Is AI Trade DEAD? 🤖 $835B Capex vs $2.4T Backlog: The Real Math💥
https://www.youtube.com/watch?v=cX2q_NngqKI
Language: en | ASR: large-v3-turbo @ 0a363e9161cbc7ed1431c9597a8ceaf0c4f78fcf

[00:00:05.580 --> 00:00:11.860] Speaker 1: hello everybody yesterday was a special day shout out to silicon phoenix first day waking
[00:00:11.860 --> 00:00:18.680] Speaker 1: up in a world of flying cars that demo was spectacular but it's also causing more
[00:00:18.680 --> 00:00:25.060] Speaker 1: polarity across the two different sides of the aisle today we're going to talk about
[00:00:25.060 --> 00:00:31.720] Speaker 1: is the ai trade dead or alive there's a lot of signs been a very rough week we're going to
[00:00:31.720 --> 00:00:37.900] Speaker 1: break out all the numbers all the charts are going to go deep into why it could be dead
[00:00:37.900 --> 00:00:43.700] Speaker 1: and why it could be alive and then a very detailed quant verdict at the very end because
[00:00:43.700 --> 00:00:50.300] Speaker 1: that's what i do here if it's dead or alive and uh thank you to the moderators in cheshaw
[00:00:50.300 --> 00:00:56.820] Speaker 1: and d cypher sliver dd tesla k8 let's go this one is a little bit tricky
[00:00:56.820 --> 00:01:03.220] Speaker 1: but boom or bust and that's the button we press whether to determine yes or no
[00:01:03.220 --> 00:01:09.700] Speaker 1: so actually it'd be fun to do a little poll is it dead or alive so if somebody can do that
[00:01:09.700 --> 00:01:14.280] Speaker 1: let me know and this is not financial advice and shout out to hodlnader glad to be back on
[00:01:14.280 --> 00:01:18.640] Speaker 1: patreon missed you guys yeah some people leave for the bear never leave for the bear the
[00:01:18.640 --> 00:01:22.540] Speaker 1: bear is where all the goodness is that's where the billions are made anyhow let's get into
[00:01:22.540 --> 00:01:29.180] Speaker 1: sentiment and signals fear and greed is at 19 extreme fear in the stock market
[00:01:29.180 --> 00:01:33.880] Speaker 1: not talking about the crypto market today because that's not relevant but we look at the
[00:01:33.880 --> 00:01:40.880] Speaker 1: week we've had by the way this snapshot i grabbed about half an hour ago and we have gone
[00:01:40.880 --> 00:01:46.820] Speaker 1: into the power hour so things bounced off of these levels but you can see it was a brutal
[00:01:46.820 --> 00:01:53.480] Speaker 1: week across the board there were no survivors nobody got away unless you were in
[00:01:53.480 --> 00:02:00.300] Speaker 1: oil or interest rates you probably were okay. And there's some rotation also into some of
[00:02:00.300 --> 00:02:06.220] Speaker 1: these stodgier sectors. But Micron down 14%. Of course, the Kospi South Korea got hit hard.
[00:02:06.720 --> 00:02:08.800] Speaker 1: Marvel nearly 17%. Alab nearly
[00:02:10.140 --> 00:02:17.120] Speaker 1: 16%. ASML 7. Tesla 7. SpaceX 11 .6%. Palantir, at last it
[00:02:17.120 --> 00:02:23.060] Speaker 1: fell down. But remember, it's been going up non -stop since 106. Hit 210. Did a little bit
[00:02:23.060 --> 00:02:28.940] Speaker 1: of a mean reversion. Cerebrus, very bad, down 22%. SK Hynix down nearly
[00:02:28.940 --> 00:02:35.380] Speaker 1: 19%. Solana down 7%. Michael Shaggy down 12%. On rumors, they're going to be booted from
[00:02:35.380 --> 00:02:42.340] Speaker 1: the MSCI or MCSI World Index Fund. But this is just a week,
[00:02:42.680 --> 00:02:47.980] Speaker 1: okay? If we look at what's happened year to date, the numbers are wild. Micron over 200%.
[00:02:47.980 --> 00:02:54.300] Speaker 1: AMD 150%. Marvel 170%. Alabs 72%. started strong fell down hard
[00:02:54.300 --> 00:03:01.180] Speaker 1: grew again asml 54 and that's after many of these names lost 20
[00:03:01.600 --> 00:03:08.260] Speaker 1: this week this is still very very positive for the year so always have that zoom out
[00:03:08.260 --> 00:03:13.940] Speaker 1: perspective now let's start with the tricky part i always like to start with the bad news
[00:03:13.940 --> 00:03:17.600] Speaker 1: first get it kind of out of the way this is the case for the bust i'm going to go through
[00:03:17.600 --> 00:03:21.820] Speaker 1: this pretty fast and pull up all the research I did and all the stuff I monitor and track
[00:03:21.820 --> 00:03:26.380] Speaker 1: every single day because I'm heavily invested in AI. So I need to know if it's going to
[00:03:26.380 --> 00:03:33.320] Speaker 1: survive or not. First of all, Bank of America, they come out and they think, ooh, we're
[00:03:33.320 --> 00:03:38.100] Speaker 1: seeing signs, signs of boom and bust that we've seen before,
[00:03:39.880 --> 00:03:46.440] Speaker 1: historic cycles, et cetera. All the people that lived through the dot -com bust, they still
[00:03:46.440 --> 00:03:50.340] Speaker 1: have the scar tissue from that. and they're hoping this will be the same thing that's why
[00:03:50.340 --> 00:03:56.560] Speaker 1: they believe it's the same all over again and they believe people are not investing they
[00:03:56.560 --> 00:04:03.060] Speaker 1: are cheerleading and that's why the banks are saying uh no things are still slowing down
[00:04:03.060 --> 00:04:08.880] Speaker 1: things are looking bad things look ugly this is classic capital cycle risk thank you boo
[00:04:08.880 --> 00:04:15.120] Speaker 1: mcduck happy friday everybody as well and thank you to mods in the chat but in addition
[00:04:15.120 --> 00:04:22.100] Speaker 1: data centers this is someone took me by surprise uh it's from tech radar data centers are
[00:04:22.100 --> 00:04:27.420] Speaker 1: more unpopular in the united states than the iran war think about that for a second
[00:04:27.420 --> 00:04:33.180] Speaker 1: because there are groups that are spreading fud about data centers oh they use all the
[00:04:33.180 --> 00:04:39.740] Speaker 1: water oh they make noise oh they do this you remember these are like it's like like trying
[00:04:39.740 --> 00:04:44.880] Speaker 1: to listen to here or factory. It's all FUD. Anyway, I'm not going to go there. But what is
[00:04:44.880 --> 00:04:50.800] Speaker 1: crazy is nearly 70 % of the US population are against AI data centers because they were
[00:04:50.800 --> 00:04:57.700] Speaker 1: told to be against them. Can they stop progress? We shall see. Also, the other
[00:04:57.700 --> 00:05:04.000] Speaker 1: problem that we have is these things take years to build. Hooking up the electric that's
[00:05:04.000 --> 00:05:09.180] Speaker 1: required, whether it's before or after the meter is extremely difficult building them is
[00:05:09.180 --> 00:05:16.000] Speaker 1: extremely difficult getting permits takes years so all of the power and electric and the
[00:05:16.000 --> 00:05:22.380] Speaker 1: facilities and the wiring etc not just sourcing the semiconductors or the memory which is
[00:05:22.380 --> 00:05:27.960] Speaker 1: also a headache really really difficult so building takes years there's a lot of big
[00:05:27.960 --> 00:05:32.120] Speaker 1: talkers saying oh we're going to have 10 gigawatts 20 gigawatts 50 gigawatts to compute
[00:05:32.120 --> 00:05:36.160] Speaker 1: next year and no they're not there's only one player that can pull that off We'll get to
[00:05:36.160 --> 00:05:42.280] Speaker 1: that later. Now, the other problem is the bottleneck is extremely physical. Again, this is
[00:05:42.280 --> 00:05:46.820] Speaker 1: not the chips, but it's all the other stuff. It's the skill trades. It's the connectivity,
[00:05:47.180 --> 00:05:53.500] Speaker 1: the substrates, the data centers, the land, the power. All of this takes a lot of time.
[00:05:53.740 --> 00:06:00.000] Speaker 1: Now, there is a huge demand. They want to try and deploy another 43
[00:06:00.000 --> 00:06:05.460] Speaker 1: gigawatts of new AI compute next year. But that's going to be very difficult, many believe.
[00:06:05.780 --> 00:06:12.020] Speaker 1: all the restrictions open ai and anthropic both want to add five gigawatts more than five
[00:06:12.020 --> 00:06:17.460] Speaker 1: gigawatts each is 10 gigawatts in total 10 gigawatts is a lot we'll share some of the
[00:06:17.460 --> 00:06:23.700] Speaker 1: details around exactly what that means in a minute also we have the credit the trade credit
[00:06:23.700 --> 00:06:29.380] Speaker 1: market for ai is getting completely smashed this week if you are a neo scaler
[00:06:29.380 --> 00:06:34.980] Speaker 1: you know one of the neo clouds you're getting absolutely brutalized i have been talking for
[00:06:34.980 --> 00:06:39.600] Speaker 1: a long time about things like oracle being potentially in hot water core weave for the last
[00:06:39.600 --> 00:06:45.160] Speaker 1: 18 months the financials don't make sense to me now the market is painting with a broad
[00:06:45.160 --> 00:06:51.560] Speaker 1: brush and they think everybody is going to get hit by this so the credit markets are
[00:06:51.560 --> 00:06:56.460] Speaker 1: flinching which means the credit dries up which means it's very difficult to find money to
[00:06:56.460 --> 00:07:02.740] Speaker 1: fund these things very difficult as well also credit is getting much more expensive. We
[00:07:02.740 --> 00:07:08.440] Speaker 1: have the Fed raised 25 basis points. 10 -year treasury is rising over 100 basis points,
[00:07:08.580 --> 00:07:15.520] Speaker 1: close to 6%. And because AI is soaking up so much capital, that actually also
[00:07:15.520 --> 00:07:21.500] Speaker 1: raised the discount rate. So it makes it harder to raise money even for other companies and
[00:07:21.500 --> 00:07:28.260] Speaker 1: also for the hyperscalers. And the credit default swap rates are at record wide spreads as
[00:07:28.260 --> 00:07:34.980] Speaker 1: well. You can see some of the top players here. Oracle, really high core weave as well. All
[00:07:34.980 --> 00:07:39.100] Speaker 1: the other players are on there too. So it's getting expensive to fund and finance these
[00:07:39.100 --> 00:07:43.320] Speaker 1: things. In addition, we just found out as well, Anthropic is weighing delaying their IPO
[00:07:43.320 --> 00:07:49.300] Speaker 1: beyond November. They were supposed to IPO in October. They said, let's wait till November.
[00:07:49.600 --> 00:07:54.960] Speaker 1: Now they're going, maybe not November. And also they're toying with the idea of a valuation
[00:07:54.960 --> 00:08:01.120] Speaker 1: rumored down to 1 .5 trillion from 2 trillion. that may reflect some of the recent numbers
[00:08:01.120 --> 00:08:07.160] Speaker 1: too so it's like somebody's put a whole bunch of sand in the gears of this whole ai boom
[00:08:07.160 --> 00:08:13.600] Speaker 1: big time the bill is huge this is the bill from just the big four or something you know
[00:08:13.600 --> 00:08:20.580] Speaker 1: amazon 220 billion of capex up from 200 billion google up to 205 billion up
[00:08:20.580 --> 00:08:27.280] Speaker 1: from 195. Microsoft, 175 billion. Meta, another 145 billion. Big
[00:08:27.280 --> 00:08:34.060] Speaker 1: four, like tons of money, 735 to 750 billion. And next year, even higher. Add in
[00:08:34.060 --> 00:08:38.520] Speaker 1: Oracle, another 100 billion. Takes to 835. And it's still going up.
[00:08:40.160 --> 00:08:43.760] Speaker 1: How are we going to pay for this? How are they going to pay for it? Can they even pay for
[00:08:43.760 --> 00:08:50.640] Speaker 1: it so that is all of the negative stuff around the case against the ai trade could it be
[00:08:50.640 --> 00:08:56.760] Speaker 1: cooked now let's talk about the case for the ai boom this is again the same approach
[00:08:57.520 --> 00:09:01.440] Speaker 1: looking at the same different type of data i'm trying to put a picture together in my
[00:09:01.440 --> 00:09:07.660] Speaker 1: little mind first of all uh joe and the channel yesterday we were talking about how i've
[00:09:07.660 --> 00:09:11.620] Speaker 1: always said for years now have you ever heard anybody saying they have enough intelligence
[00:09:11.620 --> 00:09:12.420] Speaker 1: the answer is no
[00:09:13.040 --> 00:09:13.520] UNKNOWN: So
[00:09:13.520 --> 00:09:16.960] Speaker 1: as you get more intelligence, you want more intelligence. And that's the whole Jevons
[00:09:16.960 --> 00:09:17.460] Speaker 1: paradox.
[00:09:18.980 --> 00:09:24.600] Speaker 1: And as tokens get cheaper, as intelligence gets cheaper, it'll drive more demand. And the
[00:09:24.600 --> 00:09:30.020] Speaker 1: penetration is still so low as well right now. And training has gotten so much better. So I
[00:09:30.020 --> 00:09:34.800] Speaker 1: used to cost $20 million to train the first GPT -1 model. Can now be done for $100 today.
[00:09:35.080 --> 00:09:41.880] Speaker 1: Used to cost $20 million. Now it's $100. So that's a million X. and usage is supposed to go
[00:09:41.880 --> 00:09:46.820] Speaker 1: up a million x as well over the same decade and we have that math problem that had 10 000
[00:09:46.820 --> 00:09:52.220] Speaker 1: agents solve it four million of compute that humans couldn't solve it so math is completely
[00:09:52.220 --> 00:09:59.060] Speaker 1: cooked as well so again everything is changing very fast tons of demand and demand
[00:09:59.060 --> 00:10:04.400] Speaker 1: comes in thirds a lot of people think oh all the money all all the demand is just coming
[00:10:04.400 --> 00:10:10.180] Speaker 1: from OpenAI, Anthropic, et cetera? No, it's not. So this is from Jensen. He talks about his
[00:10:10.180 --> 00:10:16.180] Speaker 1: market segment for NVIDIA GPUs. One third is classic computing. One third is across five to
[00:10:16.180 --> 00:10:19.520] Speaker 1: six frontier labs, which includes names like SpaceX and Google,
[00:10:20.740 --> 00:10:26.620] Speaker 1: Amazon, OpenAI, Anthropic, et cetera, Meta. And then you have another one third. It's
[00:10:26.620 --> 00:10:32.220] Speaker 1: beautifully spread actually across startups and NeoClouds. And there are thousands of them.
[00:10:32.220 --> 00:10:36.700] Speaker 1: and they all use open source models. And then we have a new group coming in, the Sovereign
[00:10:36.700 --> 00:10:42.760] Speaker 1: AI group, driven by players like Palantir that want people to build their own AI using open
[00:10:42.760 --> 00:10:48.440] Speaker 1: source models as well. So again, the other key point here is a lot of people say, well, if
[00:10:48.440 --> 00:10:53.040] Speaker 1: the market cap of open AI and Anthropic is so high, they need to do a trillion dollars in
[00:10:53.040 --> 00:10:57.120] Speaker 1: revenue. The answer is no. Because there's rumors that some of this business that they do
[00:10:57.120 --> 00:11:02.960] Speaker 1: is very high margin. We'll get to that in a second. And here, this one is inference. I
[00:11:02.960 --> 00:11:06.660] Speaker 1: always talk about in the future, inference will be everything, but inference is also where
[00:11:06.660 --> 00:11:13.260] Speaker 1: the money is. This shows you this little excerpt I copied here from Semi
[00:11:13.260 --> 00:11:19.460] Speaker 1: Analytics, I think it was. It talks about the gains per GPU are incredible in terms of
[00:11:19.460 --> 00:11:25.440] Speaker 1: efficiency and pricing has gone down and Blender prices per million tokens have dropped in
[00:11:25.440 --> 00:11:32.020] Speaker 1: the past few months. and despite that the margins have gone from for inference for the
[00:11:32.020 --> 00:11:38.800] Speaker 1: frontier models have gone from 40 to over 70 just in a few months despite prices falling so
[00:11:38.800 --> 00:11:43.100] Speaker 1: it's extremely lucrative and in the future when everything is learned and all the models
[00:11:43.100 --> 00:11:49.440] Speaker 1: are kind of stable and mature 99 of all compute will be inference 99 and
[00:11:49.440 --> 00:11:54.720] Speaker 1: if that's lucrative and it's high margin and it's great for everybody. Also, the revenue is
[00:11:54.720 --> 00:12:01.140] Speaker 1: beginning to show up. People thought, oh, the Mag 7 are spending far too much on AI and
[00:12:01.140 --> 00:12:05.760] Speaker 1: compute. But look at the revenue growth for these companies. Azure up 43%. Google Cloud up
[00:12:05.760 --> 00:12:12.740] Speaker 1: 82 % year to date. Amazon up 37%. That is extremely big cloud growth. So, the
[00:12:12.740 --> 00:12:18.040] Speaker 1: investment is paying off for them as well. And by the way, there's no compute available. If
[00:12:18.040 --> 00:12:23.220] Speaker 1: you're looking for compute out to 2027 or 2028, it's all sold out. There's none available.
[00:12:23.460 --> 00:12:28.940] Speaker 1: More has to be built. So, and the other thing, I touched on this briefly yesterday, are we
[00:12:28.940 --> 00:12:34.020] Speaker 1: early or late? This ties into the Jevons paradox as well, but this is the double
[00:12:34.020 --> 00:12:40.820] Speaker 1: exponential. First of all, AI needs more compute per task and the reasoning models and
[00:12:40.820 --> 00:12:46.660] Speaker 1: as agents explode, the AI agents explode, they will need more options. and of course more
[00:12:46.660 --> 00:12:53.560] Speaker 1: compute and more demand for more tasks and everything will go up to the tune of a billion x
[00:12:53.560 --> 00:13:00.160] Speaker 1: from here so there is no slowing down by any stretch and these ai agents these bots are
[00:13:00.160 --> 00:13:06.780] Speaker 1: going to drive things parabolic also the money is still coming okay it's three to four
[00:13:06.780 --> 00:13:12.620] Speaker 1: trillion it's the ai infra spend you want to follow the money because that will all trickle
[00:13:12.620 --> 00:13:17.800] Speaker 1: down to your investments, as long as you're positioned correctly. But again, AI infra is
[00:13:17.800 --> 00:13:21.760] Speaker 1: expected to reach three to four trillion dollars per year by 2030. You got all the
[00:13:21.760 --> 00:13:25.600] Speaker 1: Fudsters, all the Bank of America's, all the newspapers are loving this. They love to grave
[00:13:25.600 --> 00:13:31.300] Speaker 1: dance and say, oh, AI boom is over. The money's dried up. It's gone. No, it's not. What you
[00:13:31.300 --> 00:13:37.120] Speaker 1: also have to understand is this is the only growth in the economy. Okay. And you look at
[00:13:37.120 --> 00:13:42.120] Speaker 1: the people close to the Cantillo machine, close to the money printer what are they going to
[00:13:42.120 --> 00:13:47.900] Speaker 1: do think of the apollos the black rocks etc they get all the money and they have to deploy
[00:13:47.900 --> 00:13:53.120] Speaker 1: it where they're going to deploy it in cellular carriers
[00:13:54.760 --> 00:14:01.060] Speaker 1: food makers nike retailers no it's going to go straight into ai because that's where the
[00:14:01.060 --> 00:14:06.980] Speaker 1: return is that's where the growth is and that's what people miss so the money is tons and
[00:14:06.980 --> 00:14:12.140] Speaker 1: tons of money to be there and it's going to continue to grow also this is kind of a good
[00:14:12.140 --> 00:14:16.680] Speaker 1: news bad news piece so it's a hard to figure out where to put it but i put it in the good
[00:14:16.680 --> 00:14:21.840] Speaker 1: news area because this could be construed as a headwind for nvidia but nvidia has given us
[00:14:21.840 --> 00:14:26.820] Speaker 1: notice that they are expecting to double growth next year they have again infinite demand
[00:14:26.820 --> 00:14:32.940] Speaker 1: for their gpus in fact it's rumored that the vera rubens 40 percent of the supply has been
[00:14:32.940 --> 00:14:38.440] Speaker 1: locked up by elon musk alone one player. Remember, five players can buy everything. What
[00:14:38.440 --> 00:14:42.000] Speaker 1: about the thousands of other companies that are trying to get their hands on a Vera Rubin?
[00:14:42.080 --> 00:14:46.220] Speaker 1: They can't. And apparently on the black market, they cost $5 million each when normally
[00:14:46.220 --> 00:14:50.220] Speaker 1: they used to retail $2 million. So these things are like hen's teeth, very hard to track.
[00:14:50.460 --> 00:14:57.240] Speaker 1: But the good news about this is this impacts other parts of our IA13. The demand for Google
[00:14:57.240 --> 00:15:03.480] Speaker 1: TPUs are off the hook. The expected growth rates, As you can see here, go from TPU sales of
[00:15:03.480 --> 00:15:09.840] Speaker 1: $2 billion in 2026 to $18 billion in 2027 to $43 billion in
[00:15:09.840 --> 00:15:15.500] Speaker 1: 2028, $52 billion in 2029, $60 billion in 2030. Yeah, it's tapering off a little bit in
[00:15:15.500 --> 00:15:22.440] Speaker 1: 2029, 2030. But the point is huge demand. And who makes those TPUs? Well, they're the AMDs,
[00:15:22.540 --> 00:15:27.940] Speaker 1: the Broadcoms, the Marvels, et cetera. So, again, there's not enough chips. and this huge
[00:15:27.940 --> 00:15:32.980] Speaker 1: demand for chips and this huge amount of capital chasing all those chips. So don't fret
[00:15:32.980 --> 00:15:37.900] Speaker 1: that stuff as well. It's very good for everybody. And there's no slowdown on the capex
[00:15:37.900 --> 00:15:42.740] Speaker 1: spending. We haven't heard any players saying, you know what? Yeah, we said we're going to
[00:15:42.740 --> 00:15:47.460] Speaker 1: spend 200 billion and we're just not going to do it at all. That has not happened. In fact,
[00:15:47.640 --> 00:15:51.480] Speaker 1: what they have said is they're going to spend even more next year, a lot more, and they
[00:15:51.480 --> 00:15:57.040] Speaker 1: have to. In addition, you know, although Anthropa, you could argue, have tapered off a
[00:15:57.040 --> 00:16:01.800] Speaker 1: little bit. When you look at the growth quarter to quarter, they did 4 .7 billion in Q1, 11
[00:16:01.800 --> 00:16:08.280] Speaker 1: .6 billion in Q2, Q3, 16 .8 billion. I don't know if they're going to get higher in
[00:16:08.280 --> 00:16:14.160] Speaker 1: Q4. We shall see. There's a lot more competition now than there was before. But the point
[00:16:14.160 --> 00:16:20.940] Speaker 1: is annualized. They hope to get to 110 billion, which means a little over 26, 27
[00:16:20.940 --> 00:16:27.360] Speaker 1: billion a quarter in Q4. We'll see. And OpenAI is looking at $50 billion
[00:16:27.360 --> 00:16:31.320] Speaker 1: annualized, but it could get a lot higher. Probably going to be more like $70 billion
[00:16:31.320 --> 00:16:35.960] Speaker 1: annualized by year -end, given the growth rate I'm seeing as well. So these things, big
[00:16:35.960 --> 00:16:41.840] Speaker 1: two, are not slowing down. And in addition, SpaceX have told us they want to get to 2 .7
[00:16:41.840 --> 00:16:48.000] Speaker 1: gigawatts of compute by year -end. That would be $100 billion ARR. And remember, they are
[00:16:48.000 --> 00:16:52.700] Speaker 1: hoping to use 60 % of their compute for themselves, for Optimus and other things, and then
[00:16:52.700 --> 00:16:56.760] Speaker 1: 40 % lease it out to other players. and the other players have infinite demand for what
[00:16:56.760 --> 00:17:03.140] Speaker 1: they want and they can get the return on investment for every gigawatt they turn on in less
[00:17:03.140 --> 00:17:10.040] Speaker 1: than a year per the cfo so will they get to 2 .7 gigawatts by year end elon thinks so this
[00:17:10.040 --> 00:17:14.380] Speaker 1: is his map his plan and shout out to michael dell
[00:17:15.760 --> 00:17:22.140] Speaker 1: and elon musk and jensen wang and everybody else how much will they get to by next year
[00:17:22.140 --> 00:17:28.060] Speaker 1: Elon Musk have more than 10 gigawatts by year -end. So 2 .7 by this year -end, 10 gigawatts
[00:17:28.060 --> 00:17:34.200] Speaker 1: by year -end 2027, which means adding nearly two gigawatts,
[00:17:34.480 --> 00:17:40.900] Speaker 1: a quarter almost, a little less to ramp that. But that also, to pull that off,
[00:17:41.120 --> 00:17:48.040] Speaker 1: they probably need $100 billion in GPUs as well. So again, the plans are there.
[00:17:48.300 --> 00:17:50.500] Speaker 1: The stories are huge, but there's more.
[00:17:51.920 --> 00:17:58.200] Speaker 1: as i mentioned that before openai ananthropic they plan to add 10 to 12 gigawatts of
[00:17:58.200 --> 00:18:04.200] Speaker 1: compute next year that's between them five to six each gigawatts okay for that to happen
[00:18:04.200 --> 00:18:08.100] Speaker 1: they need a player that can do it who is that player i don't know but i think elon web
[00:18:08.100 --> 00:18:11.680] Speaker 1: services spacex will be providing a lot of that data because the others don't have it
[00:18:11.680 --> 00:18:16.540] Speaker 1: google doesn't have extra compute amazon doesn't have extra compute they're all gonna have
[00:18:16.540 --> 00:18:20.380] Speaker 1: to go somewhere microsoft is running out as well. And that's why they're building their
[00:18:20.380 --> 00:18:26.320] Speaker 1: Maya chip. But OpenA and Anthropic as well, combined revenue is expected, this is a crazy
[00:18:26.320 --> 00:18:32.460] Speaker 1: number, to go from about $100 billion this year to over $500 billion in the next 12 months.
[00:18:32.820 --> 00:18:36.800] Speaker 1: I don't see that happening. That semi -analysis came up with that number. I do not see them
[00:18:36.800 --> 00:18:41.240] Speaker 1: making half a trillion dollars in revenue in the next 12 months. But if they do, can you
[00:18:41.240 --> 00:18:47.680] Speaker 1: imagine how much money would be floating around? Even if they get to 200 billion at a 10x
[00:18:47.680 --> 00:18:54.540] Speaker 1: multiple 20x multiple that's two to four trillion dollars in value there so then
[00:18:54.540 --> 00:19:00.060] Speaker 1: they can be very easily justified and again they'll say they will get there using only
[00:19:00.060 --> 00:19:05.120] Speaker 1: about 30 of that compute for inference that's the high margin stuff 70 goes to training in
[00:19:05.120 --> 00:19:10.640] Speaker 1: fact it's rumored that open ai is taking a huge amount of their training compute and
[00:19:10.640 --> 00:19:15.060] Speaker 1: putting it towards version six and seven that they're keeping for themselves. I think
[00:19:15.060 --> 00:19:21.020] Speaker 1: they're going to try to develop drugs and become drug makers, cure things like cancer. You
[00:19:21.020 --> 00:19:25.860] Speaker 1: know, we are the beneficiaries of all of this investment. Our lives get better. So this is
[00:19:25.860 --> 00:19:32.320] Speaker 1: what people don't understand. They're anti -AI, but it's actually very good for us. So tell
[00:19:32.320 --> 00:19:38.420] Speaker 1: your friends, convince them, especially when it comes to health and medicine
[00:19:38.420 --> 00:19:45.100] Speaker 1: anyhow another simple little chart is the amount of token demand is real and rising
[00:19:45.100 --> 00:19:50.820] Speaker 1: can anybody see any any blips in this chart yeah some slight little ones maybe from some
[00:19:50.820 --> 00:19:56.780] Speaker 1: weird anomalies or how it was measured but the trend is up and that's up to like
[00:19:56.780 --> 00:20:03.440] Speaker 1: i think it's 800 billion tokens or maybe 800 i don't know
[00:20:03.440 --> 00:20:08.420] Speaker 1: what that measurement is but it's monstrous and it keeps on demanding then let's get to
[00:20:08.420 --> 00:20:13.460] Speaker 1: memory this is another big piece and i know this week the memory stocks got clobbered
[00:20:13.460 --> 00:20:19.840] Speaker 1: whether you're in samsung sk hynix or micron got completely smashed and then michael dell
[00:20:19.840 --> 00:20:25.600] Speaker 1: says he expects the memory crunch to get worse there's already no memory you can't get it
[00:20:25.600 --> 00:20:32.580] Speaker 1: you cannot get it no memory anywhere they're completely sold out for next year and
[00:20:32.580 --> 00:20:37.040] Speaker 1: elon said for sure in 2027 but if china ramps lp5 and lp6 as fast as they've ramped other
[00:20:37.040 --> 00:20:42.940] Speaker 1: manufactured products 2028 will be a different story but this is the difference high
[00:20:42.940 --> 00:20:47.220] Speaker 1: bandwidth memory is a tougher hill to climb and that's not what the chinese have although a
[00:20:47.220 --> 00:20:52.520] Speaker 1: lot of people believe they do grand duke thank you for coming it's not true hbm will still
[00:20:52.520 --> 00:20:59.380] Speaker 1: be extremely hard to get and this was rumored to hit hit the wires yesterday that google is
[00:20:59.380 --> 00:21:05.680] Speaker 1: trying to secure high bandwidth memory not for 2028 or 2027 but out to
[00:21:05.680 --> 00:21:12.660] Speaker 1: 2035 because there's none everybody is trying to lock up supply nearly a
[00:21:12.660 --> 00:21:18.760] Speaker 1: decade out think about that that's google going to micron and saying please sir can i have
[00:21:18.760 --> 00:21:23.660] Speaker 1: some more if anybody here is old enough to remember that movie let me know i thought it was
[00:21:23.660 --> 00:21:29.780] Speaker 1: a good image that i hobbled together yesterday thanks to ai let's get to the conclusion a
[00:21:29.780 --> 00:21:35.340] Speaker 1: little bit of numbers here i'm going to now measure up in a fair way the pros cons and see
[00:21:35.340 --> 00:21:41.640] Speaker 1: if it's dead or alive first of all paying for cash is it over can they afford it are we
[00:21:41.640 --> 00:21:48.020] Speaker 1: running out of cash is there a trade credit risk for ai the answer is no for over half of
[00:21:48.020 --> 00:21:53.660] Speaker 1: them they still have enough free cash flow to pay for all of this capex i think um one
[00:21:53.660 --> 00:21:59.960] Speaker 1: player was slightly negative for one of the quarters amazon was slightly negative but the
[00:21:59.960 --> 00:22:06.500] Speaker 1: rest are all free cash flow positive microsoft alphabet meta making tons of money and
[00:22:06.500 --> 00:22:12.300] Speaker 1: they're using that to fund their capex then the verdict is yes for the big four with
[00:22:12.300 --> 00:22:18.180] Speaker 1: conditions because of the growth in their revenue and the cash flow that's spinning off and
[00:22:18.180 --> 00:22:22.740] Speaker 1: the amount that they're investing and the depreciation timelines for GPUs. A lot of people
[00:22:22.740 --> 00:22:26.960] Speaker 1: just think GPUs would burn out in one or two years. No, they last five to six years. And
[00:22:26.960 --> 00:22:30.400] Speaker 1: there's companies that have H100s that are running now for seven years, no problems. And
[00:22:30.400 --> 00:22:34.940] Speaker 1: they're still renting them out at top dollar. So the credit will stay open for these big
[00:22:34.940 --> 00:22:41.940] Speaker 1: players. I'm not sure about Oracle and CoreWeave, et cetera. But when it comes to
[00:22:41.940 --> 00:22:47.440] Speaker 1: Microsoft and Google and Meta and NVIDIA and all these players, not a problem. They'll have
[00:22:47.440 --> 00:22:52.360] Speaker 1: no problem securing funding, having free cash flow to pay for it all. And I'm going to turn
[00:22:52.360 --> 00:22:57.280] Speaker 1: off my camera for this one. This is kind of my quick and dirty little summary. If you add
[00:22:57.280 --> 00:23:03.700] Speaker 1: up the math on the left on the bus side, you got about 835 billion of CapEx in 2026.
[00:23:04.380 --> 00:23:07.300] Speaker 1: Bond issuance has gone nutty, up 79%.
[00:23:08.560 --> 00:23:12.260] Speaker 1: Alphabet's Q2 cash flow did turn negative just for a minute, but it's going to be back
[00:23:12.260 --> 00:23:17.380] Speaker 1: positive next year. You got the CDS rates for CoreWeave and Oracle, totally in the toilet
[00:23:17.380 --> 00:23:24.260] Speaker 1: and yeah power is short by 10 to 15 gigawatts that and there's
[00:23:24.260 --> 00:23:29.980] Speaker 1: some circular financing in there as well that's all the nasty side but on the boom case
[00:23:29.980 --> 00:23:35.180] Speaker 1: side you got 2 .4 trillion of backing you got money coming in from all the big wall street
[00:23:35.180 --> 00:23:41.600] Speaker 1: players cloud is growing at 82 percent anthropic revenue is growing by 3 .6x labs are
[00:23:41.600 --> 00:23:48.540] Speaker 1: targeting 5x growth next year and this three to four trillion planned ai infrastructure
[00:23:48.540 --> 00:23:54.980] Speaker 1: investment per year in 2030 it's not slowing down and the backlog
[00:23:54.980 --> 00:24:01.680] Speaker 1: covers the 2026 capex by nearly 3x so when i look at the
[00:24:01.680 --> 00:24:08.360] Speaker 1: simple numbers in an objective manner the good far exceeds the bad are we going to go
[00:24:08.360 --> 00:24:14.780] Speaker 1: bust no are we going to go boom yes will there be casualties along the way yes that's
[00:24:14.780 --> 00:24:21.140] Speaker 1: always the case you know when you when you're running at 100 miles an hour breaking eggs of
[00:24:21.140 --> 00:24:27.480] Speaker 1: course there will be casualties let's finish everybody with me do you're so basic net net
[00:24:27.480 --> 00:24:32.080] Speaker 1: the air trade is not over my friends not at all but there are some signs of pain out there
[00:24:32.080 --> 00:24:38.940] Speaker 1: in the marketplace we had uh this is the fun little come on do something mean guy poking
[00:24:38.940 --> 00:24:45.500] Speaker 1: a stick this micron is the box uh micron blew out earnings forward guidance is insane
[00:24:45.500 --> 00:24:51.540] Speaker 1: and the stock didn't move and i think a lot of people felt a bit hurt by that one but
[00:24:51.540 --> 00:24:57.000] Speaker 1: remember the year -to -date map of the heat map of the returns 220 from micron even after
[00:24:57.000 --> 00:25:02.020] Speaker 1: the money gave back this week and last week so we just have to get through the storm
[00:25:02.020 --> 00:25:07.200] Speaker 1: September, my fear was September was good which is normally not therefore October will be
[00:25:07.200 --> 00:25:12.980] Speaker 1: bad and that's exactly what's panned out so with that my friends don't lose faith keep on
[00:25:12.980 --> 00:25:19.480] Speaker 1: stacking if you have not enough AI exposure things are on sale right now
[00:25:19.480 --> 00:25:24.480] Speaker 1: and we are here we got a little power a nice little recovery too out there which is good
[00:25:24.480 --> 00:25:28.640] Speaker 1: and things are even up a little bit in after markets I hope you all learned something new
[00:25:28.640 --> 00:25:34.120] Speaker 1: today once again thank you to the mods in the chat thank you to boo mcduck dbf and grand
[00:25:34.120 --> 00:25:41.080] Speaker 1: duke and of course cypher sniper jacob mavasu there too uh tnd tesla sean d and k8 thanks
[00:25:41.080 --> 00:25:47.180] Speaker 1: all for coming looks like it's sticking up a big friday see you all soon bye