think these companies are going to be profitable anytime soon. And it's unfortunate because the more money they make and the closer profitability they are, the longer this bubble goes on.

[10:23] We need to be crystal clear, and that's going to be evident throughout the whole video. This bubble will end. And if it ends while Kevin Warsh is the Fed chair, it's going to suck. And I'm not trying to create fear. I'm trying to say, just in the back of your mind, we can work together, we can make money.

[10:39] but in the back of your mind, you know, honey, maybe this month we're going to start paying off the cost. Maybe this month we're going to make an extra payment on the mortgage instead of buying the debt. Whatever, right?

[10:51] But now you're reducing a little bit of your debt risk. And that wins you optionality. You know, people always do this, oh, but, you know, you're only saving 2%. I can make more than that in the market.

[11:03] That's not guaranteed. And these numbers from opening eye tell us We have some serious tests coming up. That NCL IPO probably needs to happen before Anthropic even considers IPOing.

[11:15] And what is Opening Eye telling us? Opening Eye is telling us that they are expecting to make $350 billion. And that is oh sorry they expected to be negative What do they say here Expenses will far outrun its revenues which are projected to grow tenfold from billion this year to billion in 2030

[11:36] Ah, right, okay. And then they anticipate revenues of $840 billion in revenue between now and the end of 2030. Between now and the end of the decade, though, let's go ahead and look that out.

[11:50] It's $280 billion. So between now to 2030, they expect to be negative $278 billion, but they expect to have revenue, right, of $840 billion, which you should note there is a missing number here.

[12:14] if they expect to have a bottom line of negative $278 billion opening eyes between now and 2030, and they're expecting revenue of $840 billion, which is crazy that you could even remotely project out that far.

[12:28] It's craziness. It's going to lead to a lot of losses. Maybe Karp was right. That means you are actually, 840 plus 278, you are actually looking at spending $1.118 trillion.

[12:47] That's what you're expecting to spend. That's how you make this math work. The Financial Times didn't tell you that. But that's how you complete the sentence, basically. Now, where does that money go?

[13:00] Well, the vast majority of that money, probably somewhere around $400 billion of it, probably goes to NVIDIA. NVIDIA is just hands down the king slayer of what's happening right now

[13:13] in the entire AI ecosystem. And I'm going to show you something that's really remarkable about this NVIDIA situation. It's called model compression. Now, this is going to be a little weird,

[13:26] and I'll probably go through this in a little bit more of a detailed video this weekend because it does take some technical explaining, and I might not be doing the best job tonight, so we're only going to touch on it today.

[13:40] Here it is on my iPad. I got it up for us. Love getting it up for you. All right. So here's the script. Oh, look at that. As I say that, everything goes black. All right. So what's this?

[13:52] At the top right here, you have a sheet that shows you a breakdown of the GPU platform on the left side. The GPU platform is important. The GPU platform is the history of what we've seen here.

[14:04] A100 to the RTX Blackwell 5090 consumer grade chip. You've got the professional grade chip, which is the RTX 6000. These are great chips. The H100, which Elon has a whole lot of over at his Colossus facilities.

[14:18] And what we've done here is we've simplified this a little bit to come up with an effective output speed just to try to compare these chips. And we're going to compare them at about 200 tokens per second.

[14:34] And the reason we're comparing them at about 200 tokens per second is what you're going to find is there's actually a high percentage at 200 tokens per second of what's called stranded memory. You see these chips up here have 72, 91% stranded memory.

[14:50] When you do more tokens per second, your ability to use the memory diminishes. The reason for this is the more tokens per second you can process in a bandwidth point of view,

[15:04] frankly, the less memory you need. So you end up having a bunch of memory sitting around not doing anything. Some chips are better than others at this.

[15:17] And this is where, to oversimplify this, we get to what we typically call a max tokens per second for a chip that is practical. so a practical tokens per second for a chip model

[15:30] would be something like 36 tokens per second on a RTX 5090 or a Blastwell 6000. For a Vera Rubin, which Vera is really the... I mean, this is all a reference to an astronomer.

[15:43] Vera Rubin was an astronomer. Vera is the CPU. Rubin is the GPU. So we'll focus on Rubin here. Rubin does about 10x the tokens that an RTX 6000 does per second.

[15:58] Which actually means even though you have so much more RAM, your ability to process tokens is way more than the delta between the gigabytes of RAM here and the gigabytes of RAM here.

[16:13] I mean, think about it. The Rubin has 288 gigabytes. Just to make math simple. 288 divided by 96. that's exactly 3x. So this right here, I'll click right here,

[16:25] this has 3x the token of the RTX 6000. It's got, or install DRAM. It's got close to 3x, a little bit more like 3.3x, the H100.

[16:39] But its performance on a practical basis, because its bandwidth is so much higher, is actually 10x compared to the A100. On a tokens per second basis, it's somewhere around 8x compared to the H100.

[16:55] So what that means is, these chips are actually becoming less reliant on RAM. Now, I'm not saying that high bandwidth memory is bad.

[17:07] I'm just saying, as the bandwidth increases for these GPUs, and they become more efficient, they technically need less headroom. They need less RAM because they can process so many tasks so quickly.

[17:22] Just for quick math, because it's somewhat useful to think of, I think this makes it a little bit more relatable. I want you to think for a moment what one million output tokens gets you.

[17:34] One million output tokens, just for giggles, on a Fable model or an Astra model is going to cost you about $50. That's pretty expensive. If you use like a QWEN model or a Chinese OpenWay model, you're going to pay somewhere around 40 cents for those same 1 million output tokens.

[17:56] 1 million output tokens represents about 750,000 words. So that's sort of a relevant comparison. When you say 50 bucks, you're paying 50 bucks for 750,000 words.

[18:10] You're paying 40 cents for 750,000 words when the model's compressed. Now, why am I bringing up model compression in terms of the price compression? Because what those smaller models do that aren't sitting at the frontier

[18:23] is they actually enable price increases on the older hardware. This means the data centers don't necessarily have to be bag holders.

[18:36] Because traditionally, and this is a mind gap, okay? This is so much information. I know it's a mind gap. But this is the mind gap. Traditionally, we think, old chip, new chip comes, you throw away the old chip.

[18:49] Why are you going to use the old one? But what's actually happening right now is the new chips just help you with the frontier models. Then, the distilled Chinese models, they just spit on the old chips, but they didn't exist before.

[19:02] So all of a sudden, the frontier goes from the old chip to the new chip. But then a new product comes around, and the distilled frontier goes to work on the old chips. and so in the most perverted way

[19:14] from a technological like industrial revolution point of view we are literally seeing old chips go up in value and new chips go up in value that actually makes

[19:28] and this is a slip as soon as I started realizing this I'm like because I mean we have we have our own chips right we have our own RTX 6000 5090s I mean we've got like I don't know

[19:40] 50 machines running or more than that, processing the data that we do for our reinvest members, you know, our alpha members and the software that we're building and everything we're doing. We have a lot of hardware and it's expensive. I mean, we don't have hardware like, you know,

[19:55] or GPT, obviously, but we're a small little startup. And I've realized those chicks have gone up twice in value. And I'm looking at it like, how are they more expensive? Are we getting twice the value out of them? And then I talked to our dads and we're like, we're not getting twice

[20:08] the value of them. We're getting more than twice the value out of them. I'm like, how can this be? So then I take what we're seeing in practice, and then I apply it to the studying that I'm doing and the research that I'm doing. This is actually, these are all little sheets

[20:21] that I added onto a semi-analysis article here. I usually have issues with some of the things that I see with semi-analysis, so I go through there and kind of try to, you know, see like, what do I agree with, what do I disagree with? And the problem with it is,

[20:34] So it's a lot. It's a lot of information to process. I try to put bottom lines together for people. And what I've observed is that, yes, you actually have a case where these data centers, especially the ones that are least likely to face bankruptcy, so the ones that are not dying and drowning in debt, in other words, the ones that have other businesses that can sustain them, they are probably going to survive.

[21:01] But not just for Vaughn, they probably will get their cap-ex paid for. Even if Elon's wrong, which I think he is, I don't think he's going to get $50 billion per gigawatt. I think he might get $18 billion per gigawatt with long-term contracts.

[21:17] But even if he gets that, so what? The chips have like a two, two and a half year payback. Big deal. The big boys who can sustain it with other revenue, though, or trust and financing, Elon via the stock market and his Morgan Stanley homies,

[21:32] as well as Starlink income and Starship, you know, income, and then obviously the optionality that, you know, you can kind of merge Tesla into this as well. Optionality there to stay alive.

[21:45] Who else? Well, Meta is just another example, right? I want to be clear, like, I have exposure in some of the names we're talking about. I'm not trying to, like, shill or pump anything this Friday night. Okay, nothing that I say here I think changes anything in the market.

[21:57] I'm just trying to share information. Meta, I do think, is another interesting one because they're ad business paid for all their chips. So even if you don't get the ROI, the chips are paid for. But you get the ROI as a bonus.

[22:09] So that makes meta also a very interesting option. I get nervous. And we've talked about this. Like, watch the Anthropic video where I go through the M-scale numbers, where I go through CoreWeave and the Nebius numbers.

[22:23] Nebius is the best out of the bunch. Watch that video, the Anthropic IPO Delay and the Nebious Review. That's a very useful video. We'll give you some context on that without rehashing it all here. But that's important. Model compression is keeping this bubble going.

[22:37] So, yes, these big dogs, like opening eye, are blowing money and they're losing money. That doesn't necessarily mean, though, that these chip holders are going to go BK. The chip holders will probably make their money back at least the ones who currently hold chips they probably make their money back before these guys go bankrupt I don like necessarily I like I have my doubts about the sustainability of the open AI and anthropic business models especially in the face of open rate models and we

[23:06] see what people end up doing with software. The next year will be transformative. Like, for example, what's something else that's transformative? Honestly, four hours of this robot doing stuff?

[23:18] It's pretty cool. Now, what Figure did, and I invest in a robotic startup called Aptronic via our venture capital group. And I think we're nicely up on that. We made two venture capital investments.

[23:30] One was SpaceX at like $350 billion. One was Anthropic at like $1.8 billion. I actually, before I had the VC, also invested into Anthropic. Not Anthropic. Also invested into Aptronic.

[23:44] Aptronic. That's different. I probably screwed that up a few times, so let me make this clear. Aptronic. He's a robot developer. Humiliated robots. But anyway, this is Figure. Figure AI, the Humiliated Robotics Company.

[23:57] What they're apparently doing is they're renting 30 homes in the Bay Area, and they are putting these robots into novel situations where the robots go around and do stuff.

[24:10] I think some of this is a little, you know, weirdly arranged. Like, I'm not sure who puts, like, a nightstand on the left of the room like that, and turns the drawers that way where the drawers are going to hit that extendo outlet they've got there.

[24:23] But whatever. I'm not going to look too detailed into this. What I think is fascinating is that apparently they are training 90,000 humans wearing like little head goggles or whatever and they're collecting data from humans

[24:35] and they make a very fair point that one thing that's unique about humanoid robotics is humanoid robotics act like a human. So they work on human data. How useful is that for training?

[24:47] very freaking useful. So very useful. Okay, so good for them. So how do you invest in the robot supply chain? So what I did for course members,

[24:59] and I'm going to show you some of this, not all of it. What I did for course members is I put under the research tab a lot of information on stocks that if they all had a two peg,

[25:11] two price earnings ratio, what stocks would have the greatest upside. You can see it under suits in the app and in the research tab. And we have a sheet that includes, let's see here, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11,

[25:26] probably about 20 names. 20 names that have upside in robotics, all assumed at a two peg. And there are downside forecasts of negative 63% for some of these robotic names,

[25:39] and some of them are up 265% at the top. And then it sort of ranges through. now I'll give you a couple examples one with certainty we know that's going to take over

[25:51] the robotic stack I have no question over this because they're already in and all over the robotic stack and people don't believe people don't even think about this or in my opinion see it and it sounds basic

[26:03] but it's NVIDIA because NVIDIA literally is the brain and the system that most of these robots operate on And that's kind of crazy to think about.

[26:15] Now, of course, there are other companies we have on this list. Paradigm, Applied Materials, Lamb Research, Analog Devices. Great. Lattice Semiconductors on here. These are all examples of companies that have various exposure to the robotics stack

[26:29] and that I'm very excited about. I think there's a big future for these. This is cool. Some of these have a lot of upside. But let's understand for a moment what NVIDIA actually has cooking. NVIDIA wins because these robots have to be trained.

[26:44] Where do they get trained? On RubinChicks. We literally just saw that same company, M-Scale, that I made a separate video on, sign a $3.5 billion compute deal

[26:57] to end up installing 100,000 Vero Rubin GPUs. I think they're going to need some more money, okay? This is an insane ship. And NVIDIA is the beneficiary of that

[27:09] because N-Scale wants to license those chips to companies like Figure so they can train or rent to those companies like Figure so they can train their robots.

[27:21] So NVIDIA wins on training. But then you have to move to on-device. But NVIDIA has that too because now people have trained on CUDA and now they're operating their robots on the ROS2 software.

[27:34] They're training them in synthetic worlds on Omniverse and then the chips that these robots actually operate on are the Thor and Orin chips, which are the computers in the brain for robots.

[27:47] So you have the edge case chip. So they literally have the CPU stack, the GPU stack for training. They have ConnectX and Bluefield DPUs for comms on robot to off robot. They have the software for robots.

[27:59] They have Omniverse for synthetic training for robots. They have the CUDA that you're training on and your edge running on, Orin and Thor. In close, Tesla is a diabolical vertical integration competitor.

[28:13] In fairness, they killed Dojo, so they're expecting to make AI-5 in 2027, some Taiwan semiconductors, Samsung reportedly AI-6 will be the partner there. But they want to build their own simulators, their own edge case training, their own powerhouses.

[28:27] This is why they have a research lab out of Austin, Texas, their own circuit boards. They want to do it all. Right now, so crunching on NVIDIA for training. So there you go. You get that circle back to NVIDIA. This is why it's crazy to me that NVIDIA is the cheapest.

[28:41] But it really lifts everybody. AMD, Broadcom, Marvell, ARM, Qualcomm, MediaTek, they all win from robotics workloads. All the different degrees and different valuations. Teradyne owns Universal Robots, for example.

[28:55] This is incredible. So I'm very excited about robotics and this whole stack. I, up until I saw this figure video, I'm like, man, I kind of feel like robots are like a 2035 play.

[29:10] So in other words, it's like nine years out. Maybe even nine to 15 years out. But, I have to be frank. Even though these robots that we're seeing on this still suck at holding clothing,

[29:23] this is a step change in what we've seen in the past. I mean, just the fact, I mean, nobody's going to pay 30 grand to go pick up five toys in five minutes, right? but this is a massive

[29:36] massive improvement assuming it's not tele-operated, assuming we're legit like autonomously telling these puppies to go clean this stuff up it's still like, when it tries to fold clothing

[29:48] I mean look at this, it takes forever to pick up one, it's like, come on, like pick it up pick it up man, it's not that hard come on buddy but like it knows to go back and keep trying and so

[30:00] frankly, even if it's ten times slower than a human, is doing it 24-7. Like, that is already quite impressive. And it's not going to make the bed perfectly. Like, is it going to stretch it like you're in a, you know, in a hotel?

[30:14] No. You know, it's going to take it from looking like this to looking like, you know, so like crap. But this is 2026.

[30:26] And what this does, when people, oh, sorry, I got a little better. What this does when people see this is they throw more money at robotics. And that's how you end up financing the very NVIDIA chips that are making this stuff possible.

[30:40] It's not bad. It's very impressive. So I am excited about this. Figure is actually making a lot more progress than I thought possible.

[30:53] So this makes me more optimistic. Do we have problems potentially in the economy? Of course. Remember earlier we made a video, and this is accurate. Of course, you're always going to have the losers who just read the title of the week.

[31:05] I thought you're bullish. Why are you saying they're too past to a recession? Because there are. You can be bullish. I'm bullish right now. I think you should be buying right now.

[31:17] But at the same time, you should realize that there are risks. The big risk is the AI spend collapses somehow, that people are selling bear pelts before they have the damn bear caught. And if they go to the forest and there are no bears, there's no financing, then they can't sell the pelt.

[31:32] They can't deliver their product, they can't train their robots, the video doesn't get paged, the compute runs out, opening on an anthropic, don't get the compute they need, and you have problems. So, the circular financing needs to keep happening.

[31:45] So far, there's no sign it's slowing down, but if N-scale's IPO tanks, anthropic's IPO tanks, well, those are canaries in the coal mine. There's a second risk, and that's not bad right now, it's not bad.

[31:57] there's a second risk. I mean, the towels aren't perfect, but they're, like, good enough. All righty, good enough. It's pretty good, honestly. They're pretty good. I mean, they wouldn't get a job at Hollister,

[32:09] you know, because you had to get those, like, tight creases. You know, that's kind of crap. But it's a kitchen towel. Who cares? You know, you're not going to fold jeans like that or shirts because you're wrinkled up.

[32:22] I love that they put four hours out of this one. and this is so ballsy. Good for them. Honestly, brilliant marketing. I'm going to just let this run in the background. It's hilarious.

[32:35] Anyway, so the second fear, which I mean, just watch the recession video, but the second recession fear, and you'll see the data points to watch, is the labor market. The labor market rolls over, it's all over. So the point is, I'm bullish with two red flags

[32:50] that I watch for. I think I'm crystal clear about that. I think I'm very optimistic about what's happening. Model compression is a win for the hardware stack. It's less of a win for memory stocks.

[33:04] I'm not trying to dog on memory stocks, but it just is less of a win for memory stocks. Yes, additional headroom does enable you to potentially do more parallel compute where you're taking multiple batches of tasks,

[33:16] but I think the high bandwidth on the actual GPUs reduces your need for, ironically, quote-unquote, high bandwidth memory. so overall bullish hardware I still maintain software

[33:29] Q3, Q4, we nailed that and I think that's going to keep going and so that will continue to play out I should say obviously people are nervous it's pre-election is Trump going to talk, oh whatever

[33:43] whatever whatever, all that crap we'll figure it out now, let's talk a little bit about Warren Buffett so Warren Buffett sat down.

[33:57] It's kind of sad. Now he's chairman, emiratus, or whatever they call it, and his 70, like 70, 71-year-old son took over. One of the last things that Warren Buffett supposedly was still around for with Greg Abel

[34:12] was Greg Abel's purchase of an all-cash deal of Taylor Morrison. Taylor Morrison is a home builder. And Taylor Morrison provides a very interesting sort of segue into the real estate world Now look if you don want to hear about real estate just remember price goes up 900 bucks tonight on the Alpha membership We making a huge change

[34:32] You're going to get well taken care of if you're a course member before the end of the night tonight. So if you see it available, buy it. If it's gone, it's gone. Now, as far as real estate,

[34:45] real estate is very interesting because real estate wins in two ways. it doesn't feel like it wins right now. But if we end up having massive, massive artificial intelligence deflation,

[35:02] so we're going to go in and say AI-led deflation. This is going to be after the build-out. So that's post-build-out phase, when most of the build-out is done. Once you have AI deflation, rates come down.

[35:15] Kevin Warsh is in a money printer. when you don't print money rates going to zero is not enough you actually end up getting negative rates at base so the Fed goes to zero

[35:28] you probably have to try negative rates from base and the 30 year mortgage ends up sub 2% you're going to have 1.8% 30 year mortgages just to incentivize spend when you get AI deflation

[35:41] them not printing money means you have even less inflation which means rates go lower than ever before. I don't know when that happens, but I know, I strongly believe, real estate is going to dominate in this regime.

[35:56] If you have an AI crash, well then, you're going to have rates also go to zero, and you're going to have a massive slowdown in construction, which means you also probably get under 2% rates on the 30-year mortgage.

[36:09] Real estate wins again. where real estate has a harder time is when you have war with Iran that goes on forever and oil prices that are high forever as well as an AI compute build up

[36:24] but that doesn't last forever the bubble won't last forever real estate will have its turn it's just not now and so that's why I say if you own real estate now I think there's an opportunity to slowly just pay down a little bit more and as the bubble running on or whatever

[36:37] goes on for another 5 or 10 years You just have that much more time to build up home equity that you could break the piggy bank on in the future. And that's kind of what we talk about as well when we talk about my company reinvest.

[36:51] We'll start up. People ask me, they're like, Kevin, you know, what about IPO? When IPO? So when we started the company in 2022, we guided, hey, maybe, and everything I'm about to say is not a forecast.

[37:03] You know, we're not raising money right now. Hashtag don't sue me, bro. We could be wrong. could make mistakes with what I'm going to say. This is just ballpark, open, transparency conversation.

[37:15] We started the company in 2022, and I guided, maybe we can IPO this puppy in 2028 to 2030, a phrase I have reiterated. There were actually times that we thought about potentially being able to IPO earlier, which is great,

[37:28] but I didn't think that was the best thing to do at the time for the company. And so I look at 12 months ago, what we had, mind you, this sort of Warren Buffett concept That was sort of a segue. I'm bullish Berkshire Hathaway, bullish Warren Buffett.

[37:42] I think he sees the same thing with real estate. That's why he's getting exposed to home builders. This morning we were looking at a timber company and we're like, massive shortage of homes being built. The shortage for homes being built is literally just getting worse

[37:56] because rates are so high. So you are just like amplifying this pressure bucket in real estate. It's just going to explode. because it's going to drive prices up even more when rates come down.

[38:10] That makes unaffordability even worse. The wealth gap, which I've talked about for years, will just get worse. AI makes companies richer, and because of the AI structural deflation, whether it's boom and bust or just deflation,

[38:23] you end up making housing more expensive. So in 10 years, you could be in a place where AI has created so much disinflation, you have no pricing power for your job, or very little, unless you're in a certain type of field.

[38:36] and or homes, or both frankly, homes are so expensive that nobody can afford them. You have to usher in socialism at that point and you've got a whole bunch of other problems. But this K-shaped economy just ends up being worse.

[38:49] I believe that. And so that's why I've been sort of pounding on the table for the past few years. I'm like, folks, buy assets, buy companies. Companies are going to win from AI. Housing will win from the result of AI.

[39:02] AI deflation or a crash. Because this is in the real estate bubble. There's not enough debt. Who's buying houses right now? Who's doing all in on leverage on homes right now?

[39:14] Not a lot of people. The commercial real estate stress is already almost behind us. And when you're saying mall rates are going up, malls. Anyway, so going back to this talk. So a year ago, we basically had no software. I was looking at a flight plan, actually.

[39:27] I put it on our daily wealth yesterday. I was looking at a flight plan, and it was actually my flight back, but I was meeting our dev team, and so I had this screenshot of the flight plan because I flew us there.

[39:40] Now, you already know I don't fly anymore. I don't think anybody should hobby fly. I think that if you are going to be a pilot, you should fly three times a week, and maybe that's not a prescription for you.

[39:52] Maybe that's more of a prescription for me. I think skills decay unless you're flying three times a week, and I thought it was an excellent pilot when I was flying I mean I was flying like five times a week it was crazy how much I was flying

[40:04] but then I didn't fly as much and it didn't make sense to be a pilot I mean I'm still a pilot once you're a pilot you're always a pilot but I thought it was very interesting I hope it actually shows up here

[40:16] I'll do my HDMI plug there there we go yeah this was oh it only shows part of it this was my flight plan or whatever oh there you go this was our flight plan and I thought it was really cool kind of a little memory looking at it, but this is my flight plan back home.

[40:31] You can see the tail number there. And, Hicks Manor, Papa, Papa, Papa, clear the way, man. But, you know, that was July of last year was the date on that.

[40:44] And I bring it up for a reason. That was July 5th of last year. We didn't have the software we have now. There was no Holmes AI product. That, in visual data only, came out at the end of December 2025.

[40:59] We had nothing. I mean, we had our wedge software, which we, you know, bought for a few hundred thousand dollars, which is a bargain in hindsight. We've made way more than that back.

[41:11] We had nothing. And so we've since built that, and we released our valuation data for it on June 30th, which has been amazing, and it's just getting better. I mean, we just released another update on the back end,

[41:24] and you should see four times as many properties on the Homes AI app right now. It's incredible. In the re-invest app. It's remarkable. And then in July, we released our stock software, which is like a V0 of our terminal software that's coming out in the future.

[41:40] And we're going to do some really cool stuff for existing course members probably next week. This is why I say, like, before that $700 or $900 bump comes in, you probably want to get in like ASAP. I'm just full transparency.

[41:53] I know it's like last minute on Friday night, but oh well. Now, I just think about what we've released in the last year, and I can only imagine what we're going to be in a year from now. But the point is, people ask me all the time,

[42:06] Kevin, when IPO? And the answer is, I don't know. I think I will have a lot more clarity by January of 2028, and then maybe we can still be in that target of 2028 to 2030. But it all comes down to how we build annual recurring revenue,

[42:21] how our software grows. We think between now and then we'll probably do another fundraise. We might do a fundraise at the end of this year. We might do a fundraise, I don't know, next week. That's probably not next week. It takes a little longer to put together than that.

[42:34] But we're not fundraising right now. And what we expect is with the vast majority of that, I would guess 85% to 95% of it, we'll just throw it into building real estate,

[42:46] building in supply-constrained areas where it's hard to build, and we'll buy reg deals. mostly because our software side and sort of like securities allocation, that's self-sustaining,

[42:58] which is fantastic off of our memberships or rental income or whatever. It's self-sustaining. The company's operating and operating profits. From a cash flow point of view, that's great.

[43:10] So that's offering cash flow positive. You have to be careful with how you order this stuff because then you take out like convertible bond interest, which we expect those will probably convert next year

[43:24] and then we don't have that interesting one. Here's like depreciation on real estate which the homes have just depreciated in value but we're still depreciating. I'm pretty sure it's over like a million bucks a year. It might be more than that. I mean, whatever it is.

[43:36] Let me take a quick guess. I should honestly just pull it up. That's probably better. It might be closer to $2 million a year. Probably closer to $2 million a year which is wild. But it's great because it means when we collect revenue from software

[43:49] we're actually sheltering it with real estate depreciation meanwhile properties are growing up with value stupid but we think we'll use like 85-95% of our next fundraise to just buy more real estate because it builds

[44:01] really strong foundations in our opinion like paying it off cash paid real estate fully rented out that is a strong foundation with in-house property management so we're really excited about that and we've got

[44:13] over 10,000 people enjoying the Meet Kevin app on a monthly basis and it just keeps growing and keeps hitting new highs and new highs but it's so early and so we're working on tools to make things even easier and better for folks

[44:25] and we think the safety net of all of it is real estate in this environment. Can we be bullish AI? Can we be bullish chips? Can we be bullish software? Yes. But we're like hedge bullish and I think this is the most

[44:38] transparent way to look at all of it. So if you found this helpful consider liking the video consider subscribing go to meetkevin.com meetreinvest.com it's the same website and get into the alphabet membership before it's over. Remember what you get. You get the alpha report every morning.

[44:55] We went through our Monday sheet. We went through our Tuesday sheet. We went through our, what do we got here? We went through every single day. Ted Day we went through. First day Friday. We went through all of these. And then of course remember what you get.

[45:10] The alpha report every day the market is open. Find seller alert. Of course remember live stream short medium and long term analysis all the courses a lot of cool stuff and there's

[45:22] seriously a massive price increase going so as i said i encourage you to take advantage uh and get in before that big uh that big shift flips over to the real enjoy what's coming

[45:35] so excited to buy it why not advertise these things that you told us here i feel like nobody knows about this. We'll try to advertise and see how it goes. Congratulations, man. You've done so much. People love you. People look up to you.

[45:47] Kevin's half left there by night's limit. And you see what's neat, Kevin. Always great to get your take.