Semis vs Software Is Breaking the Market (Scarcity vs Abundance) — Jordi Visser (18 enero 2026)

[0:00] Um, we got a lot to cover today. Uh, so
[0:03] I'm not going to read through all this,
[0:05] but let's just say, uh, most of the
[0:07] story is going to be about scarcity
[0:10] verse abundance, about what's going on
[0:12] with semis and software, and then some
[0:14] of the releases on chat GPT health. But
[0:16] to start, let's go through basically the
[0:19] week. Uh, S&P small change down slightly
[0:23] for the week. Q's down a little bit
[0:26] more. They continue to underperform. Mag
[0:28] 7 underperforming worse than that down
[0:30] for the year. IWM's the outperformer we
[0:34] keep seeing small caps uh is basically
[0:37] about
[0:39] 8% uh sorry 7% in the last uh uh two
[0:43] weeks
[0:45] and when you look around the globe uh
[0:47] it's very easy to see that the worst
[0:49] performing market uh basically falling
[0:52] again the way it was last year so far is
[0:55] uh the US market. Uh this is a theme
[0:58] that I think is going to persist not
[1:00] just this year, not just from last year,
[1:02] but going forward. Uh I'm a big believer
[1:05] in foreign markets, emerging markets, uh
[1:08] and small caps and midcaps. I think this
[1:11] continues and again it's all about
[1:13] software. uh that may not be the way
[1:16] people are thinking about it, but I have
[1:17] equated uh in almost all of my
[1:19] presentations over the years uh looking
[1:22] back to 1993
[1:24] as when Netscape came out and the US
[1:26] dominated coding uh that effectively the
[1:29] dollar and the capital surplus, capital
[1:32] account surplus has grown since uh
[1:34] Silicon Valley basically dominated the
[1:36] world and now with AI that dominance is
[1:39] over uh and the disruption has started.
[1:42] So the NASDAQ, you're not going to see
[1:44] that too many times with a fully green
[1:46] board and the only market it is uh
[1:50] outperforming right now is the Swiss
[1:51] market. Uh when you go through it's this
[1:55] is not just a small cap midcap story and
[1:58] this is something I want to make sure is
[1:59] clear. This is uh about concentration.
[2:01] This is why it gets back to coding. It
[2:03] gets back to anything based on code that
[2:06] has worked over the last whatever years
[2:08] including things like Mastercard. So,
[2:11] uh, if you go through, this is the top
[2:13] names in the S&P 500. You can see how
[2:16] many names have actually outperformed
[2:17] the index. Out of the top, you know,
[2:19] this is, uh, 18 names, uh, because
[2:21] Google's counted twice in here.
[2:24] You have about three that have
[2:26] outperformed, sorry, five out of the 18
[2:28] that have outperformed the actual S&P so
[2:31] far. This is the S&P 1500. This is
[2:33] always a gauge that I've used to get a
[2:35] sense on concentration. much better than
[2:37] small cap uh midcap uh it includes it so
[2:41] it's impacted it but this story
[2:43] resonates across the S&P 500 as well so
[2:45] within large cap you're seeing the same
[2:47] thing which is this year so far has been
[2:49] about a broadening out and a lack of
[2:51] concentration this takes the S&P 1500
[2:54] all the names so far this year the S&P
[2:56] 1500 is up 1.75%
[2:59] and you have to go all the way down to
[3:00] name number 1
[3:03] so 107 out of the 1500 have outperformed
[3:07] the index. Uh which means the
[3:10] concentration you're getting u a
[3:11] broadening out. That's why you're seeing
[3:13] equal weight do well. And this is a
[3:15] dramatic shift from what we've seen over
[3:17] the last five years. This is the total
[3:18] of the last five years. So forget last
[3:21] forget um you know the prior years with
[3:23] chatbt. This is just taking the last
[3:25] five years. Uh the S&P 1500 was up 79%.
[3:30] All you had to do is go down to uh
[3:32] number 448.
[3:34] uh 448 out of the 1500 outperformed the
[3:37] index. So again, we have uh we've seen a
[3:41] change going on. And that's not the only
[3:42] change. This is uh an overlay between
[3:45] the NDX, that's the orange line, still
[3:47] sitting up near the highs. The other
[3:49] line is the Russell 1000 growth index
[3:51] relative to the S&P. So what's happening
[3:54] is growth or software. Take your pick.
[3:58] Whatever this stuff is, growth is
[3:59] getting hit. This is another major theme
[4:01] for me, but it all lines up with the
[4:03] same thing, which is to get in the
[4:04] growth side, to have the stuff that's
[4:06] been growing. We're running into a big
[4:08] disruptive phase right now where the
[4:10] commodity side, the scarcity side is
[4:12] starting to dominate the index. One part
[4:14] of the scarcity side, I've shown this
[4:16] before. I'm going to keep showing it
[4:17] because people are continuing to call up
[4:20] saying you got to buy software. Uh I
[4:23] didn't get many AI bubble uh talk themes
[4:26] showing up. No trending topics. So
[4:29] instead of AI bubbles, everybody who was
[4:31] an AI bubble seems to be trying to pick
[4:33] the bottom in software. I think you have
[4:35] to look in the mirror and say, am I
[4:36] bullish software because I don't believe
[4:38] in AI or am I bullish software for some
[4:40] other reason? Um this is the equal
[4:43] weight semi out of the 15 S&P 1500
[4:46] making new highs and now up 65% since
[4:49] the lows back in May of last year.
[4:52] Software made new lows. It's now only uh
[4:54] it's now down 6% or 5% sorry since that
[4:58] period. So we're continuing to see that
[5:00] divergence. Uh you're going to have
[5:02] periods where you get some kind of mean
[5:04] reversion. It could happen around
[5:05] earnings time could happen occasionally.
[5:08] Uh I just think this is a trend that is
[5:10] pretty much fading. Uh the prior
[5:12] technology in the decade disruptions
[5:15] were which were really the two big ones.
[5:16] Fracking which eventually disrupted and
[5:18] especially disrupted oil. Uh but the
[5:20] energy names ended up getting sucked up
[5:22] completely into it. And then you also
[5:24] had the mall names getting sucked up by
[5:26] uh by Amazon. Uh another week of just
[5:29] you can't I mean it it it's amazing. We
[5:32] have so many stories every week in terms
[5:34] of out of the White House and just
[5:36] things that people uh get bearish on and
[5:39] they end up fading away because there's
[5:41] so many of them. Um we had the PAL
[5:43] subpoena to start the week. Then we got
[5:45] the credit card proposal.
[5:48] Then we got the electricity bills geared
[5:50] towards the hyperscalers. Um I'm I'm
[5:52] bringing this one up twice because I've
[5:55] highlighted this for six months that the
[5:57] tech companies because electricity is
[5:59] just not a big portion of the cost and
[6:02] because they're going to be focused on
[6:04] building behind the meter will have to
[6:05] cover the costs on the grid. Uh I think
[6:08] now that is a given. It wasn't just
[6:09] Microsoft which was the first story uh
[6:11] that said they were coming in but now
[6:13] he's basically saying they're all going
[6:14] to have to deal with it. uh and that'll
[6:16] help on this front. Inflation continues
[6:18] to be a non-story. Uh if you're talking
[6:21] about inflation, it's more about what's
[6:23] going to happen in the future. Uh I'm
[6:25] sure we're not going to get down to 2%
[6:27] overall. At least that's my gut. Uh but
[6:29] I still believe that with house housing
[6:31] costs falling, wages uh under pressure,
[6:34] and the jobs market remaining weak. At
[6:37] the same time that we don't have oil
[6:38] going higher, which I think will change,
[6:40] but I don't think we're going to have a
[6:41] violent move higher. I think the
[6:43] inflation picture is going to stay uh
[6:45] basically in the 2 and a half to 3% 3
[6:48] and a quarter range. We're at 270. This
[6:50] is sticky inflation X shelter versus the
[6:54] core CPI. They're both in line. So,
[6:56] that's there. And then you have true
[6:57] inflation uh year-over-year core, which
[6:59] is actually making new lows and heading
[7:02] back down. So, take your pick. Inflation
[7:05] not a story. The other data out this
[7:06] week, we got IP out um the diffusion
[7:09] side, which should be the most
[7:10] correlated to the PMI. This is the
[7:13] industrial production three-month
[7:15] diffusion, five-month average related to
[7:19] ISM PMIs, which still remain low, and
[7:21] again, I talked about that last week.
[7:23] The global PMIs are actually up close to
[7:25] that level. Uh so, don't get caught
[7:27] thinking about the ISM PMIs. Eventually,
[7:29] that survey will go higher. Uh but where
[7:32] we are right now is that red line there
[7:34] is where the current reading is. This is
[7:36] the five-month average. So the
[7:38] industrial production side continues to
[7:40] get strong. This was out from the global
[7:42] macro investor team Raul and the gang. I
[7:45] will be down in Miami speaking uh down
[7:48] there and doing a live uh broadcast with
[7:50] Anthony Pompiano. So if anyone's down in
[7:52] Miami uh stop by and see me. Uh this is
[7:56] based on financial conditions. this is
[7:58] their own proxy. Again, highlighting PMI
[8:00] should go higher. And then we finally
[8:02] started to see the bounce. We'll see how
[8:04] the rest of the regionals do, but this
[8:06] is a very strong sign to start off the
[8:07] year. You have the Empire Manufacturing
[8:09] beating in a big way, coming up sharply
[8:12] higher than last month. Same thing for
[8:14] Philly Fed. Uh so the first two
[8:16] regionals to come out are highlighting
[8:17] this. I mentioned the scarcity set. I
[8:20] just want to kind of go through what I
[8:21] think the themes are that are secular
[8:23] for this year. Meaning I think these are
[8:24] the ones that are going to happen all
[8:26] year. It's not just a small cap midcap
[8:28] story. I want to be long scarcity. Um
[8:31] it's one of the reasons why whether it's
[8:32] gold, silver, uh copper, semiconductors,
[8:36] memory, um all of this is related to an
[8:39] underinvestment on the industrial side.
[8:41] Power fits into that. Short abundance is
[8:43] very simple. Uh that fits in with the
[8:45] growth story. Anything that has been
[8:47] isolated. Think about abundance as a
[8:49] broadening out. The opposite of
[8:51] concentration. uh we are running into a
[8:54] problem for anything built on abundance
[8:55] because AI is disrupting anything in
[8:57] terms of I want to be long bloated AI
[9:00] adopters healthcare companies financials
[9:02] health care companies I'll go through
[9:03] healthcare later and then short the high
[9:06] multiple AI infrastructure I still
[9:08] believe the data center buildout will
[9:09] happen I think there'll be at least two
[9:12] fear trades this year or one big fear
[9:14] trade related to bottlenecks slowing
[9:16] down the buildouts uh like we saw in
[9:18] Oracle style selloff but in a bigger way
[9:21] all of those names teams that uh have
[9:23] already built in the next couple years
[9:25] of the buildout that that are going to
[9:27] suffer. If there's any slowdowns, I'd be
[9:29] very wary of that. Uh short non AI
[9:32] native knowledge workers, long AI native
[9:35] knowledge workers. This is a theme for
[9:37] me personally. I'm spending a lot more
[9:39] time doing training sessions for asset
[9:41] managers and in particular, I'm starting
[9:44] to help a lot more college kids. If you
[9:46] guys are interested in uh I'm going to
[9:48] show some of the stuff today and
[9:49] obviously the payw wall go up but this
[9:51] is a big story. You have to get on top
[9:53] of the uh AI usage. You don't have time
[9:56] to wait and remember bottlenecks matter
[9:58] more than narratives this year.
[10:01] I released this this week on the power
[10:03] constraint. For those of you who got it
[10:05] um you got to see it. If you didn't get
[10:07] the names on it, reach out to the 22V
[10:09] crowd. I think I gave 23 24 names that
[10:13] are best associated with it. Some of
[10:15] them fit in with those bottleneck crews,
[10:17] but there's a lot on there that I've
[10:19] already kind of fallen off where you can
[10:21] uh find some names. The driving factor
[10:23] behind that was the uh this build out. I
[10:26] just wanted to show this and just make
[10:28] sure people realize what Alon Musk is
[10:30] doing with the build out of his data
[10:31] center. You can see the scale and size
[10:33] of this and his play on Microsoft macro
[10:36] hard. Um nobody else can build as fast
[10:39] as Elon Musk. And for those of you who
[10:42] keep wondering when I'm going to say
[10:43] Elon correctly, uh, I apologize. A good
[10:47] friend of mine used to work for me in
[10:49] Brazil named Elon. And I keep saying
[10:51] Elon for Elon. So, I will try to correct
[10:55] myself and be on top of it for those of
[10:56] you reminding me. Uh, Oracle, again, I
[11:00] just want to go through this fall that
[11:02] happened. I think you're going to see
[11:03] these in many of the names. If they
[11:04] miss, you're going to end up in this
[11:06] situation where they're still going to
[11:08] produce good returns for the year. I
[11:09] think they'll underperform, but I think
[11:11] it's a much harder trade. You're
[11:13] swimming upstream because of the
[11:14] crowdedness. When there's a broadening
[11:16] out, by definition, what it means is I
[11:18] got more choices. So, if you've got more
[11:20] alpha choices and that that se the stat
[11:24] I showed you of how few names uh under
[11:27] outperform the prior five years, a lot
[11:30] more names outperforming now. So, if
[11:31] you're in the crowded names, the the
[11:33] coding names, the software names, and
[11:35] you're trying to pick a bottom, I think
[11:36] you're at at risk. Um, in terms of the
[11:38] power names, I just wanted to give you
[11:40] guys a again, this is something I put
[11:42] out at the end of December. Uh, this
[11:45] remains a big story, but it became a
[11:47] bigger story this week. So, as a
[11:49] reminder, the thing that I'm kind of
[11:51] spending with people on, this is the
[11:53] process that I go through to get to the
[11:55] names, all of this in here is very
[11:58] important and part of what I'm starting
[12:00] to tell people. This human element here,
[12:02] which cannot be replaced, this human
[12:04] element here. So, in terms of picking
[12:06] the ideas and then getting to the
[12:09] selection or the filtering of the names,
[12:10] the rest of this is what you have to
[12:12] incorporate in there. This is how I go
[12:14] through my process. Um, basically to
[12:18] start off, I have a skill that's
[12:19] created. It does all of these things and
[12:23] you end up with
[12:25] at the end this. So, this is the stuff
[12:28] for 22V. I've got Qualcomm highlighted
[12:30] here because it was from last week, but
[12:31] there's a reason why it's also
[12:32] highlighted this week. But these are the
[12:34] names with inside the advanced packaging
[12:36] which fits into the stories that's
[12:38] really floating around on data
[12:40] management. I think it's important for
[12:42] you if you haven't spent the time to go
[12:43] understand why these names are doing
[12:45] what they're doing that these are the
[12:47] lists of the names. I created an index
[12:48] called package for advanced packaging.
[12:51] So all of those names here this is an
[12:53] index equal weight of all of those. So
[12:55] far it's outperformed the market
[12:57] significantly. It's up 12%. These are
[12:59] the semi-name mainly semi-names that are
[13:01] in there, but it's primarily a
[13:02] semiconductor story. There's some
[13:04] packaging things in there as well for
[13:06] AMC core and uh along the lines ASML.
[13:09] But if you look back over the last four
[13:11] years, this is not done what Nvidia and
[13:13] Broadcom have done. This is more like
[13:15] the PMIs. This is more like the XLE.
[13:17] This is more like the things that have
[13:19] been consolidating related to PMIs. This
[13:21] is the very very early innings of what I
[13:23] put out. If you want more details on it,
[13:25] again, reach out to 22V. Happy to go
[13:27] through this. Qualcomm did have a very
[13:29] good CES. It has underperformed
[13:31] significantly so far this year. It's
[13:33] down about 6% year to date. That name
[13:35] fits in with the group. It also fits in
[13:37] in a big way with the enterprise uh
[13:40] that's going to have to build out on
[13:41] premise. It will benefit from uh the
[13:44] Apple launch in terms of the iPhone uh
[13:46] AI. I've I've mentioned I got a new
[13:48] iPhone. Phenomenal uh uh increase for me
[13:52] in terms of productivity, because of the
[13:53] battery, and because of the speed. Uh
[13:55] but also it's going to benefit from the
[13:58] computer side as we do the roll out. I
[14:00] think companies are going to want their
[14:01] employees once they have the agentic
[14:03] side set up which is going they're going
[14:05] to want them to have iPhones. So I think
[14:06] the enterprise uh upgrade cycle will
[14:09] happen this year even if it's in the
[14:10] second half. We did get earnings from
[14:12] TSMC. I'm not going to read all the
[14:14] details but basically we believe the AI
[14:17] is real. Not only real starting to grow
[14:19] into our daily life. He revealed the AI
[14:22] accelerator revenue accounted for a high
[14:24] teens percent. Remember TSM has been one
[14:27] of the ones holding back uh supply out
[14:30] of fear.
[14:32] He discussed this. This is the CEO. The
[14:34] bubble commentary when asked about an AI
[14:36] bubble. He stated careless investment
[14:38] would be a disaster for us given that
[14:39] we're spending over 50 billion. He
[14:41] emphasized that Taiwan Semi does not
[14:43] build capacity based on forecast but
[14:45] based on committed demand. So basically
[14:47] was saying he already has the demand.
[14:49] They're still being very conservative.
[14:51] Uh he noted higher capex which again
[14:54] they're increasing their capex
[14:56] significantly from 42
[14:58] to 52 to 56. Uh insatiable demand for AI
[15:03] related computing power. Just remember
[15:05] how many times you've heard people say
[15:07] AI is a bubble. Uh if it's a bubble,
[15:09] everyone is still saying the same thing.
[15:11] And as I go through with Goldman Sachs
[15:13] earnings commentary and JP Morgan, uh
[15:16] not only are the uh companies that are
[15:18] making AI or producing the chips for the
[15:22] intelligence, you've got the power needs
[15:24] that are happening, you also starting to
[15:25] see the AI adoption growing rapidly. Uh
[15:28] he briefly touched on riffs. He
[15:30] acknowledged that high bandwidth memory
[15:32] shortages could constrain the overall
[15:33] server market, potentially capping how
[15:35] many logic chips it can ship if they
[15:38] can't be packaged with memory. The
[15:40] memory shortage is a massive story and I
[15:42] hear more people fading it. As someone
[15:44] who has been on Micron for a long time
[15:46] now below a hundred and continuing to
[15:49] sit in the exact same position, it's
[15:51] because I see nothing on the horizon.
[15:54] And out of every 10 people I talk to, at
[15:56] least six, if not seven of them either
[16:00] don't believe it or it's too late and
[16:01] they can't get in. Just a summary on
[16:04] everything that's gone on.
[16:07] So the Micron story is still real. Mark
[16:09] Newton, one of the guys I follow, just
[16:12] to look for some interesting commentary.
[16:15] I thought the Lag seven and the software
[16:17] names remaining in consolidation was a
[16:19] big deal. Until we get more
[16:20] stabilization from the big boys, meaning
[16:22] as he's calling them, the lag seven, it
[16:24] might prove tough to show real
[16:26] acceleration on any rally. But make no
[16:27] mistake, with transports, IWM, midcaps
[16:30] at new highs, and equal weight S&Ps,
[16:32] this is exactly what I've talked about,
[16:34] and this is exactly what I've shown so
[16:35] far. This market is stronger than what
[16:37] the cues may imply in the short run.
[16:39] Just many aren't used to seeing
[16:43] these names, transport names, chemical
[16:47] names ripping while they sit with their
[16:49] palunteer waiting. Adapt or die. Could
[16:52] not agree more. This is a different
[16:53] year. And I think you have to get used
[16:55] to this for the foreseeable future. I
[16:57] think this is the official end of the
[17:00] dominance that happened with coding. And
[17:02] that's why I said this has got to be
[17:03] understood by people from a macro basis.
[17:06] Uh and I don't think a lot of people
[17:07] have thought about equating everything
[17:08] including the capital surplus account to
[17:10] coding. Nvidia Reuben, I brought this up
[17:13] last week. This is from a data center
[17:16] thing this week. The shift will reshape
[17:18] how data centers get built this decade.
[17:20] Again, Reuben was a major announcement
[17:22] in the first week or last week in CES.
[17:25] It gets right back in to this which had
[17:28] nothing to do with Reubin. This just had
[17:30] to do with the Nvidia Grock deal and
[17:31] what Nvidia was sending as a signal and
[17:33] what was supposed to happen from the
[17:34] growth in NPUs and edge devices. This
[17:37] accelerated it and that's basically data
[17:41] movement. So, I'm not going to read all
[17:42] this, but if you want to take these,
[17:45] take a snapshot, upload it into your
[17:47] LLM, just go in there and say, "What is
[17:49] the the change from GPUs into data
[17:52] movement for edge devices?" Uh, and the
[17:55] new architecture for Reuben matter for
[17:59] new companies. And I'll basically go
[18:00] through this, but five bullet points.
[18:02] Computer is no longer the dominant cost.
[18:04] Idle GPUs are the new tax.
[18:07] Locality beats raw speed. This is going
[18:09] to be really really important as we kind
[18:11] of get into on premise uh east west
[18:14] traffic becomes critical. This goes with
[18:16] some of the names that are working. The
[18:17] reason I'm bringing this up is not to
[18:20] overwhelm you with something that's a
[18:21] little technical. It's to make sure you
[18:23] realize that we ended the GPU side in
[18:26] 2025. It probably happened midway
[18:29] through the year. That was the end of
[18:31] the GPUs being the only story for AI.
[18:34] Now we're transitioning for every GPU
[18:36] story which is Nvidia. All of those
[18:38] packaging names I showed you did not
[18:40] include Nvidia.
[18:42] You're dealing with a lot of winners now
[18:45] and at the expense of the one winner.
[18:47] It's not that I think Nvidia is going
[18:49] down. I don't think any of those, the
[18:50] lag seven, any of them in general as a
[18:52] group are going to go down. I just think
[18:54] they're going to underperform. And if
[18:55] they finish down 5% and the S&P finishes
[18:58] up 15%, the bottom 60% are going to do
[19:01] what they're doing, which is carry the
[19:02] index and be up over 25% to make the
[19:05] math work. And then the Russell 2000
[19:07] will be up 50 plus percent to make the
[19:09] math work. Um just a simple line here to
[19:12] make sure that you understand the
[19:14] transition from the GPUs to data
[19:16] management. So we also got a deepseek
[19:18] paper out this week where basically
[19:21] they're confirming that a memory
[19:24] augmented architecture for smaller more
[19:26] efficient AI is going to be the story
[19:28] going forward. So Reuben and Deep See
[19:30] confirm that memory is the key to
[19:32] performance per watt. memory becomes the
[19:34] marginal dollar that improves
[19:36] efficiency.
[19:38] Again, these are th this is not old news
[19:41] at this point. And I remind everyone
[19:43] something I've said multiple times. It
[19:45] was the summer of last year that I was
[19:47] hearing from investors that I was
[19:48] visiting that there was going to be a
[19:50] glut in memory. So, you have to think
[19:53] about the fact that that was only six
[19:54] months ago, less than six months ago,
[19:57] and I was hearing that from multiple
[19:58] people because the cellside research was
[20:00] going around that way. So you have
[20:02] Micron that have gone from 150 to 360 as
[20:06] of Friday. But it's not just those
[20:08] names. So again, direct beneficiaries
[20:10] from data movement bottlenecks. Here you
[20:12] go. There's some names. Most of those
[20:13] are on the list. I did
[20:16] SanDisk got a lot of love this week. And
[20:18] the they came out and said this is
[20:22] incremental new nan demand that is not
[20:25] previously in their current forecast. So
[20:28] again, they're admitting that this is
[20:29] news. So what has their stock done?
[20:31] Well, everyone thinks it's a bubble
[20:32] because it's gone from 225 up to 387
[20:35] already this month. But it did it in a
[20:37] basian way. It did it on new news and it
[20:39] readjusted in price and I'm telling you
[20:41] news that was not expected because this
[20:44] is a big change for right now and for
[20:46] what Reubin's going to need. And
[20:47] everyone's going to want Reuben as I
[20:49] showed you the efficiency gains and
[20:50] everyone starts realizing that the
[20:52] bottleneck is power. So the efficiency
[20:54] chips are more important than they've
[20:56] ever been. Ruben shifts a large portion
[20:58] of inference memory from transient DRAM
[21:00] usage to persistent local storage. That
[21:03] one architectural decision explodes nan
[21:06] demand. So again, just names that are
[21:09] fitting in there. Uh another one I spent
[21:11] a lot of time talking to people about
[21:13] this week, Pure Storage, which is a
[21:15] great name. It's done well, but it's
[21:16] also corrected over the course of the
[21:18] last 18 months. Uh more on this on
[21:21] there. City came out and talked about
[21:22] it. Um you've got uh people talking on
[21:25] the DDR side and just going through it.
[21:28] So you've got Micron, Samsung, SKHEX,
[21:30] all of them showing up. So when you look
[21:32] at this chart, you're like, this is a
[21:34] bubble. At the same time, you go look at
[21:37] the earnings growth that's happening.
[21:40] And then you go look at the pees. So
[21:42] again, when I first started pitching
[21:44] this stock, the PE was about six. So now
[21:47] off next year's earnings, it's currently
[21:49] at nine. So when you go through Expost,
[21:52] you'll see the inevitable bears saying
[21:54] this is a normal cycle and eventually
[21:56] you're going to have peak prices because
[21:58] this is a semiconductor. Again, I'm
[22:00] going to say this over and over again
[22:01] for the next five years. And this is not
[22:03] going to be for every part. The DRAM
[22:05] thing may peak next year. I doubt it's
[22:07] going to peak with the multiples here
[22:09] when we're still in a serious supply
[22:10] shortage. And when
[22:14] Taiwan semi-chairman Mark Leu buys 8
[22:17] million of Micron shares near all-time
[22:20] highs, he's on the board of Micron. Uh I
[22:23] don't think this would have flied
[22:24] necessarily in the US the same way, but
[22:26] whatever. Um you've got more stories on
[22:29] Micron. I'm I'm giving you guys just you
[22:31] can go do this research on your own. Uh
[22:34] it anticipated will only be able to meet
[22:36] half to twothirds of the demand from its
[22:37] key customers. It has devoured the
[22:39] global memory supply. We are in this
[22:41] situation for one reason and one reason
[22:43] only. As of a year ago, software was the
[22:47] only game in town when hardware had been
[22:48] through so many down cycles. So many
[22:51] where every time things go higher, they
[22:53] collapse because if memory memory
[22:55] bandwidth got to this situation with
[22:57] phones and the phone costs doubled, it
[22:58] would be an issue. In this case, what's
[23:00] happening is the driving factor for this
[23:02] is is not a big cost, input cost for
[23:05] artificial intelligence, and you're in a
[23:07] race for all of it from places that have
[23:09] massive amounts of money. A memory
[23:12] bottleneck dubbed the memory wall that
[23:13] now limits AI scalability and is
[23:15] expected to drive up prices for both
[23:17] enterprise and consumer devices.
[23:21] AI memory is sold out, causing an
[23:23] unprecedented surge in prices.
[23:25] Micron just broke ground on a $100
[23:28] billion memory manufacturing complex in
[23:30] history. The largest semiconductor
[23:32] facility in the US. They just broke
[23:34] ground. For those of you asking what are
[23:36] they breaking ground on, there's a lot
[23:38] of stuff that's happening this year.
[23:39] Again, belong PMIs. At the same time,
[23:42] SKHEX to invest 13 billion in new plant
[23:44] mount shortage. These are all this week.
[23:46] Everything that I've showed you is this
[23:48] week. So again, supply shortages. What's
[23:51] going on?
[23:54] One analyst estimates the DRAM industry
[23:56] can only support 15 gawatts of AI data
[23:58] center capacity over the next two years
[24:00] without cannibalizing other markets. Yet
[24:02] plans call for 40 to 50 gawatt.
[24:04] Shortages could cap deployments leading
[24:06] to underutilized delayed projects.
[24:10] So let's go now to the hopper.
[24:12] Three-year-old
[24:14] chips as Matt Seagull put out. Um and
[24:19] again you guys can watch this in
[24:20] Bloomberg. uh rental prices where Jim
[24:23] Chanos has been focusing on just are up
[24:26] 14% from the November lows. Even though
[24:28] these chips are uh
[24:31] Reuben is 90% uh be or significantly
[24:35] better. We're talking uh 20 times better
[24:37] than Hopper in terms of uh power, but
[24:40] you're saving efficiency in terms of the
[24:42] power needs by 90% relative to to
[24:44] Hopper. Um Intel almost sold out.
[24:50] Apple and Qualcomm fret over strained
[24:52] supplies of Japan's glass cloth. Again,
[24:54] I'm not going to show copper this week.
[24:56] I'm not going to show uh silver. You get
[24:59] the point. Scarcity, guys. Invest in
[25:01] scarcity. Learn how to invest during a
[25:03] commodity super cycle, which is what
[25:05] we're in. There will be sharp falls
[25:07] during the year because when something
[25:09] like Micron goes up as fast as it does,
[25:10] eventually it'll hit a level. And we may
[25:12] have hit it on Friday. We'll hit a
[25:14] level. It will fall back $50 to $70 uh
[25:17] just in a correction. And I assume
[25:18] that's going to happen with the Mag 7. I
[25:21] assume that's going to happen with fears
[25:22] over the buildout or fears that their
[25:24] revenues haven't come in. So, let's now
[25:26] jump to software, the gift that keeps on
[25:28] giving for now. Uh, every time there's
[25:31] positive news, CRM stock pews, is it an
[25:33] inverse AI ETF? Now, this gets back into
[25:36] my theme on coding. Gavin Baker this
[25:39] week coowork which was which came out
[25:41] which I'll talk about more but remember
[25:44] when Microsoft co-pilot uh the stock
[25:48] ripped higher because they did the math
[25:50] and people the analysts started saying
[25:52] how much Microsoft would get in terms of
[25:54] revenues from co-pilot that was a long
[25:56] time ago I I believe it was 2023
[26:00] uh that's a problem to have Microsoft
[26:03] co-pilot as a story in the summer or not
[26:05] even the summer of 2023 so almost 3
[26:08] years ago and Claude Co-work comes out
[26:10] and it was built in 10 days without a
[26:12] human being involved except for the
[26:13] idea. The idea came from the users. If
[26:16] you haven't read the story on it, Claude
[26:18] code opus 4.5 claude co-work. We are in
[26:22] the point of recursive self-learning and
[26:24] building stuff. So when you think about
[26:26] Microsoft Copilot, I want you to think
[26:28] Claude Co-work. I'm using it already. I
[26:31] have refused to use Microsoft Copilot
[26:34] even though it has improved because I
[26:36] find it to be useless compared to the
[26:38] LLMs. I'm a power user on every LLM. I'm
[26:41] on it all day long on my phone and my
[26:44] laptop no matter where I am all the
[26:46] time. You cannot pay me to use Microsoft
[26:49] Copilot
[26:50] instead of the LLMs. Even if I was in my
[26:54] I I think this is a major problem for
[26:56] software companies including Microsoft
[26:58] unless something changes because these
[27:00] things can be built quickly. I installed
[27:02] Cloork on my personal laptop. Since then
[27:05] it has boom.
[27:09] Morgan Stanley says anthropics code plus
[27:11] co-work is dominating investor chat and
[27:13] adding pressure on software. Um,
[27:17] again, I don't, you know, Elon Musk has
[27:20] been all over the place saying this is
[27:21] the end of software. Eric Schmidt has
[27:23] said this. Uh, Chamath has said this.
[27:27] Uh, there's a I mean, I'm going to go
[27:28] through Sequoia Partners. I I there's a
[27:31] lot going on. This is not just me, but I
[27:33] think ma most investors who are, let's
[27:36] say, mean reverting, of which a lot of
[27:38] X's, they tend to be the bears on AI.
[27:41] This is a AI bubble trade now to try and
[27:43] pick the bottom in software. I highly
[27:45] highly doubt that that's going to end up
[27:47] working out. There will be a couple that
[27:49] can adapt, but as I get through this,
[27:51] just remember incumbents have a really
[27:54] hard time changing the culture and
[27:56] changing the business and AI agents are
[27:58] moving way too fast. And I and the costs
[28:01] are going up for the models. I just
[28:02] don't know how these guys are going to
[28:04] pull it off. We'll see. Uh but for the
[28:06] time being, I'll give you some trades
[28:07] towards the end. Uh Ralph was released,
[28:10] too. It's an autonomous AI coding loop.
[28:13] you can go find out about it. These are
[28:14] the stories that make them the the the
[28:16] most important. Joe Weisenthal, who I've
[28:18] listened to a lot of the OddLots uh
[28:20] podcast this year, uh over the last
[28:23] year, I would say generally been bearish
[28:25] or skeptical of of the AI trade and and
[28:29] had a lot of bubble people on the data
[28:31] center thing, overbuild. That's gen been
[28:34] the general thesis. Well, he posted this
[28:36] this year. The AI productivity potential
[28:38] has never been more obvious in caps. In
[28:40] today's OddLot newsletter, I wrote part
[28:42] one of my takeaways from the weekend
[28:44] playing with claude code and building.
[28:46] You should really give it a read um and
[28:48] just see what someone who's not a coder
[28:50] who sat down and did the same thing I
[28:52] did with my turbulence model, which is
[28:54] just try it. You'd be shocked at how
[28:56] easy it is to build things. So later in
[28:59] the week, three million lines written
[29:01] over a week of continuous AI agent time
[29:03] with chatbt5.2.
[29:05] This is from Greg Rockman at OpenAI.
[29:11] Extending the amount of time it can run
[29:13] on its own from the cloud one. Again,
[29:14] they just keep leaprogging. If one
[29:16] releases one, then the next one releases
[29:17] a model that's better than that one.
[29:19] This will continue. And all of these
[29:21] models, I will reiterate, are not on
[29:24] Blackwell. Blackwell will come out,
[29:27] especially for Gro 5 in Q1. The
[29:30] acceleration in these whatever you're
[29:32] seeing right now in AI agents will be
[29:34] significantly stronger in the next three
[29:36] months and in the next six months and in
[29:38] the next nine months big businesses
[29:41] especially big companies where you see
[29:43] their buildings across the country like
[29:44] Salesforce.com or Microsoft very very
[29:47] difficult to navigate getting your
[29:49] entire organization to change when both
[29:51] of them Microsoft Copilot and Agent
[29:54] Force for Salesforce this has been part
[29:56] of their theme. Maybe they're able to
[29:58] turn this around. But as I go through
[30:00] this and I show you some of the papers
[30:01] I'm writing, I just find it hard to
[30:03] believe that you're going to fight
[30:04] against a tide that is just massive.
[30:06] Aaron Levy, another person to respect
[30:09] from uh Box. The capability overhang
[30:12] right now in AI is pretty massive. Most
[30:13] of the world thinks of AI as chat bots
[30:16] and that will answer a question on
[30:17] demand, but yet do not yet do real work
[30:20] for them. this concept of asking a
[30:22] question. If you do that, if you
[30:24] basically do Google search with it,
[30:26] you're not using it yet. It is not made
[30:28] for that. It is not made for giving you
[30:30] an answer. It is meant for you giving it
[30:32] context and getting a probability on the
[30:36] question that you're answering. If
[30:38] you're an if you're looking for an
[30:39] answer, you're more likely to get a
[30:41] hallucination. If you're looking for
[30:43] something to include on your thought
[30:44] process, because I've heard multiple
[30:46] people say, "I asked to do this. It gave
[30:47] me five different answers." Don't ask it
[30:49] a question for an answer. Don't ask it,
[30:51] "My knee hurts on the inside. What is
[30:53] that?" "Oh, you have tendonitis. You
[30:56] have this." It's trying to please you.
[30:58] If you go through and say it and you
[31:00] say, "What is this most likely?" Giving
[31:01] it the context, it'll give you a higher
[31:04] probability answer based on it. You have
[31:06] to learn how to use the context. That's
[31:08] the human element. We have become stupid
[31:11] due to school and due to Google search.
[31:14] We've become siloed because of
[31:16] corporations. You must think broader and
[31:20] you must think like a polymath and ask
[31:21] it the right questions. AWG from
[31:24] Moonshots is now posting awesome exposts
[31:28] every day. He only has 26,000 followers.
[31:31] I will single-handedly try to drive that
[31:33] up just because I think all of you guys
[31:34] will will be grateful in the end. He
[31:37] every day puts the news that's going
[31:39] out. I go in there because I'm looking
[31:42] for the best places to get it. His stuff
[31:44] is genius. And recursive
[31:47] self-improvement has graduated from a
[31:48] safety paper to a shipping manifest.
[31:50] Anthropic has confirmed that Claude Code
[31:52] wrote the entire new Claude Cowwork
[31:55] desktop in just one and a half weeks.
[31:57] Think about that. No humans. McKenzie
[32:00] 25,000 AI agents is now counting their
[32:03] workforce, including the digital
[32:05] employees. Get used to it. Uh, Sundar
[32:08] Pachai talking about AI agents and
[32:11] announcing the universe commerce
[32:13] protocol AI agents for shopping. Goldman
[32:16] Sachs on their call
[32:19] about AI. With AI, you will be able to
[32:21] break that paradox and scale output with
[32:23] the option of adding people or not.
[32:26] Again, getting into the efficiency and
[32:28] productivity gains. We just announced
[32:30] the launch of one Goldman Sachs 3.0,
[32:32] know our new operating model propelled
[32:34] by AI starting with six work streams we
[32:36] identified as ripe for disruption. It
[32:39] creates an ability for us in the next 5
[32:40] years to accelerate the pace of
[32:42] operating leverage creating capacity to
[32:44] grow revenues. This on premise release
[32:48] especially by the financial companies is
[32:50] going to be the gift that keeps on
[32:52] giving. Remember the paper I wrote on
[32:55] artificial intelligence is a faster and
[32:58] better version of what QE did for profit
[33:00] for profits and for stocks back during
[33:03] the prior decade. This will be a golden
[33:05] age for profit margins for companies
[33:07] that are bloated like the banks like the
[33:11] insurance companies. Uh JP Morgan
[33:15] massive investment to create autonomous
[33:18] systems that prevent headcount from
[33:19] growing with revenue. JP Morgan was
[33:21] aggressively moving beyond simple chat
[33:22] chat bots with agents. These agents have
[33:25] goal-oriented re behavior. They confirm
[33:28] their LLM suite is already handling
[33:30] autonomous trade settlements.
[33:32] Again, very very small in terms of
[33:35] what's happening there. So far, I talked
[33:37] to enough people at JP Morgan who still
[33:40] complain about their ability to use it.
[33:42] But for the workflow AI agent side,
[33:44] that's there. What is going to take a
[33:46] lot more time is breaking down the silos
[33:48] of these organizations. I will write a
[33:50] paper on how you make money on breaking
[33:52] down these silos. One of the ways is
[33:54] pure storage but I want Palunteer and
[33:56] data bricks and um snowflake all of
[34:00] these companies will be a part of it but
[34:02] I was involved as managing director at
[34:05] Morgan Stanley and I know the uh silos
[34:07] and the amount of just the intense
[34:10] amount of of software in every single
[34:13] silo. It's another reason why the
[34:14] software comp the enterprise software
[34:16] companies are screwed. When these silos
[34:18] come down, the bloat comes down. When
[34:19] the bloat comes down, the dollar spent
[34:21] outside on software will go down. That's
[34:23] what you're going to be replacing.
[34:24] People and software budgets. They guided
[34:27] for 105 billion in expenses driven
[34:29] largely by AI infrastructure and talent.
[34:31] So again, I t T T T T T T T T T T T T T
[34:32] T T T T T T S T T T T T T T T T T T T T
[34:33] T T T T T T T T T T T T T T T T T T T T
[34:33] T T T T T T T Semi
[34:36] spending uh more in terms of uh
[34:38] infrastructure and capex. The banks
[34:41] doing the same thing. Here's that
[34:42] article again for those signed up for
[34:45] 22V here. when the payw wall goes up,
[34:47] just let me know. You guys can get this
[34:49] as well. This goes through how
[34:50] artificial intelligence will flood the
[34:52] economy with just an enormous amount of
[34:55] money seen in the form of productivity
[34:57] and we're already seeing it with
[34:58] creating uh 4% GDP, no inflation, and no
[35:02] job hiring. Uh again, I've written these
[35:06] two papers. Vibe coding is the chat GPT
[35:08] moment for code. Why the great rerating
[35:11] began? When I wrote this, I got multiple
[35:13] people re reaching out saying, "This has
[35:15] already happened." Again, trying to pick
[35:16] a bottom in this stuff. I can't see a
[35:19] bottom. I I I I just can't. I think if
[35:21] you're a one-trick pony in software or
[35:23] you've got a large institution, uh, can
[35:26] Microsoft come out of this? I Yeah,
[35:28] they're a big organization. They've got
[35:30] our hands in everything and they
[35:32] certainly can, but I I think it's I
[35:35] think it's up for debate on all these.
[35:36] And I think it's very difficult. The
[35:38] revenues will speak. Uh but I think more
[35:40] importantly Claude Co-work just came out
[35:42] this week and it just highlights that
[35:44] we're only at the very beginning stages
[35:45] where the code is now good enough for it
[35:47] can doing the entire app building which
[35:49] takes things from taking months if not
[35:52] years into days and that's the critical
[35:55] part that took 10 days to build and as
[35:57] Gavin Baker said Microsoft Copilot has
[36:00] had three years to try and do what
[36:02] co-work did in 10 days think about that
[36:04] uh the Opus 4.5 inflection point so this
[36:07] is not something I haven't been on top
[36:09] of and next week I will release a paper
[36:11] titled why buying cheap software is the
[36:13] new AI bubble trade comparing it to
[36:16] people who faded the ompic side. So we
[36:20] were fighting obesity and the staples
[36:24] processed food companies suffered. I
[36:26] think software companies are basically
[36:28] had their osmpic moment and that ompic
[36:30] moment is basically co-work. Uh again,
[36:33] if you want to go listen to Sonia Yang,
[36:35] partner at Sequoa, speak about the
[36:37] software side. You don't have to take my
[36:38] word uh for this. You can go listen to
[36:40] someone, but this is the highlights of
[36:42] it in terms of what's going on. Um
[36:45] again, they're a VC place, so they're
[36:47] focused on startups, which are
[36:48] competing, but every single person, I
[36:51] mean, Mark Andre, all of them are saying
[36:53] the exact same story in terms of the the
[36:56] difficulties it'll be to challenge uh
[36:58] startup businesses and software and AI
[37:00] native places. Nate Jones did one on it
[37:03] this week in terms of uh Claude Co-work.
[37:05] He even gave a demo on it. I'm gonna
[37:07] highlight him a lot. Uh I did want to
[37:10] just highlight I took the transcript on
[37:11] this and I want to show you one of the
[37:13] things that I'm training people. So I do
[37:14] have a hedge fund analyst skill that
[37:16] takes any information. It doesn't have
[37:18] to be from a podcast. So it doesn't have
[37:20] to be a transcript like Nate Jones. It
[37:22] can be something that I read on X. It
[37:24] can be a report that I got from Goldman
[37:26] Sachs or Morgan Stanley sent to me from
[37:28] someone else or Susuana. Uh it could be
[37:30] a conference that I'm attending and I
[37:32] hear something and when I come back I've
[37:33] taken seven bullet points. I take those
[37:35] seven bullet points and I basically
[37:37] paste them in here and I say run the
[37:39] hedge fund analyst skill on that.
[37:42] This is what it does. It's creating a
[37:45] it's a skill I've created which
[37:47] basically creates a PM research memo
[37:49] output with all of these different
[37:51] components. Again, highlighting the risk
[37:53] side, the reward side, the second and
[37:56] third tier effects, looking for the
[37:58] supply chains, every single part of it
[38:00] to keep players, all the stories
[38:01] involved. It does that in a matter of
[38:04] seconds.
[38:07] This is a matter of seconds. So, the
[38:10] output ends up being 10word pages.
[38:14] I take that and then I go through my
[38:16] equity hedge thing that I showed you
[38:18] before into step two. Uh again, I'm
[38:21] showing this as a live demo now for
[38:23] people, I will try to finish the videos
[38:24] on it so they can go up on the payw wall
[38:26] uh to help uh younger people and anyone
[38:28] interested in learning this. If you're
[38:30] on the institutional side uh and you
[38:33] want to just watch me do it and kind of
[38:34] ask questions if your organization's
[38:36] looking for help on this on how you can
[38:38] utilize it, I will guarantee you that
[38:40] one hour with me, you will learn things
[38:42] that you haven't seen before on it and
[38:44] it will speed up the process of getting
[38:46] going. This is something that you will
[38:48] never get a button to push. AI agents
[38:50] are not made to solve the investment
[38:52] process. They are made to solve the part
[38:55] that is the grunt work. The human
[38:57] element is where context comes in.
[39:00] Context is important. At the end of the
[39:01] year when the last football game is
[39:03] playing, the context that matters is
[39:06] does the game matter and is the starting
[39:08] quarterback playing? Obviously, well,
[39:10] the starting quarterback playing would
[39:11] immediately change the odds and I think
[39:12] AI would be pretty good at that. But
[39:14] knowing that the game doesn't matter or
[39:16] that a lot of the players are only going
[39:18] to play for the first quarter,
[39:21] it's not going to have the context. So,
[39:23] think sports, think everything. That's
[39:25] what we do for a living is the people
[39:26] that are great in this business, they
[39:28] put into context, positioning,
[39:29] sentiment. Uh, is this a story I've
[39:32] heard before the relative nature of the
[39:34] sentiment? All of those are context.
[39:36] That was the moonshot for this week. Uh,
[39:38] I'm not going to go through all of them,
[39:39] but they did say this is an inflection
[39:40] year with rapid acceleration and it
[39:43] feels like the year of singularity.
[39:45] AGI's here is effectively being a story
[39:47] that I'm hearing repeatedly.
[39:50] Elon Musk, um, they spent a lot of time
[39:53] talking about the interview they did
[39:54] with them. Some great stuff, including
[39:56] the humanoid side, which I'll get into
[39:57] here. uh they went through a lot on opus
[40:00] 4.5 and acknowledged the fact like I
[40:03] said this is happening before Blackwell
[40:06] that this is AGI. They also went into a
[40:09] topic that I've spoken about here with
[40:11] Pomp uh and really since 2013 when I
[40:16] said it is now officially not a relevant
[40:18] statistic. It will continue to screw
[40:19] with macro people. It will continue to
[40:21] screw with economists and strategists.
[40:23] GDP is an irrelevant statistic. It's
[40:25] never been less relevant than it is
[40:27] today because of how fast exponential
[40:29] innovation. The reason 2013 is because
[40:31] that's when exponential innovation hit
[40:33] kind of the elbow point in my world for
[40:36] disrupting GDP. They talk about this in
[40:38] there and for the first time on
[40:40] Moonshots and really the first time I've
[40:41] heard anyone go through it. They said it
[40:43] can't work. It it it's not built for
[40:46] this and it's especially not built for
[40:48] the fact that AI is structurally
[40:49] deflationary, the most deflationary
[40:52] force we will ever feel. um output
[40:54] decouples from lab decouples from labor.
[40:56] I've talked about that relentlessly. The
[40:59] better AI works, the worse GDP becomes
[41:01] as a proxy for progress. Could not agree
[41:03] more. They talk about it. This was
[41:05] really interesting because when you have
[41:06] someone like Dave Blondon who has been a
[41:08] naysayer on the humanoid deployment and
[41:10] saying repeatedly that we just don't
[41:12] have the scale to get this stuff out. So
[41:14] maybe by 2035
[41:17] we could get to and I forget what the
[41:19] number is. Let's just say a million uh
[41:21] humanoids. He just upgraded everything
[41:23] significantly. And the reason was after
[41:25] they were done or while they were at the
[41:28] Gigafactory doing the interview with
[41:30] Elon, hopefully I'm doing a good job
[41:31] with his name today, guys. Uh
[41:34] he basically highlighted that he saw
[41:36] robots building robots. So he's getting
[41:38] what the Ford CEO saw in the dark
[41:40] factories in China, which is, hey, we're
[41:43] already at the point where robots are
[41:45] building robots. And that's what allows
[41:48] it to speed up. Because if you have that
[41:50] capability where you don't need humans
[41:51] to be there, then you can work 24/7.
[41:55] Everything changes 365 days a year. And
[41:58] you get back to what I talked about last
[41:59] week, which is we as humans work 34.2
[42:02] hours a week according to the most
[42:03] recent payroll numbers. Let's take that
[42:05] number down to about 20. When you strip
[42:06] out the cigarette break, the lunch
[42:08] break, the talking, the chatting, the
[42:10] flirting, the whatever else goes in an
[42:11] office that people do, you end up with a
[42:13] situation of 20 hours. Robots are going
[42:16] to work 247. Let's get to 168 hours a
[42:19] day. You're dealing with about 8 and a
[42:22] half more productivity just by that.
[42:25] This is going to be a dramatic
[42:27] beginning. This is just the beginning.
[42:30] Uh they talked about data centers and
[42:32] how they're transitioning and I agree
[42:33] they're becoming factories. I will
[42:35] highlight some of that. Uh and you have
[42:37] to think about AI factories not
[42:39] buildings anymore with servers. The
[42:41] change is happening now. So best proxy
[42:44] for you guys this year. I've talked
[42:46] about it, but we finally got a massive
[42:47] breakout which says to me the scarcity
[42:49] trade is on. Long scarcity short
[42:51] abundance proxy long Chevron short
[42:55] Salesforce.com
[42:56] both around the same market cap. And if
[43:00] you think this is crowded, oh, you're
[43:03] wrong. Here's the trade over the last 15
[43:06] years. Think of this as oh my gosh, this
[43:10] is when the capital account surplus
[43:12] really started to take off. The US
[43:14] outperformed the rest of the world. The
[43:16] smartphone came into existence and we
[43:18] had a software dominance with no
[43:21] hardware. Oil collapse. China stopped
[43:24] growing and now we're getting the
[43:25] renaissance. This is the beginning of
[43:27] the unwind of this. And maybe it only
[43:29] goes to here. I don't really care. The
[43:32] main point is from 72 to even here,
[43:35] which might happen this year. That's a
[43:36] big man. long chevron short sales
[43:40] salesforce.com as a simple proxy for you
[43:43] for scarcity over abundance. Now Chad
[43:46] GPT health came out. This is a major
[43:48] deal. Not a minor deal. This is a major
[43:50] deal. We're almost up to 20% of all
[43:54] spending nominal GDP is now healthcare.
[43:59] I you cannot even think of the
[44:04] importance of this for everyone out
[44:06] there. And I'll go through some of the
[44:08] reasons. As someone who is very focused
[44:10] on anti-aging and very focused on this
[44:12] and already does effectively what this
[44:14] thing is, which is I already upload my
[44:16] blood work, which I get done every 3
[44:18] months. I already upload any information
[44:20] that I have, all my historical data that
[44:23] goes back to 1993 I believe. all there
[44:27] so that I can have conversations and
[44:29] then when anything happens where I have
[44:30] a knee pain or anything I update it
[44:32] every day almost like a journal on
[44:34] what's going on so that it literally is
[44:36] my doctor on anything that's happening I
[44:39] use it already it introduced chat GPT
[44:41] health a consumerf facing tool aimed at
[44:43] helping patients it signals AI and
[44:45] healthcare is moving from
[44:46] experimentation to infrastructure it's
[44:49] not to replace clinicians but it is
[44:51] absolutely to make it easier for the
[44:53] consumer and easier for the healthcare
[44:55] care uh provider uh and for the
[44:58] insurance company. It has it says 230
[45:02] million users ask about health every
[45:04] week. This is a huge deal guys. I meet
[45:07] dozens of AI health startups every week
[45:09] and you can tell this is a big deal.
[45:10] Most of them will become redundant once
[45:13] this gets adopted. It will be
[45:14] disruptive. But for the bigger health
[45:16] companies which are bloated, it's going
[45:19] to help for sure. Now they not only
[45:21] launched chat GBD health for the
[45:22] consumer they also opened open AAI for
[45:24] health care. The platform is rolling out
[45:26] at major health systems including advent
[45:29] memorial salone ketering and cedars
[45:32] sinai.
[45:35] It aims to streamline clinical workflows
[45:37] reduce administrative burden you get the
[45:39] picture. Openai in the same week
[45:42] acquires torch startup
[45:45] focused on the health care sector.
[45:48] It's to enhance its capabilities within
[45:50] health care and its health platform.
[45:54] Anthropic seeing this immediately jumps
[45:57] into it launches Claude for healthcare
[45:59] challenging GPT health. The AI show
[46:03] talked about Chad GBT health and AGI co
[46:07] uh AGI uh code sorry Claude code. The
[46:10] claude code moment signals the tipping
[46:12] point for AGI level usefulness. ADI
[46:15] doesn't need AGI to transform work. This
[46:18] is really critical, guys. If you're
[46:19] waiting for a push button for this,
[46:21] again, you have to get some training on
[46:23] this. You have to do something. Chat GPT
[46:25] Health launches with deep medical
[46:27] integration. If you want to hear them
[46:28] talk about these, go listen to it. If
[46:31] you remember a couple months ago when I
[46:33] highlighted pharmaceuticals and said Eli
[46:35] Liy, which will be one of my favorite,
[46:37] one of my top five names for this year,
[46:39] especially on a risk adjusted basis, uh
[46:42] meaning V adjusted. uh Nvidia and Eli
[46:45] Liy announced
[46:48] a co-inovation lab. Now they announced a
[46:52] one a building a data factory
[46:56] back in November. May have been October,
[46:58] but I think it was November. Now they're
[47:00] doing a co-inovation lab. So one of them
[47:04] is think of it as their own LLM, meaning
[47:06] Eli Liy's got all this data in all these
[47:08] silos for all these different diseases.
[47:10] that wants to use all of the data they
[47:13] have, all of the experiments they've
[47:15] done, all of the tests they've done, all
[47:17] of the failures, all the successes, and
[47:19] they want to use LLMs and AI to
[47:22] basically be able to compare all that
[47:24] data, but the data is all in silos. It's
[47:26] in different structured scenarios,
[47:28] probably in different software, who
[47:29] knows? Well, they need first the
[47:31] factory. This is going to be about the
[47:33] innovation lab where they're going to
[47:34] work together to reinvent them, to
[47:36] discover, to develop, and to
[47:37] manufacture.
[47:39] So, you're starting to see the
[47:41] acceleration that's happening. And I
[47:43] think this is going on again. And the
[47:45] reason I love the pharmaceutical names
[47:47] because AI drug discovery is right here,
[47:49] right now. And there's going to be major
[47:51] announcements, I'm sure, this year on
[47:53] this based on what I've seen from Google
[47:55] Deep Mine and Silico and Eli Lilly and
[47:58] other places as well. So, this is going
[48:00] through what happened. Yeah, it was
[48:01] October, November. Um, one was a data
[48:04] center. This one is a co-inovation lab.
[48:07] Shopify's CEO made his own MRI using uh
[48:12] Claude code.
[48:16] China is involved in this too. So you're
[48:18] going to have advancements around the
[48:19] globe. They develop AI platform that
[48:21] achieves millionfold increase in drug
[48:23] screening speeds.
[48:25] A new AI tool could dramatically speed
[48:27] up the discovery of life-saving
[48:28] medicines.
[48:30] This is why this is so important. Here's
[48:33] pharma relative to the S&P
[48:36] 1500 all the way back to 2001.
[48:39] Do you think this is uh we had a big
[48:41] move up? I highlighted it was the
[48:42] largest move I think since around 2003.
[48:45] Uh sorry, since around this period. Uh
[48:48] this is the beginning of a long-term
[48:49] move. Pay attention to pharma. But it's
[48:51] not just pharma, guys. Look what's
[48:53] happened to healthcare. This is the S&P
[48:55] 1500 healthc care sector relative to the
[48:57] S&P 1500 since chat GPT launch. Again,
[49:02] this is a relative thing. We don't want
[49:04] to be long any of this. So, biotech
[49:06] underperformed. Now, you're getting a
[49:07] bottom here. I would suspect that
[49:09] healthcare is going to outperform a lot
[49:11] of the sectors. It's one of my favorite
[49:12] ones. I finally launched the first uh or
[49:16] my HRV substack. So the second substack
[49:19] I have for those of you interested in
[49:21] HRV, interested in learning about it, I
[49:25] do believe it is a metric for aging. I
[49:28] am going to spend every week for many
[49:30] years just highlighting all the
[49:31] different things I did. At the end of
[49:34] all of that will be a prompt that is
[49:36] there for people to learn on their own.
[49:39] Uh this is the prompt. So this is a
[49:41] normal prompt for me. This is the prompt
[49:43] that is attached at the end of what will
[49:45] be the first publication next week. You
[49:48] can copy it from there and then you can
[49:49] go paste it. It will take you on a
[49:51] journey. The first week is about the
[49:52] importance of the microbiome for this.
[49:54] If you guys watch Limitless with Chris
[49:56] Heelms Hemsworth and Peter Aia, uh
[49:59] basically that whole thing was about
[50:01] HRV. They just barely said HRV, but they
[50:03] talked about the importance of
[50:04] breathing. They talked about the
[50:06] importance of strength. They tal all of
[50:08] these different things that all equate
[50:10] back to HRV. The problem is Chris
[50:13] thought he was healthy. He's Thor. He
[50:16] wasn't. They didn't show his HRV at the
[50:18] beginning, but I'm telling you as
[50:19] someone who's gone through this. This is
[50:21] my Aura ring data. I don't have two
[50:23] Well, I basically had the same thing. I
[50:24] told you I've gone from the 20s. This is
[50:26] where I am so far this year. Continuing
[50:28] what I was I'd shown you that I'd been
[50:30] up near 70 even this year. Uh I'm still
[50:33] at 70. Uh but I've been rising higher.
[50:36] So, I'm going to go through in that how
[50:38] I used AI to get there. It'll teach you
[50:40] about AI. It'll also, if you're
[50:41] interested in health, go through it. You
[50:43] can find it at my Substack. And then
[50:45] finally, well, not finally, but we're at
[50:47] the AI trending topics. I just wanted to
[50:49] highlight this is what I went through
[50:50] this week. I ran this AI agents, Claude
[50:54] Code, AI and healthcare, efficient
[50:56] models. I highlighted these because
[50:58] basically guys, when I do these videos,
[51:00] I'm on top of everything that's
[51:01] trending. I'm on top of everything
[51:03] that's going on. The only way you can do
[51:04] that is to be listening to podcast Beon
[51:06] X. You take that information, you put it
[51:08] into my format, and then you can
[51:10] generate alpha. As Gavin Baker said,
[51:13] investing is a search for hidden truths.
[51:15] You generate alpha by finding a truth
[51:16] before others see it. There is tons of
[51:18] information to find right now real time
[51:20] for analysts and PMs and traders that
[51:22] are out there. All you got to do is
[51:24] this. Find some new alpha that's
[51:27] interesting that you haven't heard
[51:28] before. Do what I did in terms of the
[51:30] skills thing. You come up with an
[51:32] output. You put it into here. Once you
[51:33] get that, you generate reports that go
[51:35] into a deep research report. Once you
[51:37] get that, all the deep research reports
[51:39] come in here and you search for ideas
[51:41] and then you put them through a
[51:42] screener. That is the whole point of
[51:44] what I built. It's what I'm going to
[51:45] show. This is what each of them does. If
[51:47] you want to blow it up and read it, go
[51:49] ahead. Uh that's why I'm spending my
[51:52] time on this. But it's so much fun for
[51:54] me to teach people how to use AI, and
[51:55] I've been having success with it, uh
[51:57] that anyone that wants to get involved,
[51:58] again, call 22V, wait for the payw wall,
[52:01] whatever the case is. You can do this
[52:02] with fantasy football, you can do it
[52:04] with finding a job. the whole theme of
[52:06] what I said with Annie Duke, which is AI
[52:08] will help you make better decisions. You
[52:10] just have to get the context right and
[52:11] it'll lead to a better decision. Uh,
[52:14] finally on crypto,
[52:18] the banks are not giving up yet in
[52:20] fighting crypto. So, Coinbase CEO Brian
[52:22] Armstrong said they're trying to kill
[52:23] the competition.
[52:25] The bill is still being held up. Despite
[52:28] that, Bitcoin did have a breakout week.
[52:30] We broke out of the channel. So,
[52:32] basically, we broke this trend line. We
[52:34] broke the 50-day moving average, 20-day
[52:36] moving average. We're still below the
[52:37] 200 day moving average. That is the
[52:39] final line in the sand. And as I said
[52:40] last week, this line here is right
[52:42] around there, 100,000.
[52:46] So, I think we're going to have trouble
[52:47] getting out of this. We get the clarity
[52:49] act, we break above, it's there, but
[52:50] we're building a bottom here. In my
[52:52] opinion, I've been heavily involved in
[52:55] terms of the last two weeks. We did do
[52:57] three-day closes above uh 92, which was
[53:02] my trigger point for me in terms of
[53:03] thinking that everything would go. I
[53:05] love the way the chart looks. It's kind
[53:06] of reverse head and shoulders. The
[53:08] retest makes sense. We'll see what
[53:10] happens. If we fail and go back into the
[53:11] range, it's just going to take a little
[53:12] bit longer, but we do have the MACD
[53:15] weekly ready to cross. So if this is a
[53:18] bottom and we get across down here, I
[53:20] fully expect not only new highs, but
[53:21] then confirmation as we get up towards
[53:24] the new highs of it being the best
[53:26] performing asset, it has already
[53:27] outperformed the MAG 7 this year by 10%.
[53:31] So already this year in terms of
[53:32] Bitcoin's move very quickly, if it got
[53:35] up to the all-time highs, just so people
[53:37] hear, sh it will have outperformed the
[53:39] S&P this year and last year. So we'll
[53:41] see what happens. Ethereum
[53:44] also ready for a MACD crossing. All
[53:46] right, guys. That's it for this week.
[53:48] Uh, hope you enjoyed it. Keep reaching
[53:50] out. Hit the subscribe button. It helps
[53:52] me out. Trust me, more than you know.
[53:54] Keeps me going on this. Go to my HRV
[53:56] Substack and the payw wall will be up.
[53:59] Hopefully I connected on the Elon side.
[54:02] And then I hope to see a bunch of you. I
[54:04] know I will down in Miami with Raul and
[54:07] the gang. So, have a great uh extended
[54:10] weekend, the holiday. and I'll see you
[54:12] next