YT YouTube2Text

YouTube2Text → Transcripts → Gavin Baker on Anthropic's S-1, GPU Prices, and SpaceX

Gavin Baker on Anthropic's S-1, GPU Prices, and SpaceX — Transcript

by TBPN · 7,675 words · 1,310 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Next guest, is Gavin Baker from the
0:02Tradies Management, the managing partner
0:04and CEO.
0:04>> Hey,
0:05heard of him before. He's been on the
0:07show before.
0:08>> The Baconator.
0:08>> We're going to bring him into the TBP
0:10and Ultra Dome and catch up with him.
0:12How you doing Gavin? Good to see you.
0:14>> I'm awesome, man. Happiness is a choice.
0:16I'm only ever fantastic, awesome,
0:19superb. Phenomenal.
0:20>> I I mean I I completely agree. I love
0:22that happiness is a choice, but also it
0:24feels like this is a week to be extra
0:26happy. Like the Anthropic S1 went out
0:29and it seemed like a beat. It seemed
0:30like like if you were worried, oh the
0:33economics are not working secretly
0:36behind the scenes and there's going to
0:37be some bombshell in there.
0:39The haters have been kind of quiet. Ed
0:40Zitron hasn't posted. Like things are,
0:43you know, it seems like things are on
0:44track. But what is your interpretation
0:46been this week?
0:48>> I do think the Anthropic S1 was pretty
0:50clearly an upside surprise. Um, the fact
0:52that they've only burned 10 billion
0:54dollars. Yeah. The um,
0:57profitability, the cash generation, the
0:59unit economics, they're all very
1:01encouraging. Um,
1:03the w-
1:04You know, I put it on X, but I just
1:06cannot believe
1:08that whatever it is, half of the Fortune
1:10500 has adopted Anthropic and it's 0.1%
1:15of their expenses. Yeah. It's just a
1:18wild stat.
1:19>> Yeah.
1:20The The funny thing is that
1:22I read that I read that as the uh
1:24>> No, we can still hear you.
1:25>> No, no, no, yeah, we can hear you.
1:27I I read that 50%
1:28>> No.
1:29>> What?
1:30>> Yeah, yeah.
1:30>> Oh, just as like like okay, yeah, that's
1:32incre- incredibly impressive for a
1:33company in just a few years to get the
1:35half the Fortune 500 or the S&P 500 on
1:38as customers, but I was kind of like
1:40what are the other half of the S&P 500
1:42doing? I felt like every Fortune 500
1:44would work or S&P 500 would be
1:46multi-product using everything, but
1:48there's a lot of companies that just
1:49haven't really gotten through the
1:51adoption hiccups of rolling out AI
1:55broadly. Maybe some of them are all in
1:57on Google or Open AI, but by and large I
1:59think some of them are just still
2:01figuring out how AI can be deployed in
2:03their organizations. But is that your
2:05read on it that we're still sort of
2:06early on the diffusion story in
2:08enterprise?
2:09>> Yeah, very early. I mean if anything
2:11maybe earlier than I thought we were.
2:13>> Yeah.
2:15>> And it's I mean it's just astonishing.
2:17Whatever the number is, 67 billion
2:19annualized if you annualized the
2:21reported third quarter. The
2:23[clears throat]
2:24you know, whatever the ARR is right now.
2:26The but to be that big
2:30with such low penetration into the
2:32Fortune 500
2:33>> Yeah.
2:34>> is
2:35>> crazy.
2:35>> It is pretty extraordinary.
2:37>> Yeah, totally.
2:38>> Yeah, how do you how have you processed
2:40the like some of the criticism right now
2:43of the AI trade is that you have a very
2:45small number of companies driving the
2:46majority of the spend. I think if you
2:49can look historically and say that that
2:50same dynamic probably played out with
2:52like social media ads where like a very
2:54small number of companies were driving a
2:56massive amount of the spend. And so
2:58you're going to kind of see power laws
3:00in a lot of places. My read on it is
3:02that you have
3:04like there are companies based in San
3:06Francisco that are just a thousand times
3:09more obsessed with AI
3:11than the than the the say let's say like
3:13the long tail of public company
3:15employees. And so of course they're
3:17going to spend dramatically more. And
3:18it's possible that
3:20that long tail never actually cares
3:22enough, but they sort of fade to
3:24relative obscurity. And you have the
3:26disruptors that are saying like yeah,
3:28this technology is real, it works. It
3:30it's not perfect yet. It's not like a
3:32drop in co-worker that will immediately
3:34you give it a dollar and it'll give you
3:35$10 back. But it's very capable. And so
3:38if you obsess over it. And so that's
3:41been sort of my overall read. Like I
3:43don't think it's necessarily
3:45Of course you'd want a more diversified
3:47revenue base if you could.
3:49But
3:50in general it feels very natural that
3:52all this spend is coming from a very
3:54small group of people that like realize
3:56that something fundamental has changed
3:58and they're trying to take every
3:59advantage of it while the rest of the
4:00world is asleep.
4:02>> I might
4:04respectfully disagree a little in this
4:06one. There are only 250 accounts
4:08spending more than $10 million.
4:11So actually, you know, somebody in the
4:12New York Times said that, you know,
4:14maybe 10 days ago that 80% of the
4:15revenue came from eight customers.
4:18I was actually shocked how diversified
4:21their customer base actually was.
4:25And if you are uh
4:27>> Well, the the number to me was the the
4:29something like 11 or 12,000 spending
4:31over $100,000. Like that that feels
4:34diversified.
4:34>> Oh, yeah.
4:35>> But I'm but I'm talking about like the
4:36ramp data that's saying like look like
4:391% of companies are spending, you know,
4:4290% of the dollars.
4:45>> Well, I I mean, the ramp data is what it
4:47is, but we have much more authoritative
4:49data now
4:50>> Yeah, sure.
4:50>> from the Anthropic S1.
4:51>> Yeah.
4:52>> And it shows a very diversified spending
4:55base for Anthropic.
4:57>> Yeah.
4:57>> Now, maybe the ramp data is correct for
4:59AI overall,
5:00>> Sure.
5:01>> but the Anthropic S1 like relative to a
5:03thought that, you know, Meta was
5:05spending 10 billion
5:06>> Yep.
5:07>> and there were these multiple
5:08billion-dollar plus accounts that were
5:10driving it all.
5:11That just it's much more diversified
5:14customer base for Anthropic.
5:16And then for the AI broadly, I do think
5:20that something that has happened since
5:22June is you've really seen, you know,
5:24first we saw OpenAI really accelerate in
5:26the summer.
5:28And as OpenAI accelerated, so did open
5:31weight uh models.
5:33And there are so many companies, you
5:35know, whether it's Fireworks, whether
5:36it's Base 10, whether it's Together,
5:39whether it's Modal, whether it's
5:40Cognition, that are at scale
5:43really open weight first, and they're
5:45all growing faster than either Open AI
5:48or Anthropic.
5:49And that was true when Open AI was
5:51really accelerating this summer, and
5:53it's still true in October when
5:55Anthropic has had clearly a pretty big
5:57re-acceleration. Like I just I'm pretty
5:59skeptical that they would have put
6:02the you know the
6:04you know the blank space for the October
6:06ARR in the S1 if they didn't know it was
6:09going to be pretty good.
6:11>> Yeah.
6:11>> Um and I do think it'll be great for the
6:13market to get away from these monthly
6:15ARR things.
6:16>> Totally.
6:17But to me Yeah, it's interesting. Yeah,
6:19like seeing the the the models, the base
6:22tens, the fireworks, the togethers all
6:24out
6:25can probably I feel like this is common
6:27knowledge. Basically all they've all
6:29been out in market and so private
6:31markets investors have been getting
6:33these sort of updates and and a lot of
6:34these companies have been like starting
6:36fund raises, growing dramatically from
6:39the period where they got their first
6:41term sheet to when they're actually like
6:43pulling in capital. And so that that
6:46data is like broadly like I would say
6:49the East Coast is just not aware of that
6:51at all.
6:52Some of them are picking it up, but it's
6:54not like sort of common knowledge,
6:55whereas on the West I would say like
6:58over here people are like, "Yeah, these
6:59companies are we're already at scale and
7:02then they're growing, you know, double
7:04digits just during the period of like
7:07first term sheet to close or like first
7:10close."
7:10>> Well,
7:12my firm Lux Capital led the last
7:14Fireworks round and it was actually hard
7:17to price it because they were growing so
7:19fast.
7:20>> [laughter]
7:20>> Um Like we thought I thought we had a
7:21handshake. Like, "Well, we've increased
7:24revenue." Yesterday's price is not
7:26today's price.
7:27>> Yeah, you're like, [laughter] "Please,
7:28like let's let's just get the deal
7:29done."
7:30>> I mean literally it's like the valuation
7:32was going you know the multiple was
7:34effectively going down, you know, every
7:36day.
7:36>> Every day. Yeah, that's crazy.
7:38>> And if anything it feels like that
7:40ecosystem has accelerated and
7:42diversified
7:44because earlier this year cursor was a
7:46real big driver for all those in Prince
7:48clouds.
7:48>> Yep.
7:49>> And I mean I
7:50ex-cursor growth I think
7:53you know is is what's really
7:55extraordinary and I've just because you
7:56know SpaceX has been in-sourcing the
7:57cursor compute.
7:59>> Yes. So on on cursor SpaceX like the
8:02post merger integration, it feels like
8:05Grokbot has stepped forward in terms of
8:08capability is the key to growth there
8:12like we haven't seen the the models from
8:14xAI like stunned on a particular
8:16benchmark but the cursor team has always
8:18been fantastic on product and
8:19distribution.
8:21In the prior era Grok was putting up
8:23pretty remarkable results on the on the
8:26benchmarks but then there were all the
8:27political considerations and the the
8:30product was banned in some companies.
8:32It feels like Grokbot might be in a
8:34particularly different slot because they
8:35have that go to market motion through
8:37cursor with the cursor design team
8:40behind it. Is that what's driving growth
8:42there? What are you seeing on the
8:43Grokbot side that's actually showing
8:46acceleration in that product category?
8:49>> I mean as a user
8:51Grokbot is the most impactful AI product
8:55to me and to and to our you know to our
8:58to the to the firm
8:59>> Yeah.
9:00>> that we've had yet. Like I you know I
9:02clued together all the stuff with Claude
9:03code and it like really
9:06you know I'm sure it was user error on
9:07my part you know
9:0950 years old. I'm doing my best.
9:12>> [laughter]
9:12>> But Grokbot it just makes it so easy. I
9:14think I'm up to like 17 Grokbots and
9:17every day I think of a new one to add.
9:20>> That's good.
9:20>> Um
9:21>> Yeah you getting to the layer of
9:23somebody the the senior manager senior
9:26Grokbot.
9:27How are you how are you personally
9:28dividing
9:30AI agents across your personal life and
9:33your professional life because a lot of
9:35people will have like a family office
9:38with a personal assistant who manages,
9:40you know, getting gifts for their family
9:43members or planning birthday parties.
9:44And then separately, they'll have an
9:46executive assistant who plans their
9:47business travel and their calendar for
9:49work stuff. And it feels like we might
9:52be entering a world where those two
9:53merge together and you have one agent
9:55that does everything. But we also might
9:57wind up in a world where people are
9:58using Grok Bot at work or Claude Code at
10:01work. And then Instinct or Muse in
10:03personal and Open AI has products in
10:05both. And there's sort of like a an
10:08array, but it's a decision for everyone.
10:09How have you been grappling with it?
10:11>> Well, I guess you would say I I I guess
10:13I have executive assistant privilege cuz
10:15my assistant Susan is exceptional.
10:17>> Yeah.
10:17>> And
10:18>> Exactly. [laughter]
10:18>> Not going anywhere.
10:19>> I felt the need, you know, to
10:22uh yeah, I have I have of course
10:23downloaded Instinct and played around
10:25with Muse.
10:26Um but and I could see they would be
10:30super amazing if if I didn't have
10:32someone like uh
10:33Susan in my professional life, but I do.
10:36Um
10:38but for the more professional use cases,
10:43I mean, Grok Bot has just been
10:44transformative
10:47for me.
10:48>> Yeah, there is something
10:49>> of
10:50Yeah, so much of investing is making
10:51sure that the right information gets to
10:53the right person at the right time. And
10:55Grok Bot is so good at that.
10:57>> Yeah.
10:59And there's also something so satisfying
11:01with the AI agents of like
11:04it can be 10:00 p.m. on a Sunday and you
11:06can kick off like a real research effort
11:09that yes, you could probably call
11:11someone on your team and say, "I need
11:12you to like that would be crossing a
11:14professional line." All of a sudden
11:16you'd be like kind of a hassle of a work
11:18of a boss and people might not want to
11:19work there cuz they're like, "Yeah, he
11:20calls me every time he takes a shower.
11:22He has a shower thought and then I got
11:24to work for 6 hours." But instead, uh
11:26the AI is like, "Yeah, boss, I'm right
11:27here. Oh, I'd love I'd love to look that
11:29up for you and spend a a tokens figuring
11:32out, you know, what the market size of
11:34something or other. But, yeah.
11:36>> Well, my policy is actually anybody in
11:38TradFi is always welcome to call me um
11:40anytime on the weekend. I try not to
11:43call people
11:43>> Yeah.
11:44>> on the weekends.
11:45Um
11:46Yeah.
11:48>> Um so,
11:49>> Uh how
11:50uh so,
11:53I think like, you know, rewind this week
11:55is, you know, uh Anthropic S1 week and
11:58so there's there's a lot of focus there.
12:00Last week it felt like the focus was
12:03like the general attention was towards
12:04how much supply is coming online next
12:07[clears throat] year, how many gigawatts
12:10uh SpaceX is going to bring online. How
12:11are you uh
12:13what's your what's your, you know,
12:16generally you've maintained uh a sort of
12:19bullishness that that some other people
12:21on the timeline have lost here and
12:23there. But, uh how are you processing
12:26like how the
12:28how the how the industry can sort of
12:30absorb all of that supply if we hit the
12:33higher end of some of the ranges being
12:34discussed.
12:36And and and is it just Jevons paradox,
12:39you know, again and again and again
12:41uh and again until we finally ever
12:43everyone says like, you know, everyone
12:45likes to say it, but it's hard to fully
12:47internalize.
12:49>> Yeah, I think that's I think that's
12:50right. It's it's easy to say, it's it's
12:52hard to internalize and believe. And
12:54October was kind of been a strange month
12:56because
12:58the price to rent essentially any kind
13:00of a GPU has gone up every day.
13:03So, the spot market for compute is
13:06getting tighter. And it's not just the
13:08spot market. Um you know, it is
13:10I do love reading these primers on X
13:13from, you know, commodity and macro
13:16investors about these super complicated
13:19concepts called backwardation and
13:21contango and what they might mean for
13:25the forward curve of compute and the
13:27future price of compute. You know, these
13:30are contango backwardation.
13:33Um
13:34but, you know,
13:35Silica data, which I do think is, you
13:37know,
13:38my firm is an investor in them, but I I
13:39do think they're the most authoritative
13:41data source.
13:42Carmen Lee, the founder, stood up at GTC
13:45and said that the compute market is now
13:48in contango.
13:49>> Mhm.
13:49>> The you know, the forward it's not just
13:51that the spot market is tight.
13:54If you want to get a cluster a year from
13:56now,
13:57you have to pay a lot of money.
13:59>> Yeah.
14:00>> And this is And this is maybe another,
14:02you know,
14:03discontinuity between public and private
14:05markets.
14:07>> Mhm.
14:07>> But just the number of companies we have
14:10in our portfolio who are like, "Okay, we
14:11got a cluster of 4,000 Blackwells, and
14:14you know, and we got it in
14:17May or June of '26, and we want to bring
14:20on another cluster of 4,000 in September
14:24of 2027."
14:25>> Yeah.
14:25>> And they're shocked by how much more
14:27they have to pay in the future.
14:30And it's not like they're Everybody is
14:33unaware of the Everybody's aware of the
14:35supply that's coming on,
14:36but even with all of that supply coming
14:39on, it is hard
14:41for the companies in our portfolio
14:44to find compute
14:46even well into '27.
14:48>> Mhm.
14:49>> So, for you know, all the commodity
14:51experts who were focused on
14:52backwardation and what that implied,
14:55I got news for you.
14:57We're not in backwardation anymore.
14:58>> Mhm.
14:59>> And then, I do think
15:01as far as how much capacity is going to
15:03come online,
15:05one thing that I think is actually
15:07healthy about what's happened in these
15:08financing markets,
15:10is I hope it concentrates capacity and
15:12supply
15:14in the competent operators,
15:16where they can finance it through
15:18prepayments. Um you know, you can kind
15:20of finance these GPU purchases two ways.
15:22And by the way, CoreWeave is great.
15:24CoreWeave runs GPUs well, they bring
15:26them on. They They're very, very good.
15:29They just sadly locked in a vast amount
15:31of compute in long-term contracts that
15:34are now way, way under market. You know,
15:38and I think they're doing
15:40>> So, they sold their They sold their GPUs
15:42for fixed prices, and that's why they're
15:45getting hammered right now. It's not
15:46that they don't have compute, they're
15:47not bringing it on. They have the
15:49semi-analysis diamond tier. Everyone
15:51knows that the clusters work, they're
15:53good, they are reliable, but they just
15:56sold for too cheap, and they should have
15:58hold held on or had variable pricing.
16:00>> is you think Elon studied that and said
16:02like we're not going to make that same
16:03mistake, and that's why they're adopting
16:05this more short-term
16:08contract
16:09>> I don't Elon is a first principles
16:11thinker. I don't think he was probably
16:14studying CoreWeave or Nebius or really
16:16anyone else. He was probably reasoning
16:18from first principles. The conclusion he
16:20came to
16:21is prices are going up.
16:23>> Mhm.
16:24>> We if we're willing to sell on the spot
16:25market, we can have paybacks well under
16:28a year,
16:29>> Yeah.
16:29>> get big prepayments,
16:31>> Yep.
16:32>> and that's
16:33I think that's a better probably
16:36financing model for the market we're in.
16:38And what I also like about it, and by
16:40the way, CoreWeave I think is very aware
16:42of this. They're trying to get more spot
16:44exposure. They're trying to, you know,
16:46um, do less of these big long-term
16:48off-take agreements.
16:50>> Mhm.
16:50>> Um,
16:51but standing up a big GPU cluster,
16:54that I just say um, bringing on the
16:56power, standing up a cluster,
16:58it's like running a Formula 1 team.
17:00>> Yeah.
17:01>> And
17:02a bunch of people who have no real
17:04relevant experience, you know, whatever,
17:06they acquire some land and some power,
17:09and then you know, they get a zero down
17:11payment off-take agreement,
17:14um,
17:15from someone, and a lot of that is being
17:19cleaned up and I think um
17:23you know, supply is being
17:26What this is a forcing function to
17:28allocate GPUs to people who can actually
17:32bring on a cluster, energize the GPUs,
17:35run it well. Like, you know, running a
17:37Formula One team, it's like really hard.
17:39You got to like pack everything up, put
17:41it on a plane, go to a new city, unpack
17:45it, set it up, set the car up. You need
17:47all the people, you know, to like, you
17:50know, whatever, change the tires, to
17:52change the rear wing if you get in a
17:54collision, to, you know, do whatever
17:56needs to be done during the race. To do
17:57the strategy, to communicate with the
17:59driver. It's really, really hard if and
18:03it's not like, you know, some of these
18:05people, they don't have the relevant
18:07experience. You know, CoreWeave, Crusoe,
18:11Nebius, SpaceX, the big hyperscalers.
18:15You know, these are companies that can
18:16bring on GPUs, energize them, and and
18:19run them. Um
18:22But and I do think something that was
18:23announced at GTC Berlin is actually
18:25pretty important for this and that's
18:26that DGX Pod, which is going to make it
18:29a lot easier, you know.
18:30>> Oh, sure.
18:30>> Seven racks, it comes pre-assembled, you
18:33know, you just kind of literally plug it
18:35in and it works.
18:36>> Mhm.
18:37>> And right now when you get the GPUs, you
18:39know, maybe if you're if you're SpaceX
18:41and you're really, really exceptional at
18:43execution in the physical world, you can
18:45get them energized quickly. But a lot of
18:48people it takes them 6 months, you know,
18:503 months, 6 months to get them
18:52energized, really get them working. And
18:54if we can
18:56through, you know, this new product from
18:58Jensen,
18:59um get that time down, I think that's
19:02that's also healthy cuz that effectively
19:03takes a lot of working capital out of
19:06the market, you know, you have as these
19:08numbers are getting bigger, if it takes
19:096 months to energize them, you know,
19:11that could be hundreds of billions of
19:13dollars of assets that are not
19:14generating revenue. So, I think that's
19:16another
19:18that's a product innovation that's going
19:19to help with this.
19:21But, I think this is um
19:23you know, we we will see who who
19:26energizes CPUs in in
19:29in in '27. I'm I'm confident that the
19:31you know,
19:32the companies I named, I think they're
19:34going to
19:35I mean, the future is always uncertain,
19:37but these are companies with a proven
19:38track record of execution.
19:40>> Yeah.
19:41>> And we'll see about others.
19:42>> So, I have a follow-up on that. Um one
19:44is uh just one click higher from the Neo
19:47Clouds or the Neo Labs, and there's been
19:48this speculation that Neo Clouds might
19:51Neo Labs might sort of pivot into Neo
19:52Clouds. Is that dynamic going to look
19:54like they signed a multi-year off-take
19:56agreement with a Neo Cloud, and then
19:59they want to just sell into the spot
20:00market? That's what's going to happen?
20:02>> Yes, I mean, what Sarah Friar said, and
20:04Sarah Friar is an exceptional CFO.
20:07>> Yeah.
20:07>> She said, you know, everybody made fun
20:09of us for signing all these compute
20:11commitments, and we could sell them all
20:13at a three, four, or five x premium.
20:15>> Today.
20:15>> And she was being literal.
20:17>> Yep.
20:18Yep.
20:18>> It's so, if you if you signed a
20:21long-term compute contract
20:23>> Yep.
20:23>> really at any point prior to May
20:27of 2026,
20:29>> Mhm.
20:29>> you are way in the money. You're buying
20:32compute for, you know, maybe 10 or 15 10
20:35or 15 dollars a watt.
20:37>> Yeah.
20:37>> And you can sell it at a massive premium
20:40to that today.
20:42>> Yep.
20:42>> And then the point the point, you know,
20:44the Carmen made about, you know, the uh
20:46the forward curve
20:48being contango is actually really
20:49important, and just like it's not just
20:51that the spot market is tight today, you
20:54have to pay vast amount even with all
20:56this capacity coming online, you have to
20:58pay a truly vast amount to bring on a
21:00cluster
21:01in mid to late '27, and I think that's
21:04something that just maybe public markets
21:06are not appreciating even though
21:09you know, it's not
21:11all of this is public. You know, Sarah
21:13Fryer said that.
21:14>> Yeah.
21:14>> You know, the
21:15>> Yeah, yeah.
21:15>> all these inference clouds have kind of
21:17given press releases. You hear a bunch
21:20of people on X talk about how hard it is
21:22to find compute into 27. Carmen Lee
21:24stands up and says, you know, like the
21:27market in 27 is is crazy strong. So,
21:30it's all out there, but I think for
21:33whatever reason, it's not legible to the
21:35public markets.
21:36>> Sure.
21:37>> Jordan
21:38>> What talk about your
21:40outlook on biology as a as a as an area
21:43that that the labs are investing in. It
21:45feels very exciting and that there's,
21:47you know, tremendous potential, but I
21:48think the overall concern is around
21:51timelines. It's like there's very real
21:53costs like coming online, infrastructure
21:56costs, and
21:58opportunities in biology seem like
22:00they're very real, but uh may take, you
22:04know, years to get these, you know, will
22:07almost certainly take years to get drugs
22:09in market and actually drive, you know,
22:11revenue that would then
22:13finance.
22:14>> Yeah, I mean, we'll we'll see how long
22:16it takes. You know, Montana has an
22:17interesting new program for for new
22:19drugs, which I think is helpful.
22:22But
22:23I think a really big unlock, we we have
22:26a lot of compounds,
22:28and people have used AI
22:30to figure out that you can use um may
22:33maybe a drug that we had for 50 or 60
22:36years
22:37for a very rare disease that's
22:40devastating for like 10,000 kids a year
22:42and effectively cure that disease.
22:45And a friend of mine did this, and it's,
22:46you know, it's it's
22:48it's his story to tell, so I'm not going
22:50to tell it. But AI has already saved a
22:53lot of lives,
22:55and I do think, you know, the AI
22:56industry, it's, you know, they're like
22:59-5 out of 100 from like a PR
23:02>> [laughter]
23:02>> perspective.
23:03>> 100%.
23:05>> But they
23:06>> We got to be more like Pharma, the
23:07popular kids.
23:08>> Yeah, exactly.
23:10>> Well, no, they need to start telling
23:11these stories.
23:12>> Yeah, yeah, yeah.
23:13>> Um and by the way, are you guys are
23:15coming to Tomorrow Act?
23:16>> We are. We are. November 17th, 18th,
23:19we'll be there with you.
23:20>> Lineup is The lineup is unbelievable.
23:22>> is crazy. Jensen Wong, David Solomon,
23:25Mike Rowen, Lisa Su, Stan Druckenmiller,
23:27so many more.
23:28>> Li Lu, so many people.
23:30>> Yep.
23:30>> But the one thing I just want to say is
23:32I get like
23:3425 emails or texts a day saying, "Hey,
23:37man, can I come to your conference?" And
23:39the answer is you can just sign up.
23:42>> Yeah, you can just sign up.
23:42>> Anyone can come. Now, there's going to
23:44be a security screen cuz of some of the
23:46people who are coming and I do, you
23:47know, there are unannounced speakers
23:49that are, you know, comparable in
23:51caliber to some of those you referenced.
23:53So, we're going to have a lot of
23:55security at the conference and
23:56everybody's you'll be screened.
23:58But you can just sign up.
24:00>> Yeah.
24:00>> You do not need to ask me. You don't
24:02need to ask Anthony. You can just come.
24:05>> Yeah.
24:06>> And it's like the whole goal is to kind
24:08of democratize
24:09investing. And it's And in addition to
24:11those companies, you know, we are We're
24:13going to have Lynn from Fireblocks,
24:15Chase from Crusoe, Scott from Cognition.
24:18We're going to have Max from Lagora.
24:20>> Yeah.
24:20>> We're going to have a lot of companies.
24:22>> Yeah, Base 10.
24:23>> Two in. Two in from Base 10.
24:25We are going to have a lot of these
24:28CEOs there
24:30that are going to be on a stage and then
24:32we're going to have, you know, dozens of
24:34other private CEOs.
24:37And you know, it's a it's an attempt to
24:39kind of like open up the aperture
24:42of, you know, information for retail
24:45investors and democratize this. So, like
24:48if you're wondering about coming, just
24:50sign up. You don't need to ask me.
24:54>> [laughter]
24:55>> You can just sign up.
24:55>> summit.com,
24:57not Gavin Baker's email.
24:59>> [laughter]
25:00>> You can go to a URL.
25:02>> Once you're there, you can get you can
25:04get Gavin to sign
25:06>> And they will email you when you sign
25:08up. Yes.
25:10What [clears throat] what what do you
25:12think will change in the content in the
25:15conversation between last year's event
25:18and this year's event? What are you
25:19hoping to get across?
25:20>> We did this is the first time we've done
25:22this. This is the first
25:22>> Oh, sure.
25:23>> Yeah, yeah.
25:23>> Oh, sorry. Yeah, yeah.
25:24>> We've decided we're going to have you
25:25know, we're going to have by siders
25:27doing the fireside chats.
25:29Um
25:30We're going to have you know, management
25:32teams are going to have the opportunity.
25:33We've if you are an active contributor
25:36on X, Substack, Reddit You get to come
25:39for free. Um
25:41and then management teams are going to
25:43be able to do a breakout session with
25:46all of those X accounts. So, if you're
25:49if you're one of the you know, four or
25:52500, you know, kind of
25:54you know, financial commentators that
25:55we've given a scholarship to you're
25:58going to get you know, a lot of
25:59management teams are going to opt in and
26:01you're going to be able to kind of have
26:03your own session with the management
26:06team to ask your questions. Then you're
26:07going to be able to go, write what you
26:09learned for your audience.
26:13And I'm you know, I think we have a
26:14really high quality, you know, I mean I
26:16I I I picked most of them and you don't
26:18have to be bullish. You know, I want you
26:20know, I love skeptics. I love bears.
26:22>> Yeah.
26:23>> You just you just have to be polite and
26:25nice and constructive and you can be
26:26bearish while being polite, nice and
26:29constructive and you know, we're we have
26:31loads of we have loads of skeptics, we
26:33have loads of optimists. Like I think
26:35it's going to be great and then I I hope
26:37every CEO does you know, uh
26:39you know, kind of a hit on your show
26:41after they're on stage or before.
26:43>> Yeah.
26:44>> I'm Yeah, it might be the longest it
26:46might be the longest TV PN episode ever.
26:48>> I think so.
26:49>> Yeah, we How do How do with how how do
26:51you deal with
26:52um
26:53all the uh all the attention, positive
26:55and negative?
26:57Does it ever
26:58>> for me?
26:58>> Yeah, for you.
27:00For you. I know, I just feel like I you
27:02I you you talked about your your mindset
27:04right at the beginning of the show, just
27:05you know, you know, it's a choice to be
27:07happy, but uh I feel like when it comes
27:09when the the the the
27:11the spotlight, you know,
27:13once every week shines directly on you,
27:16there's a lot of love,
27:18a little bit of hate.
27:19>> No, there's there's
27:21>> The whole the full spectrum of human
27:23emotions.
27:23>> can call it what it is. I mean, there
27:25are there there are a lot of there are a
27:26lot of detractors, let's say.
27:28I'd just say, "Look, Antonio, who I'm
27:29doing the conference with, he says being
27:31on X is an exercise in cultivated
27:33equanimity.
27:35I don't take any of it personally. I'm
27:37not on X to be right.
27:39I'm right now, I would you know, I'm I'm
27:42a responsible optimist, but I'm trying
27:45to be open-minded to new negative facts
27:47and constantly search them out. But I'm
27:49really on X, you know, to learn, and I
27:53you know, all of the hate, like I try to
27:55sift through it for oh well, that is
27:57actually a very good compelling bearish
28:00argument.
28:01Um you know, like there was an
28:02interesting post,
28:04I forget it was like PE analyst or LBO
28:07analyst, and it just said, "Hey, you
28:10know, the fact that Anthropic is
28:11dedicating 74% of their compute to
28:13training,
28:15you could read it
28:16um on the one hand, it's bullish cuz cuz
28:18they could be generating a lot more
28:19revenue.
28:20On the other hand, if all this compute
28:22is being
28:23uh dedicated towards training, and for
28:25any reason somebody decides you need
28:28less compute for training,
28:30that's a lot of supply that could hit
28:31the inference market. And
28:33hey, wow, that's like a really
28:34interesting,
28:36you know, bearish perspective. And so,
28:40you know, um
28:42I'm human. I don't
28:44Like I'm human, but it's also just like
28:47I'm here to learn. I want to get to the
28:49right answer. I'm sharing my views. If
28:53somebody shares and, you know, helpful
28:55negative perspective, I try to thank
28:57them. I need to go like thank that that
28:58person, Albeo Analytics if it was. I
29:01didn't have time.
29:02Um
29:04But yeah, it's it's an exercise in
29:05cultivated cultivated equanimity. You
29:07can't
29:08You can't um
29:10You can't take it personally. Yeah. But
29:12I do I am trying
29:15you know, doing what little I can, you
29:18know, as like Gavin Baker. You know, I
29:20don't have
29:21millions or tens of millions of
29:23followers.
29:24But I am trying to help paint, I think,
29:28a more
29:29optimistic vision for AI.
29:33Doing what little what I can to help the
29:35American public understand that, hey,
29:36like you actually want a data center
29:39>> Yeah.
29:39>> in your in your town, you know. One of
29:41my favorite exchanges was somebody said,
29:43"Well,
29:45I'll wait until they start building data
29:46centers in the wealthiest county in
29:48America." Well, Loudoun County is the
29:49wealthiest county in America. They have
29:50the highest per capita data centers.
29:53>> Yep.
29:53>> Um
29:54>> Virginia, US East.
29:56>> Yeah, just understand US East
29:58availability zone.
29:59>> Yep.
29:59>> So, just understanding that like
30:02I understand why you know, some people
30:04are worried. Um there there's been a lot
30:06of fear. There's, you know, there we've
30:07all seen scary movies. But a lot of
30:10awesome things are happening today
30:12because of AI for ordinary Americans,
30:15for blue-collar Americans, for these
30:18small towns. We're reindustrializing
30:20America. We're
30:22you know, massively increasing our
30:24electrical generation capacity, which is
30:27you cannot reindustrialize without
30:28power. That's job number one. So, I
30:31think a lot of good things are happening
30:32and I'm just trying to do what little I
30:34can to at least tell that side of the
30:36story. And then be pretty balanced.
30:39Like, you know,
30:40Anthropic, I called it like I saw it in
30:42the summer. They were losing market
30:44share.
30:44>> Mhm.
30:45>> I'm calling it like I see it in October.
30:48They are reaccelerating and gaining
30:50share.
30:50>> Yeah.
30:51>> The S1, it was an upside surprise to me.
30:54I'm trying to call balls and strikes cuz
30:57my job is to just get it right.
31:01>> Yeah.
31:02Jordi mentioned and it's funny cuz you
31:04know, people will get so flustered by
31:07you know, seeing different posts on X,
31:09but imagine how boring X would be if no
31:11one ever took a stance. It was just
31:13[laughter] like, well, it might be this,
31:15it might be that.
31:16And it might be this.
31:17>> No, no, no. Every post has to be it's
31:19over or it or we're back. This is the
31:20best thing ever the cuz then we can have
31:22a debate. At least you took a stance.
31:24>> Yes.
31:24>> It's great.
31:24>> Yes.
31:25>> Jordi mentioned
31:27getting sharp on bio, starting to think
31:28about opportunities there.
31:30How deep have you gone in the AI stack?
31:32I saw a funny post that was like, you
31:34could spend all your time thinking about
31:37open AI and Anthropic and Google or
31:39>> and you know, meet all the neo clouds
31:41and meet Lynn and Tuan and figure out
31:44what's going on in the private markets
31:46or you could just buy copper and you
31:47could just sit there and hold copper,
31:49right? Cuz like it's almost a perfect
31:50proxy or something like that. But how
31:52far have you gone down into energy
31:55infrastructure, commodities? Like how
31:57how sharp are you on those? How how
31:59relevant or interesting is are those
32:01markets to you in this world since they
32:03do seem adjacent?
32:05>> No, I pay close attention to all of
32:07those. I ran a diversified fund in my
32:09life for 9 years. You know, I was an
32:10energy investor.
32:12Um,
32:13you know, natural resources. I always
32:16for many years I did
32:18would do like kind of marketing meetings
32:20with the uh
32:21with the gold analyst at at at my my
32:25prior employer Fidelity. So I always,
32:27you know, knew about the merits of gold
32:29from a portfolio construction
32:30perspective, would always have it in the
32:31fund. By the way, just cuz it triggers
32:33people, I want to say it.
32:35I find it so amusing how upset gold
32:39balls get at the idea of asteroid
32:41mining.
32:42>> [laughter]
32:44>> The idea
32:45of asteroid mining?
32:46>> Oh, yeah.
32:47>> No one No one has ever been I've never
32:51gotten more hate
32:52>> Yeah.
32:52>> than from the gold crowd about the idea
32:54of asteroid mining and the and the
32:56asteroid supply.
32:58>> I mean it's so funny.
32:59>> I mean [laughter] it was like talk about
33:01a challenge to your world view.
33:03>> Wait, are are they mad about that new
33:04algorithm?
33:05>> It's something that you have to like you
33:06need to imagine technological
33:08disruption, right? Like look at the
33:10diamond industry, right? Like if you if
33:13you understood that that technology
33:15disruption was possible, you probably
33:18built labs like, you know, 20 years ago.
33:21>> you would have done something different
33:23with that, you know,
33:24with the diamonds in your family. But,
33:26um
33:26>> [laughter]
33:27>> But, as far as I mean, I do we have
33:30spent a lot of time on like energy on
33:32the industrial infrastructure, on on
33:35natural resources. Um
33:37but we have not made it a big part of
33:40the portfolio. We've done some things in
33:42private markets
33:44as far as um trying to help with rare
33:46earths. And but this is not through
33:48acquiring mines. It's more,
33:51you know, the problem with rare earths
33:52is not that they're rare, it's that
33:54refining them is a very dirty process.
33:56You know, so we we we
33:58you know, that we've invested in a
33:59company that, you know, has a chemical
34:01process for refining rare earths and you
34:04know, I actually think DARPA has done
34:06some pretty smart stuff. Like they kind
34:07of have a
34:08a lot of different efforts around this.
34:11But, yeah, I mean, I try to pay as close
34:13attention
34:14um to all of it as I can because you um
34:18you know, as an investor, you you you
34:20never know where a weak but important
34:23signal is going to come from. And so,
34:25want to cast as wide a net as possible,
34:28but then also kind of focus maybe the
34:30our
34:31on the areas where I do feel like maybe
34:33we have
34:34the the biggest competitive advantage as
34:36a firm.
34:38Um
34:38>> How are you thinking about the evolution
34:40of like the
34:42the marginal capital decider sort of
34:46changing in 2027 potentially? Like if
34:48you go back a couple of years, you have
34:50essentially like seed stage VCs making
34:53bets on these AI labs, and then
34:54eventually some you know, more scaled-up
34:57venture firms being important capital
35:00partners. Some of them have taken board
35:02seats, but they are not deciding the
35:04race now. It's big hedge funds pre-IPO
35:07rounds, but we're coming up to a world
35:10where the fidelities of the world, the
35:12insurance agencies are going to be
35:14involved, the insurance funds, and
35:17sovereign wealth funds are going to be
35:18increasingly involved in financing the
35:20build-out. Like what does the next year
35:22look like from a perspective of like how
35:25we get the big dollars into the AI
35:28build-out?
35:30>> I mean, I think you named all the big
35:32dollar sources. Yeah. And something that
35:34I think is pretty underappreciated is,
35:36you know, American mutual funds.
35:37>> Okay.
35:38>> You're limited in the percentage you can
35:40hold in a liquid asset.
35:42>> Sure.
35:43>> And so SpaceX
35:45and Anthropic
35:46>> Yep.
35:47>> going public
35:48>> Yep.
35:48>> is actually going to free up a huge
35:51amount of buying power
35:53>> Interesting.
35:54>> in private markets
35:55>> Yep.
35:56>> Um once those are fully unlocked, cuz
35:59it's they have be public and unlocked.
36:01>> Interesting.
36:02>> So,
36:03um
36:04I think you will see
36:06mutual funds, you know, and they they
36:08they've always been involved, you know,
36:11my former colleagues, Capital Research,
36:13T. Rowe Price,
36:14Invesco, G Sam, you know, Franklin
36:17Templeton. If I forgot anyone, I'm
36:19sorry. Um I love you all.
36:23But like they've been they've been
36:24pretty involved. I mean, the sovereigns
36:26have been very involved, particularly
36:29um you know, particularly the ones in
36:31the GCC and particularly the ones in the
36:33UAE.
36:33>> Yeah.
36:34>> Like MGX, um Mubadala, all the
36:37sovereigns there. I mean, at Trafigura,
36:39we co-led Crusoe with Mubadala.
36:41>> Yeah.
36:42>> So, I think the um the UAE in particular
36:45has been extremely front-footed
36:47>> Yeah.
36:47>> with all their sovereigns with respect
36:49to AI.
36:50>> Yeah.
36:50>> Um
36:51>> And now Apollo's taking a stake in
36:53Mubadala, uh which sort of completes the
36:56the cycle.
36:57>> Yes. Yeah. Um
37:02But, I mean, that that's going to have
37:04to evolve, but it's also
37:06I do think we're probably getting
37:10has these underpriced so all these
37:12clouds, they signed contracts
37:16at well below what are below market
37:18values. And as those contracts reprice
37:20higher,
37:21as they bring on more compute with more
37:24spot like we discussed with, you know,
37:26the same thing that CoreWeave is is is
37:28trying to do.
37:30I do think you're going to see like
37:32there's a pretty big acceleration in the
37:34clouds last quarter. And you know,
37:36forget relative to expectations this or
37:39that, but I think you're going to see a
37:41pretty continuous acceleration
37:44in the sum of the big hyperscalers cloud
37:48revenue growth,
37:49the neo clouds, that number's just going
37:51to continue to accelerate.
37:54And then, you know, it's it's dropping
37:56down at very high margins so that, you
37:58know, operating cash flow is following.
38:01And then you're going to be able to, I
38:03think, finance
38:04probably more than people think of the
38:072027 buildout um
38:11>> [clears throat]
38:12>> out of cash flow.
38:13>> Mhm.
38:13>> Like I don't know if we've seen the peak
38:15of external financing,
38:17>> Mhm.
38:17>> but um I think we're getting close.
38:20>> Mhm. Yeah. Yeah, that's
38:22>> And that'll be really helpful.
38:23>> Totally. Yeah. You know, that No one can
38:25complain about that. That's just uh
38:27yeah, investing in the future.
38:28>> AI skeptics will be
38:30>> in a bad place.
38:31>> in a
38:31horrible place. We're going to have to
38:32check on check on them.
38:34You're going to have to go around
38:35>> And look, maybe I'm maybe I'm wrong and
38:38you know, but like I will change my mind
38:40when the facts change and I'm just going
38:41to try and keep calling balls and
38:42strikes and then trying to
38:45you know, be an advocate to help
38:48understand ordinary Americans
38:50that this technology can and should be
38:52awesome for them.
38:53>> Yeah.
38:54Yeah.
38:55>> Yeah, I'm I'm I'm I'm very interested to
38:56see
38:57how
38:59uh Anthropic and Open AI Open AI
39:01becoming public companies like impacts
39:03like overall sentiment specifically on X
39:06because right now you've have such a
39:09small number of people
39:11that actually hold
39:12shares in these companies, right? And so
39:14like maybe there's 50,000 people that
39:16have some incentive to be relatively
39:19optimistic one, you know, or you know,
39:21you should be able to see both the the
39:23bull case and the bear case for for any
39:25company.
39:26But when that changes and you have
39:29millions of people globally that
39:30suddenly can have exposure, how does
39:32sentiment change, right? And I think it
39:34I think people are going to be I I think
39:36it will be less much less extreme once
39:39you have the the the sort of middle of
39:41the road investor has some exposure,
39:43right? And it's not like totally binary
39:46where like very very small number of
39:47people have exposure like and and and
39:50it's direct exposure, right? Because if
39:51you happen to own shares through some
39:54retirement fund or through a hyper
39:56scalar or things like that.
39:57>> Or um
39:58>> it uh it's not the same.
40:00>> and maybe some of them will do what Elon
40:01did, you know, every child in America is
40:03going to have a share of SpaceX. That's
40:05awesome. I think that's way better for
40:06America than setting up some foundation
40:08>> Yeah.
40:09>> that you know, a board or
40:12a bunch of bureaucrats are going to
40:14decide how to allocate the money.
40:15>> Yeah.
40:16>> You know, just give people the money.
40:17>> Yeah.
40:18>> Like [laughter] I hope
40:19>> It's simple. Yeah.
40:20>> Just let people make individual
40:22decisions.
40:22>> Yeah.
40:23>> And you know, I would you know, I would
40:24encourage more company and more
40:27you know, more wealthy people in America
40:29to do what Elon and Michael Dell and and
40:32of course, you know, my friend Brad
40:33Kesner
40:34>> Yeah.
40:34>> have done. Uh,
40:36like I think that's awesome for America
40:37and great for capitalism.
40:38>> Yeah, and it's it feels so important
40:40that that infrastructure got built
40:42before the big liquidity wave, before
40:45SpaceX goes public so that it's there
40:47and it's been tested and it's been
40:48implemented and rolled out uh so that
40:51anyone who wants to participate in that
40:53can actually go and like take advantage
40:54of it uh without needing to, you know,
40:57say, "Oh, it'd be nice if the government
40:58had a program that will allow let me
41:00give shares directly to the American
41:01populace." Like, nope, it's there, it
41:03works, it's up-to-date, like it's in the
41:05app store. You can go download it. Um I
41:07love it.
41:08>> give give people the money.
41:09>> Yeah.
41:10>> Give everybody skin in the game
41:11capitalism.
41:11>> Yeah, no, I completely agree. It's
41:13great. Uh well, thank you.
41:14>> We'll see you very soon.
41:15>> We'll see you.
41:16>> Yeah, take care. Tomorrow summit.com.
41:18>> Tomorrow summit.
41:19>> November 17th, 18th, Austin, Texas.
41:22We'll see you there. Have a great rest
41:24of your day. Cheers, guys. Good to see
41:25you. How about

About this transcript

This page contains the full transcript of Gavin Baker on Anthropic's S-1, GPU Prices, and SpaceX by TBPN, generated from the public captions YouTube serves with the video. The transcript has 7,675 words across 1,310 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

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