GPT-6 Astra 已经来了。你拿错 AI 股票了吗？

On Friday, OpenAI dropped its latest model, GPT-6 Astra. Once again, the benchmarks are insane. Astra handles computer tasks better than Sol in about half the time. It scored 99.9% on OpenAI’s ARC-AGI benchmark, and it’s the first OpenAI model to reach their critical threshold for cybersecurity capability.

So naturally, AI stocks should all go to the moon, right? Right. Because for the last few years, the AI trade has been pretty simple: better AI needs more chips, more memory, more data centers, and more power. Except Astra isn’t just better; it’s more efficient. Depending on what happens next, that could be incredibly bullish for some of the stocks that you own and really bad for others. Today on Dumb Money, GPT-6 Astra is here. Are you holding the wrong AI stocks? This is Dumb Money Live. Hey there, Dave here along with Chris and Jordan. We are Dumb Money. Welcome to Dumb Money Live. We all know that Astra is 100% capable of just watching YouTube for you, but you probably forgot to prompt it to smash the like button, so go ahead and do that for us this morning. Chris, Jordan, we are back on track with back-to-back episodes because we need to regroup on what might just be the biggest mistake investors could make on this new GPT-6 platform: being right about the technology and being wrong about the stocks. Chris, I know that you posted on X yesterday that GPT-6 Astra is “simultaneously far more capable and efficient than expected, yet still far from good enough.” You said that it’s hard to imagine a more bullish outcome for AI infrastructure. Explain yourself.

Yeah, I think the biggest breakthrough here with GPT-6 Astra isn’t that it can do more; it’s that it can do more with less. So that’s not actually a bearish thesis for infrastructure. That’s what we’ve been talking about for years now. Jevin’s paradox is actually arriving in real time.

The issue, though, guys, is that Wall Street might make the same exact mistake they’ve made with every one of these cycles: thinking that falling cost per task actually results in less AI spend. Again, there are 2 sides to this. What side do you fall on? You know what side I fall on. It’s the side where AI gets 10x cheaper. I think we’re going to use 100x more AI.

And we’ve already seen that. Once Astra examples started getting out there in the marketplace, we saw that people are now just leaving. The tokens are expensive, but they’re using less of them, and they’re just leaving their computers on and having Astra build a game for them over 5 days. So it’s a different world where, yes, it is more efficient, but it just means we want to leave it running longer and have it doing more for us.

Yeah, Jordan, I want to get your take on this as an engineer. I’ve been reading every qualified engineer’s take I could find over the last few days on GPT-6 Astra. My one big takeaway is that the big differentiation with GPT-6 is that it essentially takes you from A to Z. It will really take over the process of determining what agents you need, what those agents need to be doing at any point in time, and making a lot of incremental decisions for you to get all the way down the path to Z, which is why we’re seeing so many really cool zero-shot examples where you just want something and it does it.

There’s a pro and a con to that. The pro is that it will give you a full end result for what you need and make the decisions along that path that it determines to be best, as opposed to waiting and stopping, which is a lot of what we see with Anthropic’s latest AI model. The negative is that it’s not as great when it comes to doing things in the perfect, correct way. It’s just going to make a lot of guesses along the way.

I think a lot of people are having issues with Astra in terms of controlling it in a defined manner and working with it. You have to break it out into little pieces. Otherwise, it does too much. So I think there’s a learning curve here.

As an infrastructure trade, what’s most exciting about this, guys, is that it just goes nuts with agentic AI. The compute is meaningfully down. It can do roughly 3x more with the same compute. But the amount of compute it’s willing to take on to do stuff—it will just run with it. So it feels like one of these scenarios where we’re entering a new age of AI that is just going to consume a tremendous amount of compute, more than we even thought was imaginable, and open up new use cases.

But it’s not there yet. Guys, we’re not there yet. There’s so much refinement needed on these models that the infrastructure trade is in the perfect place right now. We’re so intrigued by Astra and what it can do, yet everything that Astra is showing us it can do is opening up more questions. We have an issue here. We have an issue there. It needs to get better here.

It could do so much more, but now we want it to do so much more than that because we see what it’s capable of. It’s not quite capable of giving us a finished product that’s actually usable in any meaningful way. We’re maintaining a product, keeping up with the product.

I know I’m talking all over the place, but I just think this is the perfect scenario for the infrastructure trade. We’ve never been in a better place for the infrastructure trade than we are right now, post-Astra. Jordan, do you get what I’m saying here? I just didn’t know if you had a question.

So I have thoughts on this. Did you listen to the DHH interview with Lex Fridman? Because he basically addressed all of this already.

He’s a big user of AI. He’s come full circle from “It’s not useful” to “It’s supremely useful now.” So he’s gone through the whole spectrum of feelings, emotions, and usefulness of AI.

For engineers, it’s tough, right? With the early models, you would just tell them what you wanted very specifically. You’d lay out exactly what you wanted: “I want functions to do this,” or “I want a class to do that.” You could wrestle the LLM to do those things. You’re right: with some of the newer models, with Astra, you need to let go and just let it build, because it will do as good a job or a better job than you would have done yourself.

It’s all about letting go, trying to define less, getting less in the weeds, and letting the AI run with the problem. That’s what these models are designed to do. I think the best word for it is that the models are tenacious. They’ll attack problems to completion in a way that we haven’t seen them do before.

Jordan, the counter to that is that I’ve read that some of the ways it approaches architecting code are whatever the fastest way is to get from A to Z. It’s not always the most durable, bug-proof, sustainable codebase. Are you hearing the same thing? It’s taking a lot of shortcuts, which could technically work to give you that end product, but it’s just not the best way to go about it if you want durable, long-term code that’s not going to break, right?

Yeah, I think it depends on what kind of codebase it is, what concerns you have about the code it’s producing, whether you can define those concerns, and whether you can then figure out what you need to define to get it to produce what you want.

Yeah, right. So what you’re saying, Jordan, is that we’re only a few days in. In the raw, it has weaknesses. I think people need to change how they’re using AI to develop software. The techniques that you used for previous models aren’t going to continue to cut it. You need to learn to work within the models you have.

On the AI development side, every new limitation we’re finding is because it’s doing more. But as it does more, we’re finding more quirks and issues. Every limitation is theoretically another $1 trillion incentive to keep scaling the models and improving them with more compute, more memory, and more energy.

The deal with these models is that they’re built, as far as engineering is concerned, on all the engineering work that people have contributed to them and to open source. That’s why they’re getting so much better: they use all of this stuff. But it still takes an engineer to go through and make sure that things are being built in the way they should be built.

You can vibe-code stuff, but I think you’re going to end up with garbage. I still think you need engineers and real technical eyeballs on projects you’re trying to build. Every iteration of new models coming out gives engineers something to relearn: what is the best way to use this tool?

Does anyone remember hoarding the Mac minis because of OpenClaw? That seems like a distant memory now. But this new version is faster and more efficient, and we’re at GPT-6. Imagine GPT-12 or GPT-50 or GPT-200.

At some point, it’s going to run on some pure, pre-silicon raw sand powered by the sun, and oceanfront property is going to be even more valuable. But until that time, while we’re in a post-silicon, post-electricity, post-AGI organic form, it’s going to be confined to some world where we’re still talking about chips, memory, data centers, and power.

My thesis hasn’t changed on what is going to power all of these versions of AI until we have some major shift to a technology that doesn’t involve chips, memory, data centers, and power. Guys, is it insane that it was only a few weeks ago when the entire AI trade seemed to be unraveling because of open-source models coming out of China? Could it have flipped more 180 degrees? This is actually wild, the way that we go back and forth.

Let’s just take a moment here to realize that this is going to repeat itself. If you’re an investor, you have to be mentally prepared to understand that at some point in the next few weeks or the next few months, there’s going to be a new open model that’s going to scare the crap out of everyone and make you think that OpenAI and Anthropic don’t have a shot, right before OpenAI and Anthropic come right back at you with an ASI-looking model that makes you think, “No, these guys have such a unique advantage that the value of them is in the trillions of dollars,” which is where we are this morning with the new filing for Anthropic at $2 trillion. And people are saying $2.5 trillion is where you’re marked for an IPO right now.

I mean, guys, this is an insane time to be alive as an investor. Do you know that in San Francisco, the housing is going out. I keep hearing from people about the housing prices up there because this Anthropic IPO—I don’t know if this is true—is supposedly bigger, meaningfully bigger, like 3x bigger, than every IPO in San Francisco that came before it in terms of the look-on. It's driving to investors. We’re going to have so many wealthy people. It’s going to be disgusting.

This is absolutely disgusting. This is like a cultural moment. And I’m not saying it’s great. I’m not saying it’s good. Okay, we’re about to have a lot of wealthy people minted here between Anthropic and OpenAI. Dude, I don’t know what this does for society. What do you think about that?

It’s fucking kind of nuts. What is the average net worth—or the minimum net worth—of an NVIDIA employee these days?

Well, let me just tell you this, guys: The price of the nicest resort hotel room ever feels like it went up from $1,000 to $2,800 or $3,000 in the last few years. I feel like if you want to go to the nicest places in the world, just expect to pay $5,000 a night, $6,000 a night. That’s what it’s about to be because you have the 0.001% with infinite money. At this point, these people have infinite money.

This is not about being worth this or this or this. There is a relatively large number of people in the next year that are about to come into infinite money. Meaning, they want to go to the best places, and they will. Whether they spend $5,000 a night or $7,000 a night is kind of irrelevant because they’re all worth hundreds of millions. And they don’t—they’ve never had money before. So, guys, this is insane. It’s absolutely insane.

By the way, I’d like to see, for your example, how resorts handle that, because there will be this new infinite-money class and then there’ll be everyone else. Is it just that these exclusive places are a—you can’t get in unless you have unlimited money—or how’s it going to work?

I want to be in the private club where you have unlimited money. I don’t know, but let’s talk about trades because there actually are good ones. I think, listen, right now, this could change, and I think if you’re investing right now, you’ve got to have eyes on the market every few days to see what’s popping up. But right now, we are definitely in this cycle where OpenAI and Anthropic are real. They are going to IPO, more than likely, at like a $2 trillion-ish level, a range of value that supports basically the entire infrastructure trade.

Who are the big winners here? I don’t think you get too fancy. I don’t think you get speculative with your AI trades here. If it’s me—and it is me, and I’ve made some trades today—I’m going to zone in on the near-guaranteed winners from this cycle.

Okay, so who? Listen, I don’t want to talk about the same stocks all the time. I am going to talk about a new stock today. I’m going to talk about a new stock that we have not talked about in years, I promise. But before we talk about that stock, I did up my position again today in Amazon, and here’s why.

By the way, I read a report yesterday that Amazon was moving up some of its supply chain for AI infrastructure, increasing its order sizes and accelerating delivery on racks and everything else going into its data centers. So, again, I’m seeing almost perfect visibility into what’s happening at Amazon right now, and it’s awesome, dude. I mean, it’s absolutely awesome. How is Amazon—do you remember the deal they just worked out with OpenAI?

Okay, what is it? They upped it to like $100 billion, dude, of infrastructure commitments from OpenAI. And they have a couple of things with OpenAI that are exclusives in Amazon Bedrock now. I mean, what freaking timing?

Not to mention the fact that Amazon, if they finish out their investment, is going to have $50 billion invested in OpenAI, with the potential to do a 2x to 3x that on an IPO. That’s another $100 billion in cash, $150 billion in cash right there. Now that Anthropic is looking to be a viable IPO in the near-term future at a $2 trillion, maybe $2.5 trillion, price, what does Amazon have? Like 8% or 9% of Anthropic. You’re looking at a couple hundred billion dollars for Amazon from the Anthropic IPO.

We’ve been talking about that. Also, dude, I’ve got to own more Amazon, and I do. I went in deeper on Amazon today. I’m just going to put it out there: I went in deeper. I’m starting to get more levered in Amazon. I’m starting to get excited about it again.

So, on the negative side, we had a plane crash. Perceptionally, not great when it has Amazon Prime on the side. But does it matter for investors? Amazon actually doesn’t operate the flight. It’s a branding problem.

And the fact that it crashed into a bunch of Cybercabs—or didn’t actually hit them, I don’t think—but the way that camera shot is set up, and you see the Cybercabs in the foreground and the plane smoking in the background, the third-richest man’s stuff just crashed into the first-richest man’s toys. My favorite bit—and, by the way, people died here, so it’s an absolutely tragic, tragic accident—but when I saw those trucks that were right next to the plane, like 5 minutes after it crashed, and someone was joking that they’re getting packaged, I’m like, dude, you never know with Amazon.

That company is crazy, dude. Like I said, I’m getting my deliveries now in 20 minutes. It’s like magic. I cannot believe the stuff Amazon is pulling off. I read a lot about Amazon partnering with Qualcomm this morning on custom chips for their data centers. I mean, it’s something every day.

And that’s why I’m not playing Amazon for day-to-day news. I’m playing it for the long term. This is a stock that is going to be at the center of AI for the foreseeable future.

Yeah, but if you look at what’s happening right now, if we believe in the Anthropic IPO, which means you probably believe in the OpenAI IPO, which will come after it, I’ve been one of the very few people that have been kind of neutral on OpenAI for the last few years, going, “Hey, guys, I know everyone hates Sam Altman. I know this company has some potential management issues there.” At the same time, you just can’t deny the fact that they’ve basically been the number 1 or number 2 most-downloaded app for 2.5 years globally on the iOS store, essentially every day for the last couple of years.

I don’t know why people don’t understand that that is something that happens like once every 15 years to a company, and that the consumer adoption that continues to happen with ChatGPT is unprecedented and astonishing. The degree to which people tend to stay with a consumer app like that once they’re on it is way underappreciated. We have 1 billion people that are using ChatGPT aggressively throughout the day. Lots of them are now paying $20 a month. Some of us are paying $200 a month.

There have been times when people were like, “OpenAI is done. ChatGPT—” Not anymore. Every time another company comes out with some new fancy AI, I always try it. So, I was on Grok. I was on Claude. I’ve gone through all of them. And it’s kind of annoying because, as a consumer, I have to remember, “Okay, I was having the conversation about my dog’s medication. Which one was that? Was that Gemini?”

I have to go find the conversation because it knows stuff, and it has relevant information about me that I need it to remember. At some point, though, I realize I keep going back to OpenAI. I keep going back to whatever they have. And so it makes me want to give them more information about me so that I know it’s all in one place. Then I want to stop experimenting.

I think there’s information asymmetry and arbitrage in the OpenAI trade and everything related to it because investor sentiment has been really negative the last couple of years on Sam Altman and OpenAI. I think the investing world just has blinders on when it comes to OpenAI. The investing world wants to hate OpenAI, wants to hate—does hate—Sam Altman, maybe for very good reasons.

But it does not negate the fact that OpenAI is—with the exception of a period of time when Anthropic was on fire with commercial accounts—consistently in the leadership position. And now, with what they’re doing on the commercial side of the business, with their enterprise product, they are definitely in a place where they could theoretically go head-to-head with Anthropic. Now Gemini and Google are looking worse than they ever have.

So it’s now looking like essentially a 2-horse race for frontier models between OpenAI and Anthropic, a duopoly of multitrillion-dollar companies. Both of them have bad investor sentiment right now, which I think is putting blinders on the fact of how well these guys are doing. It’s actually incredible.

So all AI has kind of a negative feel, though, right now. You have Grok and Elon there. So, I mean, every AI has a negative sentiment, and then there’s the fear of losing jobs and having your job taken over by AI.

So it’s this kind of dance where, if you’re like me and you see the potential for this, you want to be involved as an investor and not worry about losing your job because you’re going to be in the investor class. But here’s what’s wild: if you really want to go all in on OpenAI, SoftBank is the way to do it. SoftBank owns something like 13% of OpenAI. Microsoft, I think, owns between 20% and 25%, but it’s a much larger company. SoftBank has a market cap of around $230 billion, which is roughly the value of its OpenAI stake.

SoftBank has debt, but they also own Arm. It’s a really interesting trade that I’m not in yet, but it’s one that I’m going to spend a lot of time evaluating this week. For the first time in a long time, I’m really going to look deeply at SoftBank. I already have massive exposure to OpenAI through Amazon, obviously. And I have massive exposure to Anthropic through SKM, South Korean Telecom or whatever, which I still think is an interesting opportunity going into the Anthropic IPO because I think around 20% of the company is in Anthropic shares.

I don’t know, man. SoftBank is really interesting. I think it’s underappreciated here because everyone just assumed OpenAI was in trouble, and it’s going to take some time for that to reverse. If you look at the last couple of days, SoftBank has been up quite a bit on this OAI, on this Astra move. But you could make a case that it should be up a lot more just based on OpenAI and the IPO that’s likely coming, as long as the market doesn’t get completely destroyed by some new open model or something happens in the next few months.

I do have a new stock trade on this, though, and it’s an interesting one.

This is the moment we’ve been waiting for: the new stock pick.

Yeah, this is a company that we invested in privately when it was a pretty small company, and then we all exited at the IPO. Can you take a guess what it was?

In gaming. In gaming.

Oh, yes, yes. What’s the name of that company?

Unity? Like you or something?

Unity, guys. Okay. What is the biggest use case that we’ve seen in the last few days from ChatGPT’s Astra? It’s been one-shot video games, which on the surface looks really bad for video game makers. But the key isn’t that AI makes video games. It’s that AI is now radically lowering the cost and the skill required to basically create this interactive type of environment.

Unity actually owns one of the biggest environments where the agents go to build all of this. If you look at how a lot of these games are being built right now with GPT-6, a lot of them are going through real engine operations or through Unity. Unity actually has an AI gateway that they built so that codecs can actually operate inside of the Unity editor. That’s what Astra is good at: getting in and using the software on your behalf. It’s not creating the game out of thin air. It’s actually using software to do it. It is its own computer.

I wasn’t thinking about the Unity play, but that totally makes sense. It does what people would do, and people would write software to interact with the gaming engine because that engine is a lot of the work. It’s all the heavy lifting.

Yeah, the Unity Editor is what’s actually reading the project. It’s modifying all the scenes, writing the code, and executing the actual actions with codex. More importantly, OpenAI used Unity as the case study for the Astra launch. They actually used it as the case study, and with their integration through the company PlayCo, they reported 50% fewer manual fixes versus GPT-5, again using Unity.

So I think one of my favorite GPT-6 trades right now has nothing to do with data centers. I think Astra turns millions of people who are gamers into game developers, and Unity becomes the picks-and-shovels layer for AI-generated worlds. It’s a really interesting trade.

I haven’t owned Unity in a long time. I bought some today. My position is moderate. The reason why is that, historically, people think this is a negative for gaming companies. They think people are just going to use the frontier models to make games end-to-end without touching any other software layer. That’s not at all what Astra does.

Maybe at some point in the far-off future, the frontier models will do everything end-to-end on their own. Maybe. But we’re not seeing that today. Right now, all I’m seeing is a model that’s going to meaningfully increase the use of Unity and open Unity up to so many more developers who simply weren’t in a position to code inside it prior to GPT-6. So I’m super excited about Unity, and it’s a new position of mine starting today.

It’ll be interesting to see how these software companies start to charge. There’s per-seat subscription pricing, so what if you have 100 bots that are each named something and have a different task in your Astra, or whatever GPT-12 ends up being? Do they each have to pay for a subscription? I saw people using Final Cut Pro. Do they take a cut of the actual game revenue? I’m not sure what they do.

Jordan, I almost think it’s irrelevant because that’s a business item that, if it’s one way today, they’ll make it a different way next month to adjust in this AI-driven world. They’ll figure out how to make money off of it.

Yeah, I just think Unity is really interesting here. I even looked at a company.

Yes, so we just said Unity takes a cut, and I think that makes sense. As far as incentives go, you’d want people developing on the platform. You’d want to hook them in, and then you just take your cut of the gross when it comes to the game revenue.

Completely. I could even see Unity eventually, if you think about these companies being valued in the trillions of dollars—

Okay, let’s just look. I haven’t taken a hard look at Unity in so long. What’s the current market cap on Unity today?

It’s—oh my gosh, guys—it’s $18 billion. You have Real engine and Unity. Real engine was already acquired, right? It’s gone.

So, $18 billion. Are you kidding me, guys? Is this not an acquisition target for OpenAI or Anthropic, or one of the big frontier models? They can literally buy Unity for $30 billion without even thinking about it—$35 billion.

But do they need to? Do they need to?

Yeah, I’m not saying that they do. Here’s the issue: do they need to? No. Would they potentially do it as a defensive move against their competitor, because there are really only 2 big gaming engines?

Oh, just so they could say, “The only way you can develop Unity games using AI is through ChatGPT now.”

Yeah, exactly. I think they might do it as a defensive move. I’m not saying it’s going to happen, or that it even needs to happen, but I like it, man. I like this concept of open worlds and figuring out how AI eventually bridges us to that place.

It seems really obvious that companies like Unreal Engine and Unity play a critical role in that. They become hyper-relevant now for the first time. Not that they weren’t relevant a year ago, but they’ve become critical.

Gaming seems like it’s ripe for development because it’s so boom and bust. The amount of money it takes to develop one of these super-high-end, highly marketed video games is enormous. The labor is through the roof. Could you de-rip that by using AI to develop these games?

I’ve thought about this a lot. Obviously, I’m invested long in Take-Two. The knee-jerk reaction has historically been that this is bad for gaming companies because anybody can create games. But you can’t create a universe and operate a universe in the same way. There’s a huge difference between building a universe and actually operating a universe or operating a game with millions of players, functionality, and interconnectivity. There’s just an immense amount of backend work that goes into that.

I think the beneficiaries here are gaming companies, where GPT-6 and whatever comes after it will meaningfully reduce the cost for those gaming companies to expand their universes and enrich them. When we think about GTA 6, which we’ve talked about a lot over the last few weeks, and how that universe expands over the next 10 years, they can leverage AI while maintaining the IP of the characters and maintaining the community.

Even if you had 1,000 different games that were kind of just like GTA 6, people still want to be in the game that others are in. They want that connectivity with each other. They want to buy into that communal IP and have something to talk about. It would be weird if everybody were playing a different game.

Maybe I’m wrong about that. Maybe we’ll go down a path where every human is just in their own little world that’s custom-designed for them. But even if we go down that path, I feel like that’s not happening in the next 2 or 3 years. That’s 5 to 10 years out. So, in the short to medium term, this benefits gaming companies because it allows them to do a hell of a lot more with less.

Right.

But it’s not the knee-jerk reaction, so you have to be careful. Tay 2 opened down today. Now it’s about flat. Beyond that, I would just not get too fancy. I would think of the obvious infrastructure players that are also invested in these companies: Amazon, probably Microsoft, and Oracle.

Okay. What’s Oracle doing today?

Because Oracle is so tied in, as we know, to OpenAI. Isn’t OpenAI like 50% or 60% of Oracle’s future build-out? If something bad happened to OpenAI, Oracle would be completely up shit’s creek.

Oracle’s up a couple of percent today. I mean, nothing. Today’s such a nasty market day. It’s a weird day for sure.

But Iran, right? Of course, this is going to happen, and then we’re going to get into another fight with Iran, which is what happened last night—yesterday. Oil prices are above $100 a barrel, right? So again, net short-term negative for a company like Amazon, which is also a bit of a headwind for Amazon today.

They had the plane crash, which overtook the news cycle over the last 24 hours. Now you have oil over $100 a barrel, historically bad for Amazon. Yet that’s distracting the market from the real story here: the massive flywheel that Amazon has just gotten even bigger. It’s disgusting how big it is now.

I am so pumped to have Amazon sitting where it is right now. I woke up this morning, and the first thing I did was, “Let’s go. We’re going deeper into Amazon. That’s what we’re doing today.” Dave, did you say that those flights aren’t actually operated by Amazon?

No, it’s a third party that operates the flights. I can’t remember the name, but I saw it. Like A21. I knew that they had talked about buying planes.

Do they own planes?

I’m sure they own planes. I’m sure they do, but this particular flight was not operated by Amazon.

Okay. Was it just branded on the tail?

Just branded. Amazon paid for the paint.

Oh, my gosh. Hold on. Sorry, I’m having to send a note here to my wife. Speaking of sending a note, did you remember to post this episode on X?

Yeah, I did.

Oh, good. Good. I got it during your opener, dude. During your opener, too. Perfect. So, like, for me, I added one new company, Unity. I’m now starting to do meaningful work on SoftBank as I assess the impending OpenAI IPO. I think once Anthropic has a successful IPO—and I think they probably will—everyone’s heads will immediately go to OpenAI, and then they’re going to look for OpenAI-correlated beneficiaries in the public markets.

Amazon’s obviously one of the biggest. Microsoft is a big one as well, obviously. SoftBank is kind of the purest way to play OpenAI right now, I think, and I think the market will wake up to some of that. I’m not in a huge hurry, but I’m going to do some work on SoftBank this week.

Those are kind of the 3 things I’m looking at. And then just run-of-the-mill infrastructure. You saw what happened Friday, guys. Friday’s trade showed us what the market’s thinking on memory, right? Memory’s up. It’s trickling up more today, but Friday was the big memory move.

I do think we’re going to hear more about memory efficiencies in the near term, but the market seems to finally be smart enough to realize that’s a positive and not a negative for memory. It feels very different this time.

Remember when this happened with the open models? Memory just got tanked on that. Now OpenAI comes out with a more memory-efficient model, and memory goes up. That’s a very different reaction than with the open models. It’s kind of stunning, quite honestly.

If memory holds here over the next couple of days, I think that’s a really positive sign that the market is getting smarter about what is real and what is just fear as it relates to these variables that are hitting the AI sector every few weeks.

And I love—I love—well, I mean, I say I love seeing that. I love it, but I kind of don’t, because it results in fewer inefficiencies. Before, the market was just making dumb moves, and if you had balls and cash, you could go all in. Then the market would eventually figure out that was the wrong decision, and those stocks would trade back up.

Now the market seems to be getting smarter, which means there’s probably a little less opportunity, but investors are going to be a little happier with less volatility every time we get a new story that memory costs are coming down or there’s a new whatever.

Jordan, are you making any changes?

I didn’t realize that the market was sideways to down today because Bloom is up so much. My portfolio is pure green.

By the way, Bloom is again such a clean, obvious beneficiary of this, right? We can have debates all day long about whether memory is becoming more efficient, whether memory stocks are tapped out, or whether NVIDIA’s GPUs aren’t as good as the TPUs. I don’t know. The debate can go on forever.

The energy trade still seems really clean and not widely debated, so Bloom looks great here. The only issue is that it has moved a lot, right? People have seen this move before in Bloom, and they get scared. They’re like, “Wait a second. We went from $170 to $270. I don’t know—is it at $270 yet? Are we even close to that again?”

$282.80.

Damn, dude, look at Bloom. So Bloom’s up to $280. I think investors have a knee-jerk reaction like, “Dude, that’s a big move. That’s a 70% move in a matter of weeks. Maybe it corrects here a little bit.”

But I love Bloom, and if it does, on any weakness in Bloom, I could see myself adding more. I love to see it, though. You know I went in deeper on Bloom. I went in deeper at $170. All the way down, I just kept adding Bloom, adding Bloom, adding Bloom, going, “It’ll come back.” And here we are, back at $280, right?

I love this stuff so much, guys. I love the inefficiencies in this market. I love how dumb investors are right now. I love how dumb Wall Street is. I love how scared everyone is and how quickly they get scared out of these trades.

I love this market so much because if you actually do your homework and have conviction, you’re just not freaked out when these moves happen. They’re opportunities. Every one of these moves down is an opportunity.

We’ll look back in a year, and people will be like, “How the hell did you get those?” Well, I just believed in these 3 stocks, and every time they crashed, I went in deeper.

But you don’t have to have these crazy, undiscovered stock picks. You just don’t need them right now. You don’t need to go deep into speculative stuff, which is so awesome, because I hate that. When you go deep into these speculative names, it takes so little for them to get cut in half.

I don’t think Amazon’s getting cut in half anytime soon. So I don’t know, Jordan, are you doing anything differently as a result of GPT-6 in the market? I’m curious what your concerns are—not specifically GPT-6.

Not specifically GPT-6. I mean, I’m buying energy companies, right? So I bought Vistra last week. It’s been beat up so much, and it’s just energy.

Talk to me about Vistra. I don’t understand. What’s your take there?

It’s just a pure-play power-consumption play.

It looks like that’s a company that’s pretty much flatlined. I mean, they’re all over the place, but are you just saying you love it?

Just general energy consumption. I just don’t think you can go wrong. We’re not going to use less energy over time with all of this going on with AI.

So they’re just a moderate P/E.

Yeah. It’s a safe energy play.

So boring. Okay, I mean, I get it. That’s so you. Anything else that you’re in meaningfully, or changing up at all, or adding to?

I’ve been adding to Amazon also.

Yeah, there we go. Nebius, I went in. During this big drop that we had a few months ago, I went into Nebius. I went into Amazon more.

Is it called Nebius? I’ve been saying Nebius. Is it called Nebius?

No one knows. I call it Nebius. I actually don’t know either.

Yeah, I think it’s Nebius. It’s one of those stocks I read about, but I’ve never heard anyone talk about.

I asked AI. I keep warning people: be really careful if you ask AI what the best stock picks are for something. I don’t think it’s great, because it didn’t come up with Amazon as a top-5 trade on Astra. Then I questioned it. I said, “Listen, objectively, don’t give me any crap here. Be truthful. How are you missing Amazon?”

I gave it my Amazon thesis. It put Microsoft up there. It put Bloom, NVIDIA, and CrowdStrike. It had Unity in there, which was interesting. It didn’t have Amazon.

Then it was like, “Oh, dude, you’re totally right. It’s a top 3. It should probably be higher than Microsoft as a trade.” And I do agree with it.

Amazon’s now got 3 separate ways to win. They invested $15 billion plus another $35 billion in OpenAI. Infrastructure, obviously. OpenAI expanded its AWS agreement by $100 billion over 8 years, including 2 gigawatts of Trainium capacity for OpenAI. Huge.

And then, of course, distribution. AWS is the exclusive third-party cloud distributor for OpenAI Frontier, while Amazon and OpenAI are jointly building the stately agent runtime for Bedrock. I don’t know what that word means. Maybe Jordan knows what that means.

It’s like, oh, yeah, they’re definitely a top 3. Amazon’s my play. I know it sounds boring, guys, but I’m happy to be in boring old Amazon and levered up in it, because I feel like the closer we get to Amazon earnings, the closer we get to learning just how much all of this is benefiting Amazon.

I’m starting to feel really good about this next earnings report, although it’s still quite a bit away. Where are we—7 weeks away now from the next Amazon earnings?

I haven’t been tracking it. I don’t know. 6 or 7 weeks.

I feel like the market just doesn’t—

I thought we just had an Amazon earnings, wouldn’t we?

Well, it’s been a bit. It’s been weeks.

Has it been a bit?

Yeah, it’s been a bit. It feels like weeks or days lately. Everything flies by.

I know, man. I know. October 29th.

So you’re talking, yeah, I mean, 7 weeks away. I think the closer we get to Amazon earnings, the more people are going to realize the inevitability of what’s going to happen on that day.

People are going to be like, “Wait, why did we trade Amazon down from last earnings?” I just can’t get enough.

I really can’t. And I love it. I love that we’re not here having to pick all these speculative stocks. By the way, I’m in the whole AI infrastructure chain. And I do think, generally, Astra is positive across the board. I just think there are better and not-so-great trades here on Astra, and I’d rather be a little bit less speculative.

So you’ve gotten boring. Jordan’s gotten boring. I don’t have a trade. I’m in Amazon, Bloom, and Tesla.

Yes. Although I’m in Micron, too. I bought the dip on Micron.

Have you seen the Vanguard Energy ETF returns year to date?

It’s up 44%.

Yeah, but you’re mixing in oil there, too.

Yeah, it’s a bunch of oil. I’m not long oil right now. I’m long grid power through Vistra.

Yeah. That’s just all related to the conflict in Iran.

Yeah, I do think it’s going to be really interesting, guys, to see how Anthropic plays out here going into this IPO.

Which is great. It’s just a different trade, right?

Yeah. I do think that it’s going to be really interesting. Even with OpenAI’s Astra really outperforming what people thought it was going to be capable of with GPT-6, a lot of people I hear are saying, “Hey, I still prefer the way that I’m able to work with Anthropic as a developer.”

There’s certainly enough demand to go around for both companies, including open models, which we know have been taking a decent chunk of compute. And that’s okay, because the open models are taking the more repetitive, not-so-complex tasks. So, as we can see, companies are basically taking more of a broad approach in terms of, “Hey, we’re going to utilize it as intelligently as we can.”

Does that even matter if you’re invested in Amazon? Because aren’t people either going to run them at home, which is just terrible, or put them on the cloud? And so, if you’re invested in cloud infrastructure—

My mind keeps coming back to Amazon over and over and over again. Because no matter how this AI cycle plays out, no matter how big the open-source piece of it is, I believe the most pragmatic take on everything we’ve seen the last 2 or 3 months is that it’s going to be a blend of open models and closed models for full frontier capabilities.

And so that should really only matter if you’re heavily invested in OpenAI or Anthropic. If you’re invested in any of the neoclouds or any of the traditional hyperscalers, it shouldn’t matter. It really shouldn’t matter.

For the most part, a lot of people who are using things perfectly efficiently, who are on the bleeding edge and able to optimize every dollar that they spend, are very vocal about it, especially on X. But the reality is, the global economy and the incumbent companies of the world do not operate at the bleeding edge of efficiency. Is that a safe assumption, Jordan?

Meaning that they’re probably spending more money than they should on everything, including compute, including the way Amazon operates, and everything else that they do in the cloud. And you’ve got to pay for experimentation, too, right? These companies are always experimenting with different models.

Yeah, they’re experimenting. They’re bringing more of their business into an AI world.

So even if there are tasks that are highly repetitive and can now be done meaningfully cheaper with compute through an open model, that doesn’t necessarily negate the amount of money that they also want to spend on a frontier model for best-in-class intelligence.

Also, companies like Amazon, that are able to offer you the full poo-poo platter of compute, from cheap to best-in-class, kind of win in this scenario. You need to be diversified, and you need to try to split up the type of compute that you use. Companies like Amazon can help you do that, so it feels to me like my head keeps coming back to Amazon every single time.

Also, I think one of the things we haven’t seen yet, because it’s always kind of splashier to show the entire island of Manhattan getting replicated in 5 days or a 3D world getting replicated, is advertising. Again, you can now create the most immersive digital ads with a single-shot prompt, and that is going to massively benefit digital advertising platforms online, of which Amazon is the third-largest in the world.

I think it’s the fastest-growing because now advertising quite simply gets better and more accessible to more companies and more people that want to utilize it. So the more accessible, highly immersive advertising that’s not just laser-targeted but perfect for your product—imagine having a $200,000 video for your product on a $20 budget. You can actually do that from 1 photo you took with your iPhone, and then that perfect, beautiful ad gets customized for each individual viewing it. It’s a whole new world of advertising, too.

No one’s talking about that yet, but Amazon is in the pole position as a beneficiary of being able to offer all the world’s companies—which are pretty much on Amazon, as they are also on Meta and Google—the ability to create better-performing ads. Those are the 3 big platforms.

They’re able to offer companies better-performing ads, which ultimately benefits the advertising platform, Amazon. Because when your ad performs better, you’re able to spend more money on that ad because it converts better. So it’s one of those things where everybody wins, and no matter how you look at the progress and traction of these frontier models, I think it comes around to benefiting Amazon in every aspect of its business: the infrastructure layer, the chip layer, and the digital advertising layer.

And of course, the 1 thing we haven’t spoken about yet—gosh, guys—are we not seeing such an obvious end-client, real value delivered to the end client with these new models? It’s becoming clearer, which is the last hurdle to showing us that companies like Amazon are going to get massively more efficient throughout the entire global infrastructure chain, leveraging this actual type of intelligence.

Do you know how many things I’ve read this weekend from actual founders and chief developers who are like, “This is so game-changing”? It’s actually allowing us to identify problems in our business right now and address those problems in just 4 days. In 4 days.

So imagine what companies like Amazon are able to do across the entire chain of their global business by bringing efficiencies into every part of the business. And you know how Amazon is all about efficiency, guys. Could you even imagine what’s going on there right now? Could you even imagine what they’re doing right now to say, “Hey, in 2 years, how much can we bring down this $6 billion cost line using the infinite intelligence that we have access to?” And they have about 50 of those budget lines.

That’s in the works right now, and it’s being massively underappreciated by the market. Those are the things where, all of a sudden, we’re going to wake up in 6 quarters and Amazon’s going to be like, “Oh, yeah, the margins on our retail business and the margins on our logistics business are set to now increase by this amount over the next few quarters because of new functionality that we’ve been able to deploy—first of all, assess and deploy—due to the access to infinite intelligence that we have as a company.”

Yeah. We’ll start to see some return on the CapEx they spent on building the infrastructure, and now their cost structure is going down. Although they’ll continue having to build the infrastructure indefinitely, seeing efficiencies across the board will be exciting with Amazon.

I agree. So anyway, in wrap-up, I’ve got 3 stocks here, guys. 1’s new, 1’s old, and 1’s potentially brand new, but I haven’t traded it yet. And those are my Astera trades.

By the way, that S&P inclusion on Friday—Dave, on Bloom—what a nice gift into the weekend.

Oh, for sure. I love it. And another 10% today. I love it, dude. I love it.

All right, guys. We might, if we can pull it off, do 1 more show this week because we have something to talk about. We have a topic. It’s just a question of whether we’re actually able to do 3 shows within a week and a half, or whether that would break the internet.

I want to close this whole GPT-6 thing out right now by saying that if you’re an investor, dude, this is freaking huge, dude. I think the biggest takeaway from GPT-6 is that we are not only not done, we are starting to really accelerate on the AI side in terms of its actual capabilities.

Everything we were hoping we would see—which is doing more with less—is happening. And that’s exactly what the global economy needs to further embrace AI. The next big thing to drop, and I think we’re now quarters away, not years away, is this efficiency trade.

I think we are going to start seeing big notations in earnings reports about AI massively improving the underlying business of so many of the world’s companies. And the second that starts to happen, we will get an avalanche of earnings revisions. I think we will potentially see meaningful multiple expansion across a wide variety of companies that are not in the AI world today.

These are not companies that benefit from the AI trade today. The AI trade becomes ubiquitous for all industries.

Yeah, it’s not something that happens overnight.

Well, that’s what should happen, right? If you have this technology that’s going to revolutionize all of business inefficiencies and all these things, then, yeah, you should start to see it across the entire market.

So I know we’re a little ahead of the next earnings cycle, but if I’m an investor, I am looking for those cues. I am looking for mentions of the words, “AI has saved us. AI has led to this margin increase. It’s reduced these costs. It’s allowing us to expand into this new category.”

And I would just count the number of times that you hear that early in the earnings cycle. That is the alpha. That is the information asymmetry if you can catch it early. Because if that starts to happen with a handful of companies, I think it’s an inevitability that it will start to happen with hundreds of companies.

When you start to see it happen with hundreds of companies, you’re going to start seeing sell-side banks, financial media—everyone—saying, “Whoa, we now need to revalue the global economy based on this new world of AI.” And you’ll start to see companies like Vistra just selling a ton of electricity.

Yeah.

Yeah, but Jordan, this is like you. So look, your argument about Amazon is the same as my argument about electricity. Right? If you think about what these AI companies are selling, they’re selling the output of these GPUs. And how do they do it? They do it by consuming electricity. So it’s the same thing.

Fair. You’re right. The less sexy version—an even less sexy version.

It’s less sexy, but I mean, that’s the ultimate. That’s where all these tokens get generated: through electricity generation.

So Bloom is one way to play that. Vistra is another one. Bloom is the more exciting one, but I think a lot of the upside has been consumed on Bloom. Maybe, maybe not. I think Vistra has been beaten up.

If you look at the amount of electricity that they’re estimating these data centers will ultimately consume, even in Texas, it’s about double what we’re currently consuming. I think the big question, Jordan, is, if you develop a trade window around it, when do you think the market will come to terms with that as the ground truth?

Yeah, no, I don’t really care because I’m not buying Optin, so I just own the equity. I just don’t want to be in it for years, right?

No, I get that’s not what you want to do. I can do that and feel great about it.

Yeah, but I want to make money in the next 90 days.

Yeah, and I get it. I’m patient on it.

That’s fine. I’m excited about the next Amazon earnings, even though we’re 7 weeks away. I’m starting to prepare right now. Guys, I have a lunch. It starts in 5 minutes. I’ve got to hop. All right. Well, on that note, thank you guys for watching. We’re Dumb Money. We will see you very soon.