00:00:00
Hey, 2026 could be an all-time record
00:00:03
for IPOs.
00:00:04
>> The AI IPO of the year so far. That
00:00:08
company is Cerebras.
00:00:09
>> Cerebras Systems founder and CEO Andrew
00:00:11
Bubman. We are participating in
00:00:13
something extraordinary on everything we
00:00:15
do. We are the fastest bar none.
00:00:17
>> Well, Marshall is the co-founder and CEO
00:00:19
of Planet Labs.
00:00:21
>> Space and AI are really um a match made
00:00:24
in heaven. They're getting married. In
00:00:25
fact, just like Google figured out how
00:00:28
to index the internet and make it
00:00:29
searchable, we are indexing the earth
00:00:32
and making it searchable. He's got his
00:00:34
glasses, the famous red glasses. Brad
00:00:36
Gersner is here, founder and CEO of
00:00:37
Autoimmet Capital, a leading tech
00:00:39
investment firm.
00:00:40
>> I believe that the wave is the biggest
00:00:42
wave in the history of technology, will
00:00:44
be incredibly beneficial for America.
00:00:46
I'm rooting for all of them because I'm
00:00:48
rooting for America. Ladies and
00:00:50
gentlemen, please welcome Brad
00:00:53
Gersonner, Will Marshall, and Andrew
00:00:56
Feldman.
00:01:01
>> On the couch, we switch on the couch.
00:01:05
>> Nice to see you, my friend.
00:01:07
>> Hey, big boy.
00:01:09
>> Nice to see you.
00:01:09
>> Last last time I saw you, we were in
00:01:11
Davos.
00:01:12
>> Yes,
00:01:12
>> we were in Davos causing another drop.
00:01:15
Another JL. Do you hear that little
00:01:16
Davos?
00:01:17
>> We were just, you know, it was preipo.
00:01:19
We're chopping it up
00:01:20
>> with Davos.
00:01:21
>> We're in Davos.
00:01:22
>> Hanging out at Davos.
00:01:23
>> Well, no. Listen, I was Everybody knows
00:01:25
the story. I'm supposed to go on my
00:01:27
yearly Japan ski trip. Sax calls me
00:01:29
>> with Tucker.
00:01:30
>> Yeah. Well, anyway, we don't drop that
00:01:32
name, but I'll pick it up for you. Put
00:01:34
it over here.
00:01:35
>> Tucker. Anyway, so I cancel on Tucker. I
00:01:37
cancel because Sax calls me. He says,
00:01:39
"Listen, Pus needs you, the world's
00:01:41
greatest moderator in Davos." I said,
00:01:43
"No problem."
00:01:44
>> I said, "Saxs, POTUS, and Davos." So I
00:01:48
said, "When?" He says, "In 3 days." I
00:01:51
say, "You got it." I go and they give me
00:01:53
a badge. And it's like the special green
00:01:55
badge and they buzz you through the
00:01:58
security and I look at the monitor and
00:02:00
it says Jason McCabe Calacanis with
00:02:03
Donald J. Trump.
00:02:05
>> Oh wow. How did you feel?
00:02:08
>> I thought it was hilarious.
00:02:10
So then I went and we did a great
00:02:12
interview there and we did like six or
00:02:14
seven of these great all-in interviews
00:02:16
and it was fun.
00:02:16
>> Let's start this because uh the two of
00:02:19
you guys run two of the most interesting
00:02:22
and consequential newly public companies
00:02:24
in the stock market. Andrew Feland is
00:02:26
the founder and CEO of Cerebrus. Will
00:02:27
Marshall is the founder and CEO of
00:02:28
Planet Labs. But you are also the
00:02:30
insight and a gateway for all of us to
00:02:32
understand these two big trends. One is
00:02:35
in AI silicon, the other one is in space
00:02:38
data centers. I think it would be a
00:02:39
really interesting thing to
00:02:41
>> and emerging.
00:02:42
>> And emerging. Yeah. Um but let's just
00:02:45
take one step back. Uh you just heard
00:02:48
the last conversation about being
00:02:50
public, going public early. Let's just
00:02:52
talk about that cuz I'm just very
00:02:53
curious. How's it been? It's been 3
00:02:55
weeks or so for you. It's been about a
00:02:57
year and a half or two years for you. Uh
00:03:00
>> it's more fresh.
00:03:01
>> Was it Was it everything that you
00:03:02
thought it would be? Like
00:03:04
>> what's clear so far is I need to upgrade
00:03:07
my namerop game. I mean that that was a
00:03:09
tour to force.
00:03:11
>> But by the way, you were you were in
00:03:13
Davos with
00:03:13
>> J. I was I was there but that um uh
00:03:17
>> tour to force
00:03:20
uh look I I I think you do all this work
00:03:23
and I I think it's really difficult to
00:03:26
to overestimate the amount of garbage
00:03:28
that's involved in in going public. the
00:03:31
number of meetings where you you look on
00:03:35
the the Zoom and there are 130 attendees
00:03:38
and the amount of times you review these
00:03:41
documents and the commas move and and
00:03:43
just no value added. You go there and
00:03:47
you have this enormous event and the
00:03:51
next morning you've sold no more stuff.
00:03:53
your engineering projects have made no
00:03:55
progress since the day you weren't
00:03:57
public. And you go back to work and um
00:04:02
you you have some new constituents that
00:04:04
that you have to to to address and
00:04:07
communicate with, but the core parts of
00:04:10
your business, you have more money in
00:04:11
the bank. Um but not a damn thing
00:04:15
changes in the important parts of your
00:04:18
business. um if you still if you need
00:04:21
new supply or if your relationships with
00:04:23
your vendors are bad, they're still bad.
00:04:25
If they're good, they're still good. And
00:04:27
and so I I think what what we've seen is
00:04:31
um you your employees have a party,
00:04:34
everybody's really excited, and you put
00:04:36
your head back down, you high five, and
00:04:38
you go back to work. Can
00:04:39
>> can I can I just give a little context
00:04:41
and then I want to hear from Will. You
00:04:43
know, if if I can, Andrew, you know, we
00:04:45
were investors in Cerebrus. I was on the
00:04:47
board a year earlier where we were
00:04:49
trying to go public. Um, and you know,
00:04:53
aside from just being a warrior who
00:04:55
weathered a decade worth of storms that
00:04:58
would have taken out any normal human
00:05:00
being, the path to going public for
00:05:03
Cerebras was a particularly challenging
00:05:05
one. One of their investors was the UAE.
00:05:08
So there was questions about CPHAS, you
00:05:12
know, in in the prior under the Biden
00:05:14
administration challenging to get
00:05:15
public. My observation outside looking
00:05:18
in is everything was really hard until
00:05:23
it got really easy like n 9 and a half
00:05:26
years of really hard and then 12 months
00:05:29
you know of of of really easy where
00:05:31
everybody wanted to get in.
00:05:33
>> They priced the IPO at 185 which was up
00:05:36
the range was taken up two times. Okay.
00:05:39
The stock opened at $320 a share I
00:05:43
think. Today it's at 230 bucks a share,
00:05:46
5060 billion dollars in market cap for a
00:05:49
business like you know and andrew is
00:05:52
just one of these people. Let's let's
00:05:53
get back to work and build But my
00:05:55
my just add-on question to that is from
00:05:58
an employee morale perspective like
00:06:00
distraction perspective etc. has the
00:06:04
last 3 weeks you got a lot more capital
00:06:06
you got a lot more profile presumably
00:06:08
it's easier to sell to enterprise
00:06:09
customers today. net net if you were
00:06:12
advising me if I was in a similar
00:06:14
position would you say go public?
00:06:17
>> I I I think the first thing is a lot of
00:06:19
people asked us about how we got the
00:06:22
timing right.
00:06:23
>> Right.
00:06:23
>> And I I think the answer is by getting
00:06:25
it wrong for a decade. I mean that's
00:06:27
really the right way to get timing
00:06:29
right. Um I I think um first I we we
00:06:34
we've been at this for for more than a
00:06:36
decade and and we we brought everybody
00:06:39
who'd been with the company more than
00:06:41
nine years to share and we brought their
00:06:42
families. And first I learned that
00:06:45
engineers owned ties. I didn't actually
00:06:47
know that. Um and they didn't die when
00:06:50
they wore them. And second I was
00:06:53
surprised at how big a deal it was for
00:06:55
them and their family. they they were
00:06:56
really proud in a way that sort of their
00:07:00
parents might have heard of it or uh
00:07:03
that that somehow this was like a a bar
00:07:06
mitzvah or kinera or something.
00:07:08
>> Um
00:07:09
>> and then you you had these sort of the
00:07:11
the children of immigrants, one of our
00:07:14
uh one of our leaders, her father,
00:07:17
Chinese immigrants said, "I thought it
00:07:19
would have happened faster."
00:07:23
>> Right. And but I I think um we are sort
00:07:29
of by nature
00:07:31
uh
00:07:33
sort of in the uh in the trenches
00:07:37
people. And so um we we love solving
00:07:41
hard problems. And so when when we had
00:07:43
this excitement, everybody went and they
00:07:45
were so excited and we had a party and
00:07:47
um I I think it it gave external
00:07:50
validation. And then everybody turned
00:07:52
around and said, "Now what are we now
00:07:54
back to work?" And and so I I I think um
00:07:59
>> and so you you started off kind of bang
00:08:02
right out of the gates. Will you had a
00:08:04
little bit different experience in terms
00:08:05
of you know the entry to the public
00:08:08
markets, but over the last 12 months
00:08:10
your stock has gone from five bucks a
00:08:13
share to 50 bucks a share, some 10x move
00:08:15
in the public markets. So talk us
00:08:17
through the other side of this where you
00:08:19
come public, nobody really notices until
00:08:21
they notice.
00:08:22
>> Well, we were one of the first space
00:08:24
stocks and and I think people just had
00:08:26
no idea what earth is going on in space,
00:08:29
how it was changing everything and they
00:08:31
were just like what the heck is that?
00:08:33
And and but but you know, I have similar
00:08:36
opinions. I mean, in the end, you've
00:08:37
just got to get on with executing the
00:08:39
business. uh the going public gives you
00:08:41
access to liquidity uh for early
00:08:44
shareholders whether that's the employ
00:08:46
early employees or early investors and
00:08:48
that's great it gives you cash for the
00:08:51
company that's great and I do think um
00:08:54
it helps your business as well because
00:08:57
the maturing event gives you more
00:08:59
credibility to various customers and for
00:09:02
us we work with biggest agricultural
00:09:04
customers big governments civil
00:09:06
governments defense and intelligence all
00:09:08
of those sort of actors they want to
00:09:09
know you're going to be around.
00:09:10
>> Exactly.
00:09:11
>> And not going to disappear. I mean, we
00:09:12
have countries that are fully dependent
00:09:14
on us giving them information. They
00:09:16
don't want to just disappear. So, they
00:09:18
really care that we're going to be
00:09:19
around and and being a public company
00:09:21
gives you the kind of force in the world
00:09:23
that people go, "Okay, you're here to
00:09:24
stay and you have access to capital if
00:09:27
you need it and so on, right? It's
00:09:28
legitimizing
00:09:29
>> and you know, you you know where the
00:09:31
stock is at any one day. You know, we're
00:09:33
not focused on that dayto-day. We're
00:09:35
focused on how we build long-term value
00:09:37
for our shareholders, right? And um you
00:09:41
know the market is I think started to
00:09:42
really understand where space is going,
00:09:44
why it is changing the the the world.
00:09:47
You know people forget how space is part
00:09:50
of your everyday life. Every time you
00:09:51
use a phone you're using communications
00:09:54
using satellites or GPS using satellites
00:09:57
or satellite data in some way or another
00:09:59
that's sort of integrated in your lives.
00:10:01
you may not realize it um but it's just
00:10:05
booming now there's there's a
00:10:07
>> and the the story's changed as well
00:10:09
obviously with SpaceX going public but
00:10:11
has the framing of
00:10:14
planet gone from like a data source for
00:10:17
people who need data from space and maps
00:10:18
to hey this is a tool to accomplish
00:10:21
tasks and military like post Andrew's
00:10:25
success like you probably would have
00:10:27
been bucketed into Andrew as a military
00:10:30
tech company. So is that framing what's
00:10:33
driving a lot of
00:10:33
>> I think it's a bit more nuance than
00:10:34
that. I mean firstly for the audience's
00:10:36
benefit what planet does we have
00:10:38
satellites doing earth imaging. We have
00:10:39
the largest earth imaging fleet about
00:10:41
200 satellites. They image the entire
00:10:43
earth every day. So think of it like the
00:10:45
Google uh satellite lay on Google maps
00:10:48
that you can look at except it's today's
00:10:50
date rather than 3 years old and we have
00:10:52
every day going back. So it's a time
00:10:54
series analysis of everything going on
00:10:56
on the earth. That's useful for farmers.
00:10:58
It's useful for energy companies. It's
00:11:01
useful for civil governments, flooding
00:11:03
and fires. It's useful for security
00:11:06
applications like you're getting at. And
00:11:08
it's a wide variety of use cases. Um, I
00:11:11
think that where we're seeing this is
00:11:13
that AI is now enabling the it's
00:11:18
basically reducing the barrier to entry
00:11:20
so that more people can get access to
00:11:22
this, right? And uh, you know, there's a
00:11:24
lot more to say on that, but AI is only
00:11:26
as good as the data it's trained upon.
00:11:28
percentage is military. I'm curious.
00:11:30
Sorry.
00:11:30
>> What percentage of revenue customer bas
00:11:34
um uh security is part of the initial
00:11:36
thing that we said we would do um out of
00:11:39
the gate, but it's true there's a bigger
00:11:41
fraction today than perhaps we would
00:11:43
have guessed. Um but the needs of the
00:11:45
geopolitical situation right now demand
00:11:48
what we're doing. Um you know, just as
00:11:50
an example, what this does is enable
00:11:52
them to see threats around the corner.
00:11:54
>> Yes. and then uh you know give them
00:11:57
weeks or months advanced warning of
00:11:59
things and then that enables them to
00:12:00
more likely do things that stop
00:12:02
conflict. So we believe this is you know
00:12:05
really better for the world.
00:12:06
>> Are you reticent to be perceived as a
00:12:09
military company?
00:12:10
>> Not really. I but I wouldn't say we're
00:12:12
limited to being perceived like that,
00:12:14
right? um we we are helping farmers, we
00:12:16
are helping uh you know uh energy
00:12:19
companies, civil governments, we work
00:12:21
with NASA, we work with what have you
00:12:24
and and so um it's a it's a bigger it's
00:12:27
a bigger play than that but back to the
00:12:28
space uh piece of it what has changed
00:12:31
obviously rocket costs have come down
00:12:35
about four or 5x over the last uh 10
00:12:37
years which has helped tremendously but
00:12:40
a thing that people don't know that is
00:12:42
actually perhaps more important is that
00:12:44
we've had a miniaturization of
00:12:46
satellites. So that the same satellite
00:12:48
that used to cost a billion dollars and
00:12:50
weigh 20 tons now cost a few kg or a few
00:12:54
tens or hundreds of kg
00:12:56
>> and can do just as much stuff if not
00:12:58
more. It's it's the same as the sort of
00:13:00
mainframe computer to desktop revol
00:13:03
computer revolution for space and it's
00:13:05
unlocking just like mainframes to
00:13:07
desktops unlock loads of applications.
00:13:10
This is unlocking loads of applications
00:13:12
and it's so both go in combo the launch
00:13:15
cost coming down and this
00:13:16
>> let's build let's build on this. So I
00:13:18
think I I'd like first you maybe take a
00:13:20
few minutes and then I I want to talk to
00:13:22
Andrew the same question. Both of you
00:13:24
guys are at the foot of what are
00:13:26
probably huge secular trends in
00:13:29
technology. How I would frame this is we
00:13:32
are rebuilding the data processing
00:13:35
infrastructure that has existed on the
00:13:38
earth in the sky.
00:13:40
And first you do the satellites, but I
00:13:43
would love for you to explain
00:13:44
space-based data centers because I think
00:13:46
everybody's hearing about that.
00:13:48
>> Are they really viable? What are they?
00:13:49
How will they work? etc. And then
00:13:51
Andrew, this is the rebirth of silicon.
00:13:54
We're going to find the next version of
00:13:56
Moore's law, which I think is more
00:13:58
timebounded, not transistor density
00:14:00
bounded. We now hear a lot about domain
00:14:03
specific architectures. we hear I mean
00:14:06
your chip was just a complete
00:14:07
transformation in terms of the design
00:14:10
principles that you know like at Grock
00:14:12
we took a very different approach
00:14:14
Nvidia's taken a very different approach
00:14:16
you took a big pizza shaped die and said
00:14:18
it yolo this is it and you were
00:14:20
right just explain where we're going in
00:14:23
silicon so maybe will you start and then
00:14:25
Andrew you start
00:14:26
>> I mean what we're seeing firstly in
00:14:28
space is is all these new applications
00:14:30
based on data and AI so you know we
00:14:33
we're collecting vastly more data about
00:14:35
the planet. And with SpaceX and Starlink
00:14:38
and One Webb, they're they're
00:14:40
transporting far more data around the
00:14:42
planet. As you say, we're sort of
00:14:43
changing um uh the nature of data using
00:14:46
satellites. And that's basically doing
00:14:49
what was once the province of
00:14:50
governments only and giving everyone
00:14:52
else access to uh satellite
00:14:53
capabilities. And that's going to I I
00:14:55
mean I I estimate there's a 75 to$100
00:14:58
billion market just on Earth
00:14:59
observation. this kind of data we
00:15:01
collect and AI on top of that unleashing
00:15:04
all that application. So that's the
00:15:05
near-term thing. Applying large language
00:15:07
models to earth imagery data, unlocking
00:15:10
agriculture, you know, um energy, civil
00:15:14
government applications, permitting, you
00:15:16
name it. This is going to make
00:15:17
everything more efficient. Um and then
00:15:21
where we're going is indeed space is is
00:15:25
we did a study with uh our partners at
00:15:28
Google about eight or nine years ago uh
00:15:30
looking at what are the costs of data
00:15:32
centers on the ground, what are the
00:15:33
costs that it would take to put them in
00:15:35
space and when might it make sense to uh
00:15:38
do it non-aterrestrially.
00:15:40
And we figured out that when launch
00:15:42
costs come down to about $200 to $300 a
00:15:44
kilogram um uh it would be cheaper, just
00:15:48
simply cheaper to put the data centers
00:15:50
in space. Now we're about $1,000 a
00:15:52
kilogram, just over that in right today.
00:15:55
Uh but that's come down about 10x in the
00:15:57
last 10 years. Um on the current
00:16:00
trajectory with Starship in particular,
00:16:02
I would expect it the launch cost come
00:16:04
down there in 2 to 3 years. Elon might
00:16:07
say it's next week, but at least
00:16:10
realistically a couple of years. So
00:16:11
we're not far away from it literally
00:16:13
just being cheaper. Then in addition and
00:16:16
the intuition there that is that helps
00:16:18
people understand that is you would
00:16:20
naturally use solar panels for doing uh
00:16:23
the the data centers are a power
00:16:25
problem. It's a power game and you would
00:16:27
normally use solar panels. That's the
00:16:29
cheapest way to get a watt uh today by
00:16:31
far. But you you don't want intermittent
00:16:34
power. So then you have to have
00:16:35
batteries or you then you have to have
00:16:36
gas or then you have to have nuclear and
00:16:39
then it gets really expensive. Um in
00:16:42
space you can put a solar panel in a
00:16:44
suns synchronous uh dawn dusk orbit
00:16:46
where you're 24/7 looking at the sun. So
00:16:49
you can have a solar panel that collects
00:16:51
and gathers five times more energy per
00:16:55
solar panel than on the ground and you
00:16:57
don't have to have batteries or anything
00:16:58
else. Uh so the infrastructure for
00:17:01
comput in space is literally just solar
00:17:03
panels and the chips and then the RF
00:17:06
signals up and down. So it's actually
00:17:09
really quite simple. It was just a
00:17:10
question of when it's going to be
00:17:11
cheaper to launch all those solar panels
00:17:13
and chips into space than putting on the
00:17:16
ground. And it turns out that's going to
00:17:17
be in a few years. So we're partnering
00:17:19
uh with Google to launch some of their
00:17:21
TPUs into space. We've already launched
00:17:23
Nvidia's uh uh some of Nvidia's GPUs
00:17:27
into space. We're launching Google's
00:17:28
TPUs into space on an early test.
00:17:31
There's lots of technology to figure
00:17:33
out.
00:17:34
>> Let's have a conversation. Um, but it's
00:17:37
an early it's early days, but I think no
00:17:40
question within 10 years most compute
00:17:43
will be putting in space, which to give
00:17:46
you a sense is a lot of money, like
00:17:49
trillions, and um will be bigger than
00:17:52
any of the other space businesses today.
00:17:55
comes Earth imaging. This is why we're
00:17:57
getting into this.
00:17:58
>> Do you believe this? Do you believe
00:18:00
sending data centers to space makes more
00:18:02
sense or is it just the regular
00:18:05
>> Can you have him explain the business
00:18:06
first and then
00:18:07
>> Oh, yeah, of course. Yeah. So, I think
00:18:09
they're, you know, with all due respect,
00:18:11
one or two hard problems still left be
00:18:14
beyond putting putting GPUs in in in
00:18:17
space right now. I I think um we we
00:18:21
we're not super good yet at
00:18:25
building the clusters in space necessary
00:18:28
for the communication
00:18:30
>> between
00:18:30
>> between Exactly. between
00:18:33
>> We're not good at doing it on the
00:18:34
ground.
00:18:34
>> We're not good at doing it on the
00:18:35
ground. We're really not good at doing
00:18:36
it in space. I I think this is an
00:18:38
extraordinarily important and
00:18:39
interesting problem and one we should be
00:18:41
spending money and attacking. I've got
00:18:43
it in a slightly different time frame,
00:18:45
but one that certainly will occur. And
00:18:48
the the hard part is is it is it one of
00:18:51
those problems where uh the last 10% is
00:18:55
80% of the time.
00:18:57
>> Now, self-driving was a problem like
00:18:58
that, right? Where the last 10% proved
00:19:01
to be a decade's worth of work and just
00:19:04
now we're over the hump. And we don't
00:19:06
know yet, but I think the interesting
00:19:08
work they're doing uh at Planet is
00:19:10
really important. And I think the
00:19:12
fundamental driver to experiment to even
00:19:14
get insight into whether I'm right or
00:19:16
not is to get down the cost of launch
00:19:19
vehicles. Then you can start doing
00:19:21
experiments and getting it wrong and
00:19:23
fixing it and figuring it out. And until
00:19:26
then it was unpaid.
00:19:27
>> For the foreseeable future, you're going
00:19:28
to be terrestrial. explain your business
00:19:30
and how you made these critical
00:19:32
decisions that kind of took you on a
00:19:35
different path and you know you versus
00:19:37
Nvidia versus AMD and what you think the
00:19:39
future of AI silicon looks like. I I I I
00:19:42
think there were two parts your your
00:19:44
first question was around sort of the
00:19:48
rise of silicon in general and I I think
00:19:50
what AI did and it's it's rarely sort of
00:19:53
framed this way but it allowed computers
00:19:56
to address a class of problems that
00:19:58
before AI computers were bad at.
00:20:02
We were bad at images for almost the
00:20:05
entire history of compute. We could
00:20:07
store them and that's about it. We were
00:20:10
bad at language. We could store it but
00:20:12
that's about it. We could transform
00:20:15
numbers. We were magical with numbers.
00:20:18
And what AI did starting in about 2015
00:20:22
16 is it opened the door the aperture to
00:20:25
say maybe we could use computers on
00:20:27
images.
00:20:29
All right. Maybe we could find insight
00:20:31
in images. Maybe not only could we store
00:20:34
language but we could generate it. All
00:20:37
right. maybe we could understand it
00:20:40
rather than storing it and regurgitating
00:20:42
it. And what this did is it it opened up
00:20:46
sort of to compute
00:20:48
huge areas that were previously
00:20:50
foreclosed and at the same time we were
00:20:53
adding to those areas. we were taking
00:20:55
vastly more images,
00:20:57
all right, terrestrially
00:21:00
in satellites. And what this did is it
00:21:02
it simultaneously opened up this entire
00:21:05
area and allowed compute to attack it.
00:21:09
And this is what's underpinning both
00:21:10
Nvidia's growth and and sort of all the
00:21:12
growth you're hearing about in in AI
00:21:15
compute is as a as a processor builder
00:21:18
as a hardware builder suddenly our tools
00:21:22
could attack more and different parts of
00:21:25
knowledge
00:21:27
and and that was sort of the first part
00:21:29
to to to answer your question. Now how
00:21:32
you do that there are lots of different
00:21:34
strategies tons of different ways to
00:21:35
skin cats. What we saw in 2015 were
00:21:40
several things. First, we saw that AI
00:21:43
would be an enormous consumer of
00:21:46
compute.
00:21:48
All right. And historically for computer
00:21:50
architects,
00:21:52
new workloads were the opportunity for
00:21:55
share to change,
00:21:57
right? Share changed when the rise of
00:22:00
graphics emerged and you got the
00:22:01
dedicated GPU. That's how Nvidia was
00:22:03
born. Share changed when cell phone
00:22:06
compute emerged and Intel and AMD who
00:22:09
had fabs and the best architects got
00:22:11
zero share and it all moved to ARM.
00:22:14
Right? Share changed in the late '9s
00:22:17
when Nortell and all these companies
00:22:19
we've forgotten about couldn't build
00:22:22
chips and couldn't do uh uh data
00:22:25
networking and and what you got with
00:22:28
Cisco and Juniper and Arist and this
00:22:30
collection of new companies. So we knew
00:22:32
that um this new problem
00:22:37
would present an opportunity for massive
00:22:40
change.
00:22:41
So we saw that we we made two bets. Um
00:22:45
the first was dedicated silicon would be
00:22:47
the answer. The second was it couldn't
00:22:51
look like a GPU.
00:22:54
And our view as computer architects is
00:22:57
if you want to be 20 times better than
00:22:59
somebody, right? Your architecture can't
00:23:01
look like them,
00:23:03
right? It it it can't. They have they
00:23:07
have enjoyed and and eaten all the
00:23:08
lowhanging fruit. So, if you build a
00:23:11
GPU, the odds that you're better than
00:23:12
Nvidia in our view are approximately
00:23:14
zero. That led us to a fundamentally
00:23:16
different architecture. All right. The
00:23:19
hard part here, the hard part is moving
00:23:23
data from memory to compute.
00:23:26
>> This is the fundamental problem in AI.
00:23:30
And we solved it with a way that that
00:23:33
very few others had even attempted,
00:23:35
which was to build a very big chip and
00:23:37
to put memory right next to compute. And
00:23:40
by building a big chip, a chip the size
00:23:42
of a dinner plate, whereas most chips
00:23:44
are the size of a postage stamp, we
00:23:47
could use a different type of memory.
00:23:50
And by using a different type of memory,
00:23:51
a memory that was vastly faster, we
00:23:54
opened up all sorts of opportunity.
00:23:57
So when OpenAI uses us, we're 15 or 18
00:24:00
times faster than a GPU.
00:24:03
That means your answers are delivered
00:24:06
more quickly. It means your engagement
00:24:08
with with the AI is more enjoyable. It
00:24:11
means you can use the AI to solve harder
00:24:13
problems and not wait.
00:24:16
And the way to think about this is sort
00:24:17
of to ask yourself the the counter
00:24:20
factual question. How big is the market
00:24:22
for slow search today?
00:24:25
Right. Right. Is zero. How big is the
00:24:28
market for dialup? It's zero. H how long
00:24:32
do you wait for a website to resolve
00:24:34
before you click away? 3 seconds, 5
00:24:37
seconds. You will not wait for AI. We
00:24:40
have to deliver it to you in a in in
00:24:43
real time. And that's what we saw.
00:24:46
That's what we built.
00:24:47
>> So the panel's on going public. A lot of
00:24:49
LPs in the room. They need to get
00:24:51
liquid. I'm curious about the journey
00:24:53
for your investors. Yeah.
00:24:56
>> Okay. So, well, you guys went public
00:24:57
what year?
00:24:58
>> Uh 2021.
00:24:59
>> 2021 by way of a spa.
00:25:01
>> Correct.
00:25:02
>> Okay. And your VCs were who?
00:25:06
>> Um Drake Professor Jverson was one of
00:25:09
the earliest Capricorn. Um Peter Te's
00:25:12
founders fund. Then we got uh Yuri
00:25:15
Milner's DST.
00:25:17
>> Okay. So your your investors come in,
00:25:19
you go public at 2 billion via a spa.
00:25:24
>> Now we're four years later. Really, it
00:25:26
wasn't until year three or four that 90%
00:25:29
of the value was created. Okay. So did
00:25:33
those early investors capture this 90%
00:25:37
move? Did they stay in it?
00:25:38
>> Most of them did. Yeah, most of them
00:25:40
did, which is really smart on their
00:25:42
part. Obviously, I think they should
00:25:44
hold on even more. Uh, if I didn't think
00:25:47
that, you should.
00:25:48
>> I'm a little bit self-interested.
00:25:50
>> What's interesting about this is
00:25:51
>> No, but really they did. And and I mean,
00:25:53
Google hasn't sold a share. They're our
00:25:55
largest single investor. Um, uh,
00:25:58
Capricorn didn't until very recently.
00:26:00
So, basically, most of them stayed
00:26:02
really well in and they got all of that
00:26:04
upside. And good for them. And the
00:26:07
reason I think this is so important
00:26:09
>> is that there are a lot of LPs in this
00:26:10
room
00:26:11
>> who they're like when a company goes
00:26:12
public give us the shares.
00:26:14
>> No no
00:26:15
>> give us give us the shares. This is a
00:26:16
counter example right? This happened to
00:26:19
us in 10 years ago. We invested
00:26:22
preipo at a billion dollars. We
00:26:24
distributed the shares I think at three
00:26:26
or four billion and then it went to 50
00:26:29
billion over you know the course of the
00:26:30
next 24 months. And we had people who
00:26:32
called us who said well why didn't you
00:26:34
hold on to the shares? And we're like
00:26:35
because you're pounding on us, right, to
00:26:37
distribute to the shares. So you're an
00:26:38
example. Now
00:26:40
>> in your case, Andrew,
00:26:43
>> you have an innovation, right? You're
00:26:45
just now uh public. So all of your
00:26:49
investors are still under lockup like
00:26:51
like Altimeter, but you guys have
00:26:53
innovated with the banks on what I call
00:26:55
a dribble lockup. So over 6 months the
00:26:58
shares can be dribbled out according to
00:27:01
a bunch of performance hurdles which
00:27:03
SpaceX is going to have a very similar
00:27:05
>> when did we start this process of the of
00:27:08
the dribble
00:27:08
>> the dribble
00:27:11
>> concept you started it years ago but
00:27:13
>> yes
00:27:13
>> but u we're all of that with respect to
00:27:16
the lockup I think this is the the most
00:27:19
innovative and I think SpaceX is going
00:27:21
to have a very similar innovation but
00:27:23
Andrew for your investors if you were
00:27:25
talking to ILPs right in the room.
00:27:28
Should Alimter be distributing the
00:27:30
shares when they come out of lock? How
00:27:33
do you think about you know your VCs
00:27:36
holding on to the shares kind of post
00:27:38
lock?
00:27:39
>> Well, I I I think historically more
00:27:40
money's made after IPO than before.
00:27:43
>> Yeah.
00:27:43
>> I I I think every single study shows
00:27:46
that uh there is more money to be made
00:27:50
both in percentage and in in what we
00:27:52
care about which is absolute.
00:27:54
>> Yes. and and and so uh the amount of
00:27:57
money that it's possible to put to work
00:27:59
in most venture companies is very
00:28:01
modest. I mean there are two or three or
00:28:02
five outliers but for the most part you
00:28:05
can only put a relatively little bit of
00:28:06
money to work. Um by the time we get
00:28:09
public there's a lot more money there if
00:28:11
things are going well and the
00:28:12
opportunity to make vastly more is after
00:28:15
IPO not before. Hey, if I could just add
00:28:18
on that, one interesting question is
00:28:20
what's going to happen with SpaceX on
00:28:22
this because you know a lot of the value
00:28:25
is is in the future, right? But I mean
00:28:28
most of the big tech companies went
00:28:30
public at a few billion,
00:28:32
>> not a few trillion. Like there's a lot
00:28:34
of zeros in between those, right? And
00:28:36
you got all this upside afterwards. Now,
00:28:39
for the equivalent liftoff, SpaceX would
00:28:41
have to be aiming at quadrillion
00:28:44
valuations. Now, I know Elon has those
00:28:47
sort of ambitions, but you really have
00:28:48
to believe in that uh to get, you know,
00:28:51
>> this is this is kind of the point I'm
00:28:53
getting to, right? We have three mega
00:28:55
IPOs, you know, we keep talking about
00:28:58
that are multi- trillion.
00:28:59
>> Yeah.
00:29:00
>> All of that value acrewed to private
00:29:02
market investors.
00:29:03
>> Planet Labs is a great example of
00:29:07
venture capital in the public markets
00:29:09
where the 10X has occurred in the public
00:29:12
markets. We're all advocates of these
00:29:14
companies coming public sooner. Had
00:29:16
Andrew had his way, he would have been
00:29:18
public 18 months ago, probably at $10
00:29:21
billion rather than $50 billion. And
00:29:23
that 5x over the course of the last two
00:29:26
years would have gone to public market
00:29:27
investors. So go ahead.
00:29:29
>> Way better to be lucky than good.
00:29:31
>> Yeah.
00:29:33
>> So, so I think that I hear a lot of
00:29:36
people thinking that anthropic open AI
00:29:38
and SpaceX are the new normal. I
00:29:41
actually think the public markets may be
00:29:43
shifting back in this direction and a
00:29:45
lot of the companies in our portfolios
00:29:47
are now thinking about going public at a
00:29:49
billion or three billion or 5 billion.
00:29:52
>> We had this period of a decade where
00:29:54
Andrees was really pushing stay private
00:29:56
forever and I see the pendulum swinging
00:29:59
back to companies are like man I want to
00:30:02
be like Planet Labs and get public,
00:30:04
>> right? and and have to play in the big
00:30:06
leagues and do it in the public markets
00:30:08
like
00:30:09
>> here's what I'll say maybe just like to
00:30:10
the two of you guys both of you guys
00:30:12
have had enormous pressure because
00:30:14
there's visible competition that's
00:30:16
always sort of in your periphery
00:30:19
but I do think that getting public
00:30:21
sooner having the scrutiny of public
00:30:24
markets having the scrutiny of having to
00:30:25
deliver sharpens the focus it steel
00:30:29
sharpens steel iron sharpens iron
00:30:31
>> and I think innovation tends to get
00:30:33
better And so the idea that you allow
00:30:36
everybody to participate but you also
00:30:38
put yourself in the spotlight to me is
00:30:40
where great things happen.
00:30:41
>> I agree.
00:30:42
>> Um and so anyways I just wanted to say
00:30:44
to both of you um just as we wrap you
00:30:48
guys are an incredible testament to
00:30:49
entrepreneurship both of you. I mean
00:30:51
we've been talking literally since day
00:30:53
one me and Andrew because we went in
00:30:55
different paths and then we kind of
00:30:57
recon converged and then Will same with
00:30:59
you.
00:30:59
>> I'm happy it worked out for you Jimoth.
00:31:01
>> Well it's worked out for both of us so
00:31:02
it's fine. Um you guys are incredible
00:31:05
testament to entrepreneurship and uh I
00:31:07
just want to say thank you for
00:31:08
everything you guys are doing and the
00:31:09
next few years are going to be really
00:31:10
spicy.
00:31:11
>> Yeah, if I could just spend a 30 seconds
00:31:13
on the next few years because I think
00:31:14
it's going to be so exciting with um as
00:31:17
I mentioned AI and space merging
00:31:19
together. We're going to see a take off
00:31:20
our applications. I like to say like all
00:31:23
the cool stuff that we're doing on the
00:31:25
with LLMs now is really based on just
00:31:28
the text of the internet being absorbed
00:31:30
into these models which is incredibly
00:31:32
powerful already but they don't know
00:31:35
about the real world. I call them
00:31:36
blind to you know they don't know about
00:31:38
that farm field that flood that security
00:31:40
situation around the corner. If you give
00:31:43
them real world data, then they can
00:31:45
answer real world problems. And that's
00:31:47
going to open up gazillions of
00:31:49
applications for these uh AI models. I
00:31:52
call them instead of having large
00:31:53
language models, large earth models or
00:31:56
uh instead of AI, planetary intelligence
00:31:59
where you have planetary sensing systems
00:32:00
in space, planetary compute systems in
00:32:03
space. And we can disagree or agree on
00:32:06
on exact time frame, but I think it's
00:32:07
going to happen. and and then that's
00:32:10
going to enable a huge economy. So, it's
00:32:13
an exciting time in the next few years.
00:32:15
>> Will Andrew, thank you guys very much.
00:32:17
Well done.
00:32:18
>> Thanks, guys. Okay.
00:32:22
Thanks, buddy. Appreciate you. Really
00:32:24
appreciate it. Great seeing you,
00:32:25
brother. Congrats.
