YT YouTube2Text

YouTube2Text → Transcripts → AI Investing Got Weird

AI Investing Got Weird — Transcript

by Equity Empire · 4,007 words · 617 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00So, right now, Open AI and Anthropic are
0:02adding revenue faster than I think any
0:05company in the history of American
0:07business, but something is changing. Cuz
0:11over the past 3 weeks, I've sat through
0:12more than two dozen earnings calls, and
0:15every executive is starting to sound
0:18exactly the same. And here's the crazy
0:20thing. A year ago, all these executives
0:23were trying to get closer to Open AI and
0:26Anthropic. They were trying to name drop
0:27them in a conference call. trying to get
0:30out a press release with the company's
0:32names in them that they're working
0:33together. But today,
0:35the biggest companies in the world are
0:37quietly and kind of quickly trying to
0:39build as many ways around these AI
0:43frontier labs. The reversal is really
0:45only taken about 8 weeks. And look, by
0:49the end of today's video, you're going
0:51to know exactly how to spot what I
0:54believe is going to be the next winners
0:58in AI. And I'm calling this the AI layer
1:02trade. So, here's what I'm going to walk
1:03you through on today's video. Number
1:05one, I've pulled up several key moments
1:08from conference calls that I've listened
1:10to over the past few weeks that is going
1:12to tie this AI layer framework that I
1:16have together. Number two,
1:18there's a big giant paradox that makes
1:22some of this that I'm going to talk
1:23about honestly kind of confusing.
1:25Because like I said at the top, Open AI
1:27and Anthropic are growing faster than
1:31ever before while their own biggest
1:33customers are literally trying to
1:36actively replace them. I'll show you why
1:40both of those things can actually be
1:42true at the exact same time. Number
1:45three, I have a six-layer framework.
1:49This is really important that I'm using
1:51to decide which AI companies that I am
1:55going to buy and I'm going to recommend
1:57to subscribers. This framework, I think
1:59it's it's not only going to help you
2:01pick stocks to invest in, it also helps
2:05you hold on and have conviction around
2:09stocks because as we've seen over the
2:11past few weeks,
2:13some of these AI stocks can go down a
2:15lot very quickly. And so if you have
2:18high conviction, you can buy more and
2:20you can continue to hold for the next
2:22leg higher. So,
2:24one quick note though, before I got to
2:26get into these things cuz I I think this
2:28is going to help you understand my
2:30perspective here. So, I've spent the
2:31past 25 years
2:33building stuff on the web. This is front
2:36end, this is back end, these are
2:38websites, database applications, even a
2:40handful of websites that I scaled up and
2:43I sold them on Flippa when that was like
2:46kind of an easy thing to do. Now, it's
2:48probably too easy and there's not enough
2:50money there. So,
2:52long story short, I'm not just an like a
2:55user of this technology, I'm not just an
2:57investor, I literally rely on it every
3:00single day. It's paid my family's bills
3:02for a long time. And I think when you
3:04actually build with this stuff, when you
3:06build with AI, I I think you have a
3:08pretty good read on which technology is
3:11worth paying for and which technology
3:15you can eventually stop using. So, let's
3:18jump into the earnings calls because I I
3:21think this is what's really important.
3:23So, when you sit through like over 20 of
3:26them like I have over the past two and a
3:28half three weeks, you start to notice
3:31patterns showing up company after
3:33company. Different industries, different
3:35businesses, they're all essentially
3:37starting to say the same thing. It's
3:38it's crazy. When Microsoft reported
3:41their earnings a few weeks ago, they
3:42told investors that this major shift was
3:46happening with clients that they were
3:48working with. In fact, I I the example
3:50they gave was like Levi's Strauss or
3:53something. So, instead of Microsoft
3:55Copilot relying on a single model like
3:59it did when it launched, customers
4:01started to use multiple AI models
4:05depending on the task and how much that
4:08model was costing them. Here's the CEO
4:11of Microsoft and this is what he said on
4:14the conference call.
4:15>> Since the start of the year, we have
4:17seen five x increase in the number of
4:19customers building with models from
4:21multiple providers.
4:23>> So, I think his investors and maybe even
4:24consumers in technology were somewhat
4:26used to one dominant operating system
4:30like Microsoft or one search engine like
4:32Google, but Microsoft's customers have
4:35stopped committing to a single AI
4:38company and they're actually starting to
4:40use several at the same time. This is
4:42what they want and Microsoft actually
4:45took it a step further than just making
4:48this swap possible. On the same earnings
4:50call, Microsoft announced that it built
4:53more than a dozen additional AI models,
4:57ones for image generation, voice
5:00transcriptions, coding, cybersecurity,
5:02all these different types of things.
5:04So, Microsoft is no longer selling its
5:07customers just OpenAI's technology.
5:10Microsoft is selling essentially an AI
5:13store that carries a bunch of different
5:17AI models. So, in Microsoft
5:19wasn't the only one.
5:21This version of companies wanting to use
5:24multiple AI models, it literally played
5:27out conference call after conference
5:30call. The company that probably said it
5:33the most bluntly was Palantir. Palantir
5:36builds, as probably you know,
5:39some of the most widely used AI software
5:41for corporations,
5:43government agencies. I would consider
5:46Palantir one of the largest builders of
5:49real production AI custom systems. And
5:53here's what Palantir's chief technology
5:56officer told investors on its conference
5:59call.
5:59>> The assumption that the frontier is
6:01actually the best performing is just not
6:03born out in practice. Within 24 hours of
6:06bringing Nematron Ultra into our stacks,
6:08we found five production tasks where a
6:11standard Nematron Ultra model without
6:13post training beat frontier models.
6:16>> So there's a lot he said there and let
6:17me kind of break it down for you.
6:18Nematron Ultra is an AI model that
6:21Nvidia built and gave away for free.
6:24We'll put free in air quotes because
6:27anyone can download it and run it on
6:29their own expensive Nvidia hardware, but
6:33other than that, it doesn't actually
6:34cost anything. Palantir took that free
6:38Nvidia model, didn't change anything
6:40about it, and it pointed it at stuff its
6:42customers were already paying the
6:45frontier labs like Open AI and Anthropic
6:48to do. And inside of a single day,
6:51Palantir found five real legitimate
6:54production jobs that the free model that
6:57was made by Nvidia actually produced
6:59better results than the models built by
7:03the most valuable AI companies in the
7:05world. Then Palantir went on to explain
7:09why it would rather do business going
7:12forward in this way.
7:14>> We built a partnership with Nvidia.
7:15We're expanding our application layer.
7:17We are going to
7:19enter the market and already entering it
7:20in the classified space as Sham alluded
7:22to of of fine-tuning models so that the
7:25models actually fine-tuned by us in our
7:28enterprise on an Nvidia stack outperform
7:31frontier models and you own the weights,
7:33you own the alpha, you own everything.
7:36>> So owning the weights is essentially
7:39owning the AI model itself. It's kind of
7:42the same way we used to own music. Like
7:44we used to own the DVDs and the CDs and
7:47those types of things rather than
7:49streaming them and renting them.
7:51Palantir is telling its customers they
7:54can own the model outright. That
7:57honestly is great for governments and
7:59their types of clients. So instead of
8:01renting access to somebody's else's
8:03model that can be changed, re-priced, or
8:07even shut off without warning. And all
8:10three of those things, models changing,
8:13models changing price, and them being
8:16cut off and shut off without warning,
8:19all three of those things have happened
8:21over the past couple of months. And
8:23Palantir's customers can't deal with
8:25that. So the largest technology
8:27companies in the world are adapting to a
8:30world where AI models can easily be
8:34replaced. Which should mean
8:37these freaking AI labs like Anthropic
8:40and Open AI are in big deep trouble,
8:42right?
8:44Except that's not what's happening at
8:46all. Open AI and Anthropic are growing
8:48faster than ever right now and it's
8:50really any point in their companies'
8:52history. It's actually been pretty good
8:55and well documented at this point that
8:57Anthropic ended December of last year
9:00with about $9 billion
9:03in annualized revenue. So less than a
9:05billion dollars per month in revenue.
9:08Now outside researchers and Anthropic
9:11probably leaking the information out
9:13into the public,
9:14they're at over 70, 70,
9:19billion dollars of annualized revenue in
9:23July. I can't think of a company in the
9:25history of American business that has
9:28added that much revenue at that speed. I
9:31mean it is remarkable. Open AI is also
9:34seen a re-acceleration of its business
9:36particularly on the consumer side. At
9:39the end of July, OpenAI said it passed 1
9:42billion active users. So, the two
9:44companies
9:46whose products are supposedly becoming
9:48replaceable, and that's what the
9:50executives are saying, they're actually
9:52growing
9:53faster than ever before. And over those
9:56same weeks, their largest customers
10:00are actively trying to spend less with
10:02them. Even firms outside of big tech are
10:05kind of seeing this trend and trying to
10:06participate in it. Here's the chief
10:08executive officer of Uber describing
10:12this on the company's earnings call.
10:16>> A few years ago, many expected AI to
10:18converge around a single foundation
10:19model. Instead, multiple frontier models
10:22have emerged alongside a growing
10:24open-source ecosystem.
10:26>> Now, look, I I
10:27I realize we live in a world where it's
10:29always good versus evil. It's Democrats
10:32versus Republicans. It's Lakers versus
10:34the Celtics. It's thumbs up or thumbs
10:36down. And by the way, if you're still
10:37watching this video, give me a thumbs up
10:39or a thumbs down on the video, either
10:41one, however you feel. But here's the
10:43crazy thing about AI right now. The
10:46demand for artificial intelligence is
10:48real. I think we can all agree on that.
10:51And I think we can all say it's off the
10:54freaking charts right now. The companies
10:56that make the AI models are capturing an
10:59extraordinary amount of revenue and
11:01growth right now. And their own biggest
11:05customers at the exact same time are
11:07building ways around them. All three of
11:10those things are true all at the exact
11:12same time. It's literally the most
11:15extraordinary thing I think I've seen in
11:16technology. And also, I think this makes
11:20investing in artificial intelligence a
11:22very hard thing because the horse race
11:25of what model is ahead and what
11:27companies are using what, and it's
11:29difficult to track all of this stuff in
11:32real time. So, a while back I actually
11:33stopped trying to figure out what model
11:37was the best and which model people are
11:40using, which one's in the lead. Instead,
11:42I'm starting to track where the money
11:45settles in real time. I'm doing this.
11:48So, the way I'm doing this is I'm
11:51starting to treat the whole industry as
11:53a stack. Every piece of it sits upon
11:57another piece, the way kind of like a
11:59house is built. So, you have the
12:00foundation below the building, then you
12:02have the floors, you have the walls, and
12:04eventually you've got the roof. The
12:07kind of the original version of this
12:09framework is is not mine. I'll admit
12:11that. It actually comes from Jensen
12:13Huang at Nvidia. He laid out what he
12:16called the five layers of this stack in
12:19an interview at Davos earlier this year.
12:23And I thought it was actually one of the
12:24cleanest explanation that really he or
12:27almost anybody has given about what's
12:29happening right now. He described AI as
12:32a five-layer
12:34cake. Now, since then, I've actually
12:37added a sixth layer of my own. And this
12:40sixth layer is actually one of the
12:42biggest reasons why I'm making this
12:44video, and I think it's really important
12:46you understand this. So, let's go
12:47through the cake right now. At the
12:49bottom layer of the AI cake or the AI
12:53framework that I have is the power
12:56industry, the utilities, the nuclear
12:59operators, the solar companies, anybody
13:02building electrical or electrical
13:04equipment or anything that moves powers
13:07into buildings.
13:08Companies that are popular here are like
13:10GE, Vernova, and obviously the power
13:13providers. These are pretty popular
13:15stocks, and I don't think you need me to
13:16tell you which ones to invest in. But, I
13:19will say over the past year I've
13:21actually recommended Eaton and Emerson
13:24Electric to our subscribers. And both of
13:27these stocks have comfortably
13:29outperformed the S&P 500 this year. So,
13:33above the power layer is the
13:36semiconductor chips. This is obviously
13:38Nvidia, Broadcom, AMD. You also have the
13:42companies that manufacture the
13:44equipment. This is Taiwan Semiconductor.
13:47This is ASML, Applied Materials. You
13:49even have the memory makers now. This is
13:51Micron and this is Sandisk. This layer
13:54is extremely difficult to replicate. And
13:57the competition in this layer
14:00doesn't just come out of nowhere. These
14:02firms have very, very durable moats
14:05around their business. Now, above the
14:07semiconductor layer
14:09is the data centers. So, this is Amazon
14:11Web Services, Microsoft Azure, Google
14:14Cloud, Oracle. There's also the smaller
14:16Neo Cloud operators like CoreWeave
14:19and Nebulous. So, above the data center
14:22layer is the layer that I've added
14:25myself. I call it the infrastructure
14:28layer.
14:29It's easily the best performing stocks
14:32in what the market often refers to as
14:34{quote} software stocks. I'll come back
14:36to this in a moment cuz these aren't
14:38software stocks. This is way better than
14:40software. Now, above the infrastructure
14:42layer are the AI models. OpenAI and
14:45Anthropic, Google's Gemini, Meta's doing
14:48some stuff, Elon Musk is doing something
14:50with Grok.
14:51You also have all the open models coming
14:53out of China, Nvidia, other companies
14:56out there. And finally, above that is
14:59what I will call the application
15:01software layer. These are the
15:03applications that you click on, that you
15:06visit, and that you download. This is
15:08Salesforce, this is ServiceNow, this is
15:10Adobe, among many others. Now, here's a
15:13key thing to this entire cake or this
15:17framework, however you want to call it.
15:19It's the test that I run on every layer
15:23before I decide if the company is worth
15:26investing on.
15:27And you can run this test yourself in
15:30like 10 seconds if you understand how
15:32all this works.
15:34If the company disappeared tomorrow,
15:36[clears throat]
15:37so you're going to run this test. If the
15:39company disappears tomorrow morning,
15:41what happens to artificial intelligence?
15:43Let's start with ASML. This is the one
15:45company that builds the world's
15:46lithography machines that print the most
15:49advanced chips. If ASML vanished
15:52tomorrow, the entire leading edge
15:55semiconductor industry and the AI
15:57industry
15:58essentially would stop in its tracks.
16:01Let's take Nvidia. Every frontier model
16:04on earth including the open source stuff
16:06being built in China.
16:07It's largely trained and built on top of
16:09Nvidia hardware and software, honestly.
16:12Remove Nvidia and you set the whole AI
16:15industry back years, probably. Now let's
16:18take Anthropic.
16:20Anthropic could cease to exist tomorrow
16:22morning. In fact, some of its models
16:23have actually had that happen to them.
16:26And the AI industry didn't even blink.
16:28The work would route to Google, it would
16:31go over to Open AI or they would
16:33download a free model
16:35uh probably by the end of the week,
16:36maybe even by the end of the day. Remove
16:38Adobe and people would just find new
16:41ways to edit videos and you know, edit
16:44photos. It wouldn't be that hard.
16:47That is the whole framework
16:49in one single test. And it honestly, if
16:51you do that framework across all of
16:53technology, it produces a pattern. The
16:56money, all the profits, all the profits,
17:00all the cash flow is accumulating at the
17:03bottom of the stack and it starts to
17:06thin out the higher you start to go. The
17:09reason is because of substitutes. There
17:12is one company on earth that makes the
17:15machines that makes all the chips.
17:18There's only three or four companies in
17:20the world that can actually design and
17:22manufacture leading edge chips for AI.
17:25There are like four or five companies
17:26that can actually afford to build data
17:28centers at scale.
17:30And so by the time you reach Anthropic
17:34and OpenAI at the model layer,
17:36there's dozens of options and several of
17:39them now are free open source. And so
17:42the very top is the application software
17:46layer. There are literally thousands of
17:48companies
17:49at that layer and you can now write code
17:52and build replacements for them
17:54internally and that's what companies are
17:55doing.
17:56The higher you climb in this AI stack,
18:00the easier you become to replace. Easy
18:04to replace
18:06means you don't get to set your own
18:08prices. It means another company can
18:10come out of the woodwork and start
18:12competing with you and I honestly don't
18:15think it's where you want the bulk of
18:17your money as a tech investor. Now,
18:22this now brings me to the layer that
18:24I've added personally.
18:25Between the data center and the AI model
18:29layer
18:30is a layer that almost nobody pays
18:32attention to or they do the mistake and
18:34they put it at the top layer. They put
18:36it up with software.
18:37It stores and organizes your company's
18:40data. It watches the entire AI system.
18:44It watches and makes sure AI agents are
18:47behaving and have permissions and doing
18:49all those types of things. It secures
18:51all of this as well and it connects
18:53everything together in a way so all the
18:56pieces can talk to each other.
18:59This is databases. This is data
19:01warehousing. It's monitoring. It's cyber
19:04security. It's integration software.
19:07This is Snowflake. It's Palantir. It's
19:09MongoDB. It's Databricks. It's Data
19:11Dogs. It's CrowdStrike. it's Palo Alto
19:13Networks. On the large cap side, you
19:16have Microsoft, Google, and Amazon. They
19:18all have these integrated tools inside
19:20of their software and their data centers
19:23as well.
19:24This layer sits a little higher on the
19:27stack, and by that logic, you'd probably
19:30think,
19:31"Um maybe it's not worth that much."
19:32But,
19:33it is worth a lot, and here's why.
19:36Everything above this layer is generic
19:39and in some cases open source or easily
19:42swappable. Everything inside of the
19:45infrastructure layer is yours and hard
19:48to switch away from it. Now, I've been
19:51putting money behind this framework for
19:52a while now. And and I'll tell you how
19:55it went, including the the part that was
19:58somewhat painful for a couple of months.
20:00So, I recommended Snowflake to Equity
20:03Empire subscribers last year. And it was
20:05entirely based on this framework logic
20:07that I've presented for you here. Not
20:10because it was an AI company or it was
20:12software, it was because it sat in the
20:15layer that I just described that I knew
20:18after 25 years of building web
20:20applications, it was very hard to
20:23replace a company like that once you
20:25started working with them. And look, I
20:27was early. The stock went down. And it
20:31when I say Snowflake went down, it went
20:33down a lot, and it stayed down. And
20:35subscribers started to question the call
20:38and started questioning if I was still
20:39committed to it. And I said publicly on
20:42this channel that I completely
20:44overestimated Wall Street's ability to
20:46understand what this company and any
20:49company in the infrastructure layer
20:51actually does. It's not software. And
20:53so, I averaged down the entire way.
20:57Fast forward to today, and Snowflake is
21:00up, I think it's over 80% in 6 months.
21:03It's I think literally the best
21:06performing quote software stock in
21:09really almost the entire stock market.
21:11Now, I'm not telling you this
21:13to brag or take a victory lap. I'm
21:16telling you because the framework is
21:19what helped me make the decision to
21:21invest in this company.
21:23Here's the most important part. The
21:25framework allowed me to hold and buy
21:29more when it got really ugly. The
21:32framework gave me the conviction to buy,
21:36hold, and keep buying
21:38until investors saw the critical piece
21:41of infrastructure software that
21:43Snowflake and many others are. So,
21:45that's the system.
21:47It's got six layers to it. It's got a
21:48single test. You can
21:50run it on any stock that you're looking
21:52for. Now, what I do for Equity Empire
21:55subscribers, if you're curious,
21:57is that's the part that comes kind of
21:59after this framework. Every company I
22:02work with is kind of mapped, at least in
22:05my mind, to what layer it sits in. I
22:07also give recommendations on what prices
22:10to buy, a buy range, give you the
22:12conviction when it starts going down
22:13like Snowflake, to like, "No, we've got
22:16to keep adding here." The price also, I
22:18would sell this. There's risk
22:19management. There's stop losses. We
22:21don't just hold and hope. But, in the
22:22case of Snowflake, we actually set the
22:24trade and the investment up for some
22:26volatility. It was actually pretty well
22:29executed other than being a little
22:31early. Now, also, what I do when
22:33earnings come out or new technology
22:34emerges, I go through the conference
22:36call. I make videos. I record videos. We
22:39look at the technical charts. We look at
22:40everything. And I let the subscribers
22:43know where I think the company is. Have
22:45they moved up? Have they moved down? Has
22:46new competition caught up? All those
22:48types of things. The framework though
22:50that I gave you on today's video is
22:51free. It's yours to keep. You can do
22:53whatever you want with it. The
22:55implementation on what to buy and the
22:58deeper analysis, the videos, the
22:59specific earnings videos, that's what I
23:02provide to subscribers. There's always a
23:04link to that in the description below.
23:07Whether you click on that or not, it's
23:09completely up to you. But, here's what I
23:11want you to take away from this video.
23:14Take any AI stock that you own, put it
23:17in one of the layers, figure out where
23:19it goes. Again, it's a little bit easier
23:21if you've spent the last 25 years of
23:23your life in these layers and digging in
23:26them and figuring out where everything
23:27goes. But, I think most investors like
23:30yourself can do that. Then ask yourself,
23:33"What happens to the AI industry if this
23:35company disappears tomorrow morning?" If
23:38the answer is everything stops in its
23:40tracks, then it's probably a great place
23:42to put your money. If the answer is
23:45"Somebody else would do the job by
23:47Friday," then you're near the top of
23:49these layers and I I wouldn't honestly
23:52allocate as much of my investment into
23:55companies at the top. Most people are
23:58going to spend though the next 2 years
24:00doing just that. They're going to buy
24:02the top of the stack because that's
24:05where all the headlines are.
24:07Those are the companies making the new
24:09software, the new thing that you click
24:10on, the new thing that consumers are
24:12using.
24:14That's where all the headlines are going
24:15to be, but all the money is going to
24:18trickle down to the layers I described
24:20at the beginning. So, I hope you have a
24:22better understanding of this. It's a
24:24crazy time. AI is on fire and at the
24:28same time companies are trying to figure
24:30out how to not use the AI labs. It's
24:33unbelievable. Now, we've just gotten
24:35past the major earning season. I've
24:37recorded a ton of videos for paying
24:38subscribers. That means that frees up a
24:40lot more of my time. I'll be back later
24:42here on the channel to help us walk
24:44through it and try to make sense of it
24:46the best we can. Thanks for tuning in
24:48today's video. If you like my content,
24:50please subscribe. Please like the video.
24:53Please tell a friend. I appreciate it.
24:56I'll see you guys again soon. Good luck
24:58with your investments.

About this transcript

This page contains the full transcript of AI Investing Got Weird by Equity Empire, generated from the public captions YouTube serves with the video. The transcript has 4,007 words across 617 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.