00:00:00
Guys, this is one of the most amazing entrepreneurs that you're going to meet. Jim Leinsky, this founder and CEO of MP
00:00:06
Materials. Thanks. Good to be here. Jason, how are you?
00:00:11
So, let me let me set this up. Um, Jim was a hedge fund guy running a
00:00:16
pretty successful hedge fund. Uh, and he ended up basically investing in
00:00:22
something called Molly Corp, which went out of business. Yep. Yeah. And you did this incredible thing, which
00:00:27
is you said, "You know what? screw this. You essentially shuttered the fund, took
00:00:32
over the company, and fast forward many years later, you are the uh largest and
00:00:38
only, I think, supplier and refiner of rare earth materials and maker of
00:00:43
magnets inside the United States. We're 100% of the American industry. 100% of the American industry. You
00:00:49
just did two really incredible things actually in the last couple weeks. One was you announced an enormous public
00:00:55
private partnership with the DoD $400 million etc. And then the second is
00:01:01
you uh announced a really big deal with Apple. Yes. Okay. So take us take a huge step back.
00:01:08
Talk to us why rare earths matter. Tell us about the supply chain for AI. Tell us why you're doing this.
00:01:14
Rare earth magnets are really the feed stock to physical AI. Um you know robots, drones, everything we're talking
00:01:20
about today, the biggest industry in the world to come. Um, essentially electrified motion requires rare earth
00:01:26
magnets. Um, so you mentioned the predecessor went bankrupt. Um, there
00:01:32
there was a feeling when I when I took over this site with my co-founder and this is this goes back to 2015. Uh,
00:01:38
there was where where is the site? This oh it's in Mountain Pass, California. So if you you'll be familiar if you take a 45minute drive from the
00:01:44
Las Vegas strip uh just over the border in California uh is the site. You actually can see it from the road. Uh,
00:01:50
and it's actually the really the best rare earth or body in the world. Um, the thing about rare earths is that when you
00:01:56
mine them, you also have to refine them. And it's really expensive and difficult to refine them. It's really a specialty
00:02:02
chemical process. And so it's really a think of it as a multi-billion dollar refinery uh that you need to have just
00:02:08
to separate them. And then once you separate them, you need to turn them into metal and then a magnet. And so
00:02:13
there's a multiple layers of this stream to get this supply chain. And of course, you could have all the rare earths in
00:02:19
the world, but if you don't make the magnets, you're sending it to China. Or you could have all of the magnetic capability in the world, but if you
00:02:26
don't have the rare earths, you're relying on China. And so our vision from day one going back to we we originally
00:02:32
bought these assets out of bankruptcy officially. It was a two-year battle, took it out in 2017, and there was a
00:02:39
perception that we just couldn't compete against China. And what we discovered actually is we could. It's a world-class
00:02:44
site, but we had to we had to reorganize the process flow and then we had to make investments uh to move downstream. So
00:02:51
over the last eight years, we invested about a billion dollars. Um Jim, as you know, we took the company public in
00:02:57
2020. We built out the refining capability. Uh and then uh about four
00:03:02
years ago, we announced we were going to build a magnetics factory in Texas. We built that factory. We have GM as a
00:03:08
foundational customer. We're now producing autograde magnets uh G to GM spec and we'll be ramping up uh sales to
00:03:14
GM at the end of this year in in magnets. And then Chimath, you referenced a couple it's been a busy few
00:03:21
months for us. Um we announced a a pretty transformative public private
00:03:26
partnership with the Department of Defense. DoD is um there's really three pillars to this deal. uh DoD is becoming
00:03:33
our largest economic uh investor um as well as they're going to provide
00:03:39
uh a price floor for our commodity so that the Chinese sort of Chinese mercantalism we can get into that uh
00:03:46
won't take the price of the commodity below the cost of production and then as a result of the DoD investment we're
00:03:52
going to accelerate the buildout of the magnetic supply chain so we're expanding our facility in Texas for Apple I'll
00:03:58
talk about that in a second but we're then going to build a 10x facility to 10x our capacity with DoD as our 100%
00:04:05
offtake partner um uh customer and business partner because we'll be we'll be splitting profits 50/50 with D.
00:04:12
So to to just translate this, it's not a handout from the government. They didn't gift you $400 million. They invested in
00:04:18
your company. They have warrants. They have equity. Yeah. So they invested they they both are an an owner. Uh they also are an
00:04:25
upside participant in our commodity to the extent that the prices uh uh take off. And then they're also 100% offtake
00:04:32
customer. We have a guaranteed level of profits to want to build out this facility, but above a certain threshold,
00:04:37
they're a 50/50 economic participant. So there's really you, the taxpayer. Yeah. So this is a and maybe I'll say
00:04:44
something wild here. This is a true win-win. Obviously great for MP shareholders. um great from a national
00:04:50
security and commercial national security standpoint because we're going to have enough magnets to provide you
00:04:55
know real certainty in the supply chain for the physical AI revolution and and other industries. Um but it would not
00:05:01
surprise me if when we 5 years from now hopefully we'll do this conference and Chimath you'll say to me Jim you know I
00:05:08
remember that deal that was the first of its kind that you did with DoD and the government made money on you u the
00:05:15
taxpayer made money on doing this and I'll say yeah I actually think that that's going to be the outcome um
00:05:21
because there's sort of an element of mutually assured economic destruction if
00:05:26
the Chinese believe that America has national champions do then there's no point in subsidizing the rest of the
00:05:32
world. And so I think you can start to see prices uh normalize for some of these things and free up our ability to
00:05:38
invest and expand. Why go to the government for this investment as opposed to the private markets? Yeah. Well, because it's that issue that
00:05:46
this is sort of one of those, you know, obviously you have to go back to World War II or the railroad boom where you
00:05:51
really need government and credit. I mean, this administration uh did
00:05:56
something, you know, totally unique that Why do you need the government? Merkantalism. Straight up mercantalism.
00:06:02
Because the Chinese will sell magnets for below the cost of raw materials. And so every time there's somebody who makes
00:06:08
progress, they can put them out of business overnight. And so it's difficult to want to make the investment. And so frankly
00:06:15
with the Department of Defense, the scale that they wanted us to build on the time frame that they wanted us to build. We there was no way we were going
00:06:22
to make that commitment. Um we're fiduciaries, right? We have shareholders. There's no way we're going to make that commitment without
00:06:27
certainty that we would not be destroyed by mercantalism and that we would have a customer for the magnets.
00:06:32
How big of an industry is physical AI? Meaning, we see the robots, we're told the robots
00:06:39
are coming, we we're told there's going to be billions of them. Are they actually being deployed at the
00:06:46
scale and at the pace that we that we've been told? Yeah. Well, I think that that is a question for there's much smarter guests
00:06:53
on this. for the rest I'll give a plug the rest of the day obviously you have the you know the best of the best providing that feed stock um I will say
00:07:00
that I think one of the big drivers of our deal was the as we've seen um in
00:07:06
Ukraine and the Middle East the future of warfare is physical AI right robots and drones and I think irrespective of
00:07:13
the scale that robotics is ultimately going to be and and certainly the commercial business will be bigger than
00:07:18
than you know the the defense needs but just from a defense standpoint this is this is an a really important supply
00:07:24
chain that we must have, right? We there's we can't be funding cutting edge drone and robotics companies and then
00:07:31
say, "Okay, but we're going to buy those magnets from China." That makes no sense. Do we have talent capacity or do we have a talent shortage? Secretary Bergam gave
00:07:38
me a stat which was pretty shocking to me that we only graduate 200 people a year in the United States in mining,
00:07:45
which is orders of magnitude different than China. What do we need to do to be competitive to build the industry here?
00:07:52
It's a great It's a great question. I think Jason, I think about this question a lot because
00:07:57
what's that? Oh my god. Dave, sorry. No, it's all good. I'm a huge fan of the pod and I just embarrassed myself all the time. It's
00:08:05
it's the only token whites and you know I'm a fan of the pod since day one and I'm totally embarrassed
00:08:11
myself. That's only one correction messing with you. Nor am I messing with you? Was this intentional? Um, so
00:08:18
huge fan of the pod. Yeah, huge fan of the pod. Who are you again? I'll take a selfie later.
00:08:23
And I'm not the AIAR. Go ahead. Um, so we have we have uh 850 employees
00:08:30
today at MP. We're we're going to hire when we include what we're building out for Apple uh coupled with what we're
00:08:37
going to build with DoD. We're going to need a couple thousand more people easily, not to mention the construction
00:08:43
jobs. So this is um this is a key existential question for all of us as we
00:08:48
build out this is is where are we going to get the talent? I think what we have found um you know at Mountain Pass and
00:08:54
we we hire it all electricians um maintenance you know operators is you
00:08:59
get people in you train them and then obviously you give people a career and so we we've been training a lot of
00:09:05
people and it's a little bit more painstaking but there's there's absolutely talent out there people are hungry to do it. Why do you think it's
00:09:11
been so hard to establish that that idea like meaning you find it straightforward to find good hardworking people to get
00:09:18
into these jobs but the se the thought is always that wow these jobs are not
00:09:23
desirable but they really are desirable by many people. Yeah, absolutely. I mean we you know our
00:09:28
median wage is now pushing $100,000 a year. Um and there's you know relative
00:09:35
to some of the opportunity set I mean uh these are great these are great jobs. And what are the salaries?
00:09:40
What's that? What's the starting salary? Just curious. So, it really depends on the job function because you know there's
00:09:46
there's I mean I I think the easiest way to think about it is you can you can certainly as an operator make close to
00:09:52
$100,000 a year with us because by the way um everybody's an owner. We have an
00:09:57
owner operator culture. Everyone got stock when we went public in 2020. But somebody coming out of high school they can make 40 50 60k or more. Yeah.
00:10:04
Or it depends. Are you if we we can't find enough electricians. We can't find enough maintenance workers. A
00:10:11
maintenance worker can uh an electrician they can make six figures today. Tell us you said earlier that you
00:10:17
suspect 5 years from now we're going to look back and this deal with the DoD was a blueprint. Yeah.
00:10:23
give us other areas of either physical AI or software AI or other markets where
00:10:29
you think these public private partnerships are really necessary to
00:10:34
embellish US supremacy. Yeah. Um there are some major categories obviously we've all heard about ship
00:10:40
building and advanced pharmaceutical ingredients. I mean I think I think those are important ones and then there
00:10:46
are a number of sort of niche areas like industrial diamonds that are important
00:10:51
for quantum computing and some of these things that you never would have thought of um where there it's a vertical where
00:10:58
there might not be a market large enough to to need five players but a a good public private partnership can just
00:11:04
solve that problem and then there's some other verticals in critical minerals straightforward for you to find the right person within the Trump
00:11:10
administration that said of course this is obvious. Let's sit down and hash this out like Yeah. Well, and I I think that's you our
00:11:17
our particular deal was led by DoD and so I have to say that the Pentagon leadership is extraordinary. Um and uh
00:11:24
you know this was a mandate though directly from the president to solve this problem. And so again they deserve
00:11:29
a lot of credit for being you know bold here. And and to be clear cuz I you know this story is not out there. Our process
00:11:36
this was I've never worked so hard in my life. I mean this was this was like a true aggressive private equity style
00:11:43
investment and negotiation. The transaction documents are public. You can look at that.
00:11:48
So that's you saying that they're tough. Yeah. They are this was this was as tough as it gets. Tougher than you know
00:11:54
think of any you know blue chip private equity or or uh distress lender type
00:12:00
negotiation. That's what this was. And the key thing um was they were going to
00:12:06
hold our feet to the fire to execute on an aggressive timeline. They were going to hold our feet to the fire on the
00:12:11
cost. And so we're exposed if we get the costs wrong. You know, we we're making this investment. And and so the key
00:12:19
piece of this and which I think is a good model for all of us and is actually will be really effective is the goal I I
00:12:27
don't speak for them ask them but I think their goal was we're going to take the things off the table that you can't
00:12:32
control mercantalism you know certain customer issues um we're going to be
00:12:38
held to account for the things that we can control our ability to execute our ability to um execute on a good timeline
00:12:45
and our ability to control costs. So when we think about a lot of these historically the government sort of
00:12:51
investing in a sector and quote picking a winner usually there's sort of money
00:12:56
given to someone and it's sort of public risk private upside right this is not
00:13:01
that this is private risk public risk public upside private
00:13:08
upsideing it's a true shared win-winwin and again like I said whole me to these
00:13:14
words I I hope uh I I I hope I'm right on this, but I think the uh to the credit to the Trump administration, I
00:13:20
think they will make money on this and have solved the national security. All right, we appreciate you coming, Steve. Yeah. Oh, thanks.
00:13:25
Thanks so much. Thanks, brother. Yeah, it's great. All right. Take care, Steve.
00:13:30
Thanks, Jacob. I appreciate it. Yeah. Okay. Hi, Lisa. Lisa, it's a pleasure.
00:13:36
Hi. Nice to meet you. Hi. Hi. Hi. Hi. Hi. Well, thanks so much for being here with us today. Um, we don't have a lot of
00:13:42
time, so we want to get into it. In April, it was announced that you achieved your first silicon output at the TSMC facility in Arizona on that
00:13:48
2nometer line. Um, this administration and the private sector have talked a lot about onshoring semiconductor
00:13:55
manufacturing. Would love your thoughts of the on the ground experience in Arizona? How's it going? What's not
00:14:01
going well? What does America need to do to get this right? Well, absolutely. First of all, it's a pleasure to be here. Um, love the theme.
00:14:07
I think we're all super excited about winning the US um, AI race. And I thought if we're going to talk about
00:14:12
chips, David, I should actually bring one. Oh, awesome. That's okay. Little bit of showand tell. So, this is our latest generation um AI chip. It's
00:14:21
our MI355 chip. 185 billion transistors. Takes about 9 months to build. Lots of
00:14:27
technology on it. If I just nimeter, this is um 3 nanometer and uh 6
00:14:32
nanometer. So, lots of different uh I'll be putting this on eBay later. I'm going to take it with me when I'm
00:14:38
How is that? Thank you. Uh but look to answer your question I think look these AI chips are
00:14:43
extremely extremely complex. Um they have uh so much technology on it. Uh we're super excited about the progress
00:14:50
in US manufacturing. I would say you know 12 months ago people weren't sure that we could do leading edge
00:14:56
manufacturing um in the United States. Uh we've been very early in Arizona with TSMC and we did get our first uh chips
00:15:04
out. um they're actually 4 nanometer but what we see from it is where there's a will there's a way and I think all of
00:15:11
the uh conversation about onuring manufacturing has been you know super good for the semiconductor industry and
00:15:17
and you know frankly for you know all of us in um semiconductors uh we're in such
00:15:23
an interesting place because you know chips are so essential to ensuring that we are able to win the AI race that um
00:15:30
you know we want to make sure that there's a lot of geographic diversity and capability there but the reports out where that TSMC
00:15:35
couldn't get good qualified trained employees. They had to bring folks over.
00:15:40
Is that accurate? And like again, like if we're going to scale, like what's the the order of magnitude we're going from here? Is it 10x, 100x? And how are we
00:15:47
going to build a workforce to support this industry, which is a completely new industry for America? And Lisa, you have permission to speak
00:15:53
freely. Yeah. The best way to say it is no matter when you start something new, it it's going
00:15:59
to take work, right? It's it's going to be hard. So sure in the beginning there there were some issues of you know the
00:16:05
the you know TSMC has like a formula for how they build and they just you know
00:16:10
rinse and repeat and they've learned how to do that well in Taiwan. So they had to learn how to do it well in the United
00:16:16
States. But I have to tell you, we've been super impressed with the progress. And you know, if we look at the the main
00:16:22
thing that we look at is, you know, yields and just how many chips do we get out on a given wafer? And I would say
00:16:28
it's equivalent between what we get in Taiwan and what about cost in Arizona because it's unrealistic to think the United States
00:16:34
could compete on cost. Am I correct? We're going to pay a little bit more. Give us the ballpark. 50% more, 20%.
00:16:41
Not not 50% more. I mean look it it it it's going to be you know more than 5% but you know let's call it less than
00:16:48
20%. So low low low double let's say low double digits and how does that impact the business if
00:16:54
at all uh in terms of competition globally? Well, I think the important thing is I mean just think about like
00:17:00
everybody wants a GPU, right? Like if you look across the industry, you really
00:17:06
say, you know, the people who are going to win in AI want to have as much compute in their foundation as possible
00:17:12
and they want assurance of supply. We want to be able to supply this no matter what happens. And so if you put that in
00:17:17
context, you know, the fact that you're not going for the the lowest cost, you know, every minute of the day, um, is
00:17:24
okay. It's okay. Like obviously we're not going to build um not everything needs to be in the most advanced
00:17:30
technologies and so we have a very geographically diverse supply chain. You know I think Taiwan continues to be important um in that view but the the
00:17:38
focus um from this administration on getting uh onfirm manufacturing in a big
00:17:44
way not in a small way I think is is very good for the country. How much time do we have? If there was a disruption,
00:17:50
for whatever reason, we can come up with hypotheticals in Taiwan and we were unable to get chips from those
00:17:56
factories, what would that look like globally? Yeah, you have to look across um the supply chain, but you know, from a
00:18:04
structure standpoint, we all want to keep reserves for, you know, those those times. Uh but it's months, it's not
00:18:10
years. Lisa, there was uh two really interesting posts over the last couple of days. One was from Elon where he said
00:18:17
in 5 years he projected 50 million H100 equivalents just for XAI and the second
00:18:23
was Sam Alman they signed a deal for a four I think gawatt data center 30
00:18:28
billion a year with Oracle that just pretends an enormous amount of chips
00:18:33
that are necessary and power and if you forecast that how do we
00:18:39
actually meet all of that what needs to happen that's not happening today inside of the United States to actually do that
00:18:45
Yeah, it's it's a great great point. I mean, that's that's what we're seeing. We're seeing this uh incredibly large
00:18:51
demand uh for AI and they're coming from, you know, Sam and Elon are certainly uh the you know, the the
00:18:57
leaders, a couple of the leaders. Uh there's there's a lot of demand elsewhere, too. I mean, if you think about it, nations want their own AI. So,
00:19:04
there's uh very high demand. We're we're imagining that just the accelerator market, so the chips for these um you
00:19:10
know, AI large computing systems will be like, you know, over $500 billion in a couple of years. So very high growth.
00:19:17
And when you say, you know, what do we need to do? Um it's the entire ecosystem needs to scale up. So we need to scale
00:19:24
up um certainly what we're doing in in chip design is trying to get chips out as fast as possible. Uh but we're also
00:19:31
scaling up the entire manufacturing ecosystem. And you know, as I said, I don't I think the US is going to be a
00:19:37
huge piece of it. So, it's not just about the silicon. There's all of the various other pieces of the ecosystem
00:19:43
that have to come to the US. And and I think look, I think today's um AI action plan is actually a really, you know,
00:19:50
excellent blueprint. And how do you see the market evolving in these next five or six years? Is it
00:19:55
there's a standard set of chips for training, a standard set for inference or do you just see an explosion like a Cambrian explosion of different AS6,
00:20:02
different designs, different use cases? Yeah, I I I like that question because I I am a believer in there will be
00:20:08
diversity of chips. Uh and the reason is there's so many use cases, right? If you think about use cases from, you know,
00:20:15
whether you're talking about science or manufacturing or design or backend or, you know, frankly personal AI, I think
00:20:20
we're going to see AI in everything that we do. you know certainly in your phones in your PCs and so you have all these pieces you're going to have different
00:20:27
types of chips um that do that uh you know certainly the for the largest systems uh we tend to believe that uh
00:20:35
you know you need the most compute you can get and so you know GPUs are there but lots of AS6 are um in the in the
00:20:41
process and you know we'll see a variety of different chips. You opened up a really interesting line of questioning
00:20:47
there. When mainframes uh were so expensive and then eventually wound up having PCs that were more expensive on
00:20:54
their desktop, you alluded to AI being run locally. Yes. When would we have a local computer, a
00:21:01
laptop, a desktop computer that would have the power we're seeing to run some of these LLM models in your mind? And do
00:21:08
you see that as a specific market to go after? I look I definitely see the um the idea that AI will be at every part
00:21:15
of our ecosystem um is a uh is a real thing. I think that's one of the advantages. If you think about the power
00:21:22
of AI um you want it everywhere and you want it across all different applications. And I think when you think
00:21:28
about PCs today, we're putting significant amount of AI um in them to run local models. And why would you want
00:21:34
that? It's like well maybe I don't want all my personal data, you know, all over the place. On on that point, can you make a prediction on when the the market
00:21:42
for physical AI chips is greater than the market for chips and data centers?
00:21:47
Uh that's a great question. I'm a big believer in physical AI. I still think it's let's call it five years.
00:21:54
You think five years? Is that that fast? It's at least five years. So you're saying five plus.
00:21:59
Five plus. Yes. Okay. But that but that is ultimately the biggest end market. Do you think is it phys you think physical AI becomes the
00:22:05
biggest end market? I think it becomes a significant end market. Um I think you look at chips in data centers and you
00:22:11
look at chips at the edge, they're also, you know, significant markets. When you look at the most cutting edge
00:22:18
techniques today, EV lithography, all of this whole stuff to make chips.
00:22:24
One of the things that's observable is we're only as good as what humans have been able to invent.
00:22:30
And I often ask the recursive question, what happens when the AI is able to invent its own method of manufacturing,
00:22:36
different materials, different material sciences, different approaches that we may not necessarily understand? Is any
00:22:42
of that R&D happening whether at AMD or in other places? Like how are we trying to get beyond the physical limits of
00:22:50
electrons shunting across a junction? I I think this idea that the um AI can
00:22:57
be extremely smart and extremely capable like we think about how AI can design the future chips and it will design
00:23:04
pieces of it but there's still a a creativity of bringing it all together
00:23:09
that I think humans are still absolutely at the center of that. So I don't necessarily see the AI designing our
00:23:16
next generation GPU, right? Right. Uh but I do see it helping us design the next generation GPU much faster and more reliably. So
00:23:24
you talked about the need to like reshore more parts of like the you know ecosystem. Obviously you guys are there's a world class chip design the
00:23:30
fabs are getting reshored but how do you think about things like lithography like does that need to be you know sort of reshored or like does ASML need to start
00:23:36
building you know sort of machines in the United States or is it okay to have that type of you know supply chain risk on an ally? Look, I think we're going to
00:23:44
we we have to accept the fact that it's a global supply chain. Like even if you were to reshore, you know, X number of
00:23:51
components, you would still have Y components that are across the world. I think it's important for us to have our
00:23:57
allies together. So that's a a key piece of the conversation and ensuring that
00:24:02
you know we have access to the latest generation technologies and uh you know that that is you know something that we
00:24:08
protect given our intellectual property and going going to first principles and asking you the open-ended question what
00:24:15
should be done about American education. I'm going to ask this a lot today.
00:24:20
Assume there's no college, high school, nothing. You arrive in America, the situation is what it is today. What do
00:24:26
you do? How do you build an education system to prepare the next generation for the evolving workforce?
00:24:31
Yeah, I'm probably a little bit biased as maybe some of your uh guests are today. Uh I'm a big believer in, you
00:24:37
know, science and technology background as being, you know, sort of the STEM background is, you know, so helpful when
00:24:43
we think about the future workforce. And the earlier we can get into the um you
00:24:50
know sort of the process I think the better. So some of the work that's being done to kind of revitalize the uh
00:24:56
curriculum I think is is pretty important in the um sort of the next generation workforce. And one of the
00:25:01
things when I think about you know how we win in AI like there there's so many aspects of it but ensuring that you know
00:25:07
America is the best place for AI talent is is also you know a key piece of that. So kind of inspiring people when they're
00:25:14
young to uh really you know study you know science. Lisa when you um go to bed at night and
00:25:20
you think about the best case scenario for this technology and this trajectory we're on which is accelerating and
00:25:27
you're enabling. What could the world look like in 10 years? We let's say it's pretty obvious we're hitting artificial
00:25:33
general intelligence at this moment. I think we'd all agree we're starting to see that. But super intelligence can't be far behind that. I assume you agree
00:25:40
with that. um assume we hit that super intelligence what will the world look like in 10 years in the most optimistic
00:25:47
scenario if we do this right well I think the exciting part about it and you know I can say this ex very
00:25:53
sincerely I mean this is the most um transformational technology sort of in our lifetimes I mean that's the way we
00:25:59
should think about orders of magnitude orders of magnitude and the reason is it's not just going after one aspect
00:26:07
right you can actually take AI and make science better you can take AI and make medicine better. You could take AI and
00:26:14
make manufacturing better. You can take AI and make every aspect of your business better. And so, you know, in my
00:26:20
mind, 10 years from now, we'd like to believe that we are um, you know, really leveraging it to solve some of the
00:26:26
world's most important problems. I I like to say like, you know, AMDers get up in the morning and they say, you know, how can I use technology to solve
00:26:33
some of the most important challenges um in the world? And, you know, AI is really our mechanism for doing that. I
00:26:40
have a business strategy question. If we went back 20 years and we wrote the tale of three companies, Nvidia, AMD, Intel,
00:26:48
and then you fast forwarded 20 years, two have just absolutely thrived
00:26:54
and one has not. And if you had made the bet back then, it would have been very inconclusive that you would have picked
00:27:00
Nvidia and AMD. And if anything, there was a an amount of inherent belief that
00:27:07
Intel had just figured it all out. Can you just tell us sort of like the lessons learned of why you've thrived
00:27:15
and maybe what you take away from their journey that you make sure AMD doesn't play out
00:27:22
well you know as a CEO we have to be paranoid every single day right so we don't rely on the past but I think there
00:27:27
are lessons of the past and I I think the probably the most important lesson that I can uh say for technology is you
00:27:34
have to shoot ahead of the duck like you have to be thinking what is the most like your question Jason great question
00:27:40
we think about that all the time how do we shoot ahead of the duck and you know you have things that change you know
00:27:47
technology is a beautiful place because you see big inflection points like 5 years ago AI was around but we wouldn't
00:27:54
be able to gather this audience to talk about AI because people be like who cares but the fact is you had to invest
00:27:59
many many years ago to be where we are today and I think you know I I like to say that you know you will you will be
00:28:06
able to judge whether we've done a good job or not by how we perform 5 years from now. Like the decisions we're
00:28:11
making will take, you know, five plus years to play out. Uh but that's the key thing in tech like nothing is fast but
00:28:18
hopefully it's quite lasting in what it do you think is happening in countries not in the United States like what do you think is happening in chip design
00:28:24
and all of these mech capabilities in China and other places right now? We should believe that it's super super
00:28:31
competitive. I mean, at the end of the day, I think the world has recognized that um that semiconductors and chips
00:28:38
are essential. They're essential to national economies. They're essential to national security. And so, assume that
00:28:45
everyone's investing. Um I'd like to believe that we have a great head start, you know, because of the innovation
00:28:51
pipeline, because of, you know, the great uh companies that we have here, but we should not be, you know, confused
00:28:57
that everybody's investing and we need to keep our up our investments as well. And I I think that's why, you know, this
00:29:03
whole idea of any one company can provide every solution that's necessary just isn't the case, right? I I I love
00:29:09
the idea of open ecosystems of uh you know uh companies collaborating of
00:29:15
collaboration across the ecosystem. So hardware, software systems um you know
00:29:21
collaboration across public private partnerships because that's what it's going to take like for us to win. We have to be, you know, front-facing and
00:29:28
realizing that bringing, you know, the the countries that win bring all of the smartest people and the best, you know,
00:29:35
capabilities together and let them go as fast as they possibly can. Great. So, thank you for being with us.
00:29:41
Wonderful afternoon. Been great. Appreciate it. Thank you. Thank you. Great. Pleasure to meet you. Thank you.
00:29:47
I'm Chase Locker, the uh co-founder and CEO of Crusoe, and I'm here to talk to you about the AI industrial revolution.
00:29:55
Um, I'm going to start with a quote and it's from Warren Buffett in his 2020 shareholder letter shareholder letter to
00:30:02
investors and he said, "In its brief 232 years of existence, there has been no
00:30:08
incubator for unleashing human potential like America. Despite some severe interruptions, our country's economic
00:30:15
progress has been breathtaking. Our unwavering conclusion, never bet against America."
00:30:23
Buffett's words were true then and as we enter this global race for technological
00:30:28
dominance of a artificial intelligence uh they ring even truer today. American
00:30:34
dynamism has always prevailed and it will continue to do so. So in in in sort
00:30:39
of the uh history of of of really what made America great is uh you know we
00:30:46
live in a nation that's the freest nation in the world and we have uh we
00:30:51
are just as rich in land and resources as we are in human ambition uh to drive
00:30:56
progress. And one of the things that's fundamentally enabled that progress to happen and that ambition to be unleashed
00:31:03
is the leading investments that we've made in infrastructure over the course of his lifetime. Uh Warren Buffett got
00:31:09
to witness uh investments in in power in transportation and in uh in power in
00:31:17
transportation and in natural resources to uh to enable people to go pursue their dreams and live a better life.
00:31:27
Let's see. There we go. Now, in 2025, we stand at
00:31:32
a, you know, the start of a new era of infrastructure, the infrastructure of intelligence, and it's driving the
00:31:39
biggest capital investment in human history. Uh, this investment is being led by the hyperscalers uh who are
00:31:45
investing hundreds of billions of dollars per year per year um to make this happen. These are the companies
00:31:52
with the biggest balance sheets in the history of business that are quite literally going allin to to make this
00:31:58
happen. And they're not the only ones. You know, there's also startups like Crusoe and uh there's there's even
00:32:03
nation states um that are following suit. So So what's going on there? What's the what's the prize that that
00:32:09
they're going after? Um the you know the opportunity here is that uh uh for the
00:32:15
first time in human history, we've actually been able to manufacture intelligence. um intelligence is the uh
00:32:22
scarcest economic resource um in the history of the economy and and for the first time we're actually able to make it. Um and the opportunity here is to
00:32:30
actually unlock access to what has uh historically been that scarce economic resource. So um this is why the data
00:32:40
centers of the future are not being referred to as data centers. They're actually being referred to as AI
00:32:46
factories. It's a factory that takes as inputs data and algorithms and chips and
00:32:52
energy and it outputs intelligence. This is the alchemy of intelligence.
00:33:00
So this newly manufactured intelligence will spawn a new chapter of unprecedented productivity and
00:33:05
development and that will serve to improve human quality of life. So the
00:33:12
IDC estimates that AI will generate $20 trillion in economic impact by 2030. So
00:33:18
even if you can earn a small slice of that, that hundreds of billions of dollars of investment will earn an amazing return. For each dollar invested
00:33:26
into uh business related AI is expected to generate $4.60.
00:33:31
Uh as my friend Jensen would say, the more you buy, the more you save. Um or in this case, the more you buy, the more
00:33:38
you make. uh and we can grow the pie together and usher in a new era of AIdriven abundance.
00:33:46
So when we look at the history of American energy production and consumption, uh as as as the US
00:33:52
industrialized, we we really ramped up energy generation and uh and also consumption. But if you look at this
00:33:58
chart, you can see that, you know, it's it's kind of flatlined over the last 20 years uh where we're, you know,
00:34:04
generating and consuming about 4,000 terowatt hours per year. AI is fundamentally transforming this demand
00:34:10
picture and uh and and energy is quickly becoming the bottleneck to growth.
00:34:16
Uh data centers are forecasted to account for 20% of the growth in power demand between now and 2030. And uh data
00:34:24
center total power consumption is going to go from 2.5% of US power consumption to 10%. So what this means is that the
00:34:30
technology industry that's sort of willing this infrastructure into existence fundamentally needs to bring its own power to support that growth. Um
00:34:37
which means massive investments not just in data centers but also in the energy infrastructure to support them. And this
00:34:43
will require people lots of people to build operate maintain uh and run these
00:34:49
these large scale energy investments. Um so if we look at data centers you
00:34:54
know by the numbers I think it's important you know as people are sort of throwing around gigawatt scale data centers uh of looking at the amount of
00:35:00
data center infrastructure that exists today. Uh Northern Virginia is sort of the center of the world for data centers
00:35:06
but it's only you know at the end of 2014 it was only 4.5 gawatt. Today we have companies that are looking at building single 5 gawatt facilities. And
00:35:13
if you look at this growth we're building more than a uh northern Virginia every single year in the forecasted future.
00:35:21
So we need new in So if there's one thing that you're going to take away from this presentation, it's that we need new infrastructure. We need lots
00:35:28
and we need lots of it and we need lots of people to build, operate and maintain it. This is what Cruso is focused on
00:35:34
solving. Cruso is in the business of activating energy for intelligence of
00:35:40
building operating uh AI factories at scale from from the steel to the
00:35:45
silicon, from the electron to the token. And if you look at our pipeline, we have about 40 gawatts of capacity that spans
00:35:52
um all sorts of energy resources from new energy technologies like uh you know like like uh small modular reactors to
00:35:59
uh renewables and and natural gas to to power this uh innovative future.
00:36:06
Uh so revisiting my formula here, I think we left off one critical component which is the people. Uh AI will be AI
00:36:12
infrastructure will be the largest job creation catalyst that we've ever seen. So I think it's important to sort of
00:36:19
look at you know what this looks like in practice. For the last year um Cruso has been building um uh a large large scale
00:36:26
AI factory in Abalene, Texas. And uh you know speed is paramount. Again this event is winning the AI race. Uh in
00:36:34
order to win a race you really need speed. Uh, and Cruso's really been focused on on on using modular
00:36:40
components on on rapidly scaling investment in in construction and infrastructure to support this. And and
00:36:46
and and we've actually built a lot of different modular components in in factories and brought them to site. And
00:36:52
they're kind of like uh they're kind of like Lego blocks that sort of fit together to to build one of these AI factories at at rapid scale and speed.
00:36:59
Um so if you look at if you look at what this looks like today um you know this is uh this is this is this this site
00:37:07
will consume over 1.2 gawatts of power and 400,000 NVIDIA GPUs all in a single
00:37:14
coherent cluster. So uh this will essentially be a gigawatt scale computer to drive human progress forward.
00:37:22
Um you know it's really amazing what you can kind of accomplish in a year. You see just one year ago, this is what the site looked like and this is what it
00:37:28
looks like today. Um, so what does this mean from a jobs perspective? We have 4,000 people
00:37:34
working on site every day to, you know, make this facility happen. Um, and you know, it's it's it's a bunch of
00:37:39
different trades, electricians and plumbers, um, and construction workers. Um, and it's required a lot of capital,
00:37:44
too. We raised $15 billion to basically put this facility and and and and bring
00:37:49
it into existence. Um, and it's also required manufacturing
00:37:55
and that's and a lot of the critical components have happened off-site in these controlled manufacturing environments.
00:38:01
Um, but this isn't the only one. This isn't a one-of-a-kind. We also are building AI infrastructure and AI
00:38:07
factories across America. This site in West Texas is going to be a gigawatt facility uh behind the meter with wind
00:38:12
with incremental gas and grid interconnection. Uh we did a partnership with Redwood Materials where we built
00:38:17
the largest uh uh we built the the largest micro grid with uh in the United
00:38:23
States with 60 60 megawatt hours of of batteries, end of life EV batteries and uh 20 megawatts of solar to power an AI
00:38:30
factory. We have a partnership with GE Verova and engine number one uh for 4.5 gawatt of new uh gas generation capacity
00:38:37
to power future AI data centers. And finally, uh, we want to announce a new
00:38:42
partnership that we're doing with Tall Grass Energy in Wyoming that will initially power 1.3 gawatts of total
00:38:47
compute load uh, alongside two 2 gawatts of power generation and ultimately we feel like this can scale to 10 gawatts
00:38:53
of power. Uh, so we're really thrilled to partner with Tall Grass. So uh as a vertically integrated AI
00:39:00
infrastructure company built here in America, we believe that AI factories will be the ultimate economic engine
00:39:05
creating utility for society uh and and and and new jobs for the economy. Um
00:39:11
this will usher in a massive new era of AIdriven prosperity for the United States. And I want to leave you with,
00:39:17
you know, my final quote from Warren Buffett that, you know, in this AI race, uh never bet against America. Thank you.
00:39:28
So, is this stuff real? You guys, you started off as like a, you know, sort of Bitcoin miner and now somehow all the
00:39:33
hyperscalers are asking you to, you know, build non-stop data centers. Why you guys?
00:39:38
Um, you know, I think again it comes back to this being a race and one of the things that Cruso's been able to do
00:39:45
better than anyone is execute at speed and scale. And I know there's been like some of the biggest constraints around, you know,
00:39:51
sort of, you know, water, energy, um, you know, the land for this type of stuff. like where have you seen what
00:39:56
parts of the country you know are you guys able to actually do this or have you seen any of the local regulators start to step up to you know make this stuff easier for you? Um, you know,
00:40:03
we've been building quite a bit in Texas. You know, Abene, Texas is this, you know, initial facility that's gotten a lot of coverage. Uh, you know, we we
00:40:10
just sort of announced another facility in Texas. Uh, Wyoming's been a, you know, big area of investment for us.
00:40:15
But, um, you know, there's a number of other states that were sort of evaluating, investing uh, to build large scale AI.
00:40:21
Is it only going to be the like, you know, sort of more rural, you know, sort of red states or do you think that like, you know, Oregon, Washington, etc. will
00:40:26
start to, you know, sort of get together and realize they've got cheap hydropower and, you know, cheap water and we'll try and get you there. Uh, no. You know, believe it or not,
00:40:32
we're actually looking at something in California. Wow. California. Gavin Newsome is gonna bring you in.
00:40:38
I would imagine that's gonna take like 50 years with their regulations. Maybe. We'll see. We'll see. And you the Do you think that the, you
00:40:44
know, sort of hyperscaler demand? Obviously, we were just, you know, on with Lisa Sue talking about the demand for chips over the next couple years.
00:40:49
That's obviously correlated to the, you know, demand with data centers. Do you think that's actually going to play out the way that all the public markets are,
00:40:54
you know, sort of projecting or are we like in 1999 peak, you know, everybody thinks that fiber is going to be deployed all over the world? Turns out
00:41:01
all those projections were totally off. Um I I think the important trend to watch is sort of the capital investment
00:41:06
that's happening and and the term over which that's happening. So I thought like Meta backed off on it a little bit like didn't they like for a
00:41:12
little bit talked about they were going to deploy like crazy and then pulled back although he's obviously spending a billion dollars on chief AI scientist
00:41:17
now. Yeah. I think you know the the investments they're making in people are actually rounding errors compared to the
00:41:23
investments they're making in infrastructure. And I think that's something to sort of appreciate in this moment in time. Like people are betting
00:41:28
their entire balance sheets. These are the biggest, you know, and best balance sheets in the history of business and they're betting their entire balance
00:41:34
sheet on on, you know, the future infrastructure that's going to power the modern economy. And then in like the, you know, data centers like Texas, like
00:41:41
what's the limiting factor? Like is it like workforce to actually go build these things? Is it like materials? Is it the cooling towers? Is it the chips?
00:41:47
Is it the hyperscalers giving you the like, you know, you know, sort of contracts? What's the, you know, sort of limiting reagent?
00:41:52
Um, you know, uh, labor is definitely like a major constraint. you know, like I I said, you know, we have about 4,000
00:41:58
people on site um every day. Uh we're going to have multiple sites that are operating with thousands of, you know,
00:42:04
uh folks uh basically building this infrastructure. So, you know, uh labor
00:42:09
is definitely like one of the one of the big bottlenecks and we're, you know, we think it's really important for America
00:42:14
to make these massive investments in the workforce uh to really build the infrastructure of the future. And you think that requires some real real
00:42:20
reskilling where it's like people from oil and gas or like construction having to go into just totally net new fields or is it something where you guys are
00:42:26
actually able to pull on pre-existing talent pools pretty quickly? Um both you know there's there's a lot
00:42:31
of existing uh you know labor uh you know at that facility in Abene where we're actually pulling labor from all 50
00:42:37
states at this point believe it or not. Um so making like a company town importing people in. Yeah. you know we we have about 50% of
00:42:43
the people are are are uh you know from Texas but um you know the uh we we are importing a lot of labor to to make the
00:42:49
project happen and um you know do you do you see the company starting to go more full stack beyond just like the operations of like
00:42:54
the you know sort of data centers or how do you think about like you know you started off with you know focus on like you know sort of energy arbitrage now to
00:43:00
data centers where do you see you guys goeselves going over time yeah is a vertically integrated AI
00:43:05
infrastructure business so you know data centers is a key component to that and you know I think one of the most important pieces to building right now
00:43:11
and and one of the hardest things to do at speed. Um but we also have you know this managed AI cloud services layer
00:43:17
that um enables innovators to build uh large scale AI applications on the platform.
00:43:22
Makes sense. Well yeah Chase, thanks so much for uh you know joining us on stage and um appreciate the talk. Yeah, thanks D.
00:43:29
Okay everybody, we got a real treat for you. Jensen Wong is here. Sit here. All right. Sit here. Sit here.
00:43:34
The hot seat. Thanks for coming. Thank you. making that happen. Thank you.
00:43:40
The number one podcast in the world. We were saying the number one company in the world.
00:43:45
Wow. It's You're a fan of the pod. You listen to the pod. This is This is Norman, our host.
00:43:52
Yes. Yes. And there's Steve. Uh what's the story with the jacket? You got one of those. You have like six.
00:43:57
I have something like 50 or 60 of them. So you really? Yeah. Wow. What is that? Tom Ford.
00:44:04
I think so. This one is I think so. Yeah. That's nice. I like it. I tried I tried that on. It was like way too much money.
00:44:09
Well, you guys are all so fashionable. Coming from you guys, it actually means something. Yeah. Oh, yeah. Oops.
00:44:14
Oh, look at you. Look at you. Uh, hey, we we've been talking a lot about opportunity. You've talked.
00:44:20
So, mine is like a model. He is. He is. Okay. Good idea. He's definitely in his head. He's like,
00:44:27
"Is Tom Ford your favorite?" Who's your favorite, right? I My favorite is whatever my wife gets me. Ah, she dresses you, right?
00:44:32
As soon as she gets it for me, it's my favorite. Yes. Same with me. Same with me. Smart man. Um, how you got up. Nobody wears a Nobody wears a suit
00:44:39
better than Jacob. Good God. Yeah, Jacob. He's a handsome man. Just trying to keep up with you guys. Um, I got I have two questions for you.
00:44:45
Take them in whichever order you like. Um, we've been talking a lot about job
00:44:50
displacement, opportunity, short-term, long-term. Obviously, you get to see
00:44:56
everybody applying the technology because, hey, listen, you've got the best product in town to build on.
00:45:03
Therefore, everybody explains to you their hopes, their dreams. So, you have a unique way of looking at the playing field. You have complete information
00:45:09
that we don't have. So, I want to know what you think. Don't worry, we'll fix it. Um,
00:45:17
what you think that edit what you think about job creation, transfer, displacement, etc. And then
00:45:23
the second one, I've just always been curious. You got all these important people knocking on your door. You got Zuck, you
00:45:29
got E, you got uh, uh, Sam Alman. He seems like he's a little bit of a
00:45:34
headache. I'll be honest. Um, but he's great. He's great. I'm joking. I'm joking. How
00:45:40
do you allocate the H100's and whatever else you're selling them and still have them all
00:45:47
like you? Because they must ask sometimes, hey, can I get extra? I'll pay you extra. So, just the allocation
00:45:52
of a finite amount of resources and then jobs. First of all, I wrote off $5 billion
00:45:59
worth of hoppers. If anybody would like to have some extras, you got them.
00:46:05
you know, just give me a call. Uh jobs, uh uh we use AI across a whole company.
00:46:11
Every single software engineer today uses AI. Not one left behind. 100% of our chip designers use AI. We are busier
00:46:17
than ever. And the reason for that is because we have so many ideas that we want to go pursue. AI makes it possible
00:46:23
for us to go pursue those ideas now that we're not doing the mundane stuff. And so I I think the first idea is the more
00:46:30
productive you are as a company um so long as you have more ideas, you could
00:46:36
pursue those ideas, you'll go after those ideas. And I I think that that AI in my case is creating jobs. It causes
00:46:43
us to be able to create things that other people would uh customers would like to buy. Uh it drives more growth.
00:46:49
It drives more jobs. You know, all that goes together. The other thing that that to remember is that AI is the greatest
00:46:55
technology equalizer of all time. Okay, explain. Everybody's a programmer now.
00:47:00
Yes. You used to have to know C and then C++ and Python and you know in the future
00:47:07
everybody could program a computer, right? Just have to get up and if you don't know how to program a computer, you don't even know how to program an AI, just go up to the AI and say, "How
00:47:13
do I program an AI?" And the AI explains to you exactly how to program the AI. Even when you're not sure exactly how to
00:47:18
ask a question, you say, "What's the best way to ask the question?" And it'll actually write the question for you. It's incredible. Yeah. And so it's a
00:47:25
great equalizer. Everybody is going to be augmented by AI. Everybody's an artist now. Everybody's an author now.
00:47:32
Everybody's a programmer now. That is all true. And so we know that AI is a great equalizer. We also know that uh
00:47:40
it's not likely that although everybody's job will be different as a result of AI, everybody's jobs will be
00:47:46
different. Some jobs will be obsolete, but many jobs will be created. The one
00:47:52
thing that we know for certain is that if you're not using AI, you're going to
00:47:58
lose your job to somebody who uses AI. That I think we know for certain. Yeah. There's not a software programmer in the
00:48:04
future who's gonna be able to hold their own. I mean, you know, typing by
00:48:09
themselves. Yeah. You can't raw dog it. No. No. Not anymore. No raw dog. Not anymore. You can't raw dog it. And
00:48:17
tell us. I'll be sure to go home and tell people. Yeah. Exactly. You're not going to raw dog this. Yeah. Get your co-pilot on. Now, what
00:48:23
about the allocation of all the Okay. So, the way we allocate is this. The way we allocate is this. Uh, place a
00:48:29
PO. Okay. That's it. That's you go to the register, you pay, you order
00:48:35
first you you know first in in the old days with Hopper it happened so fast it
00:48:40
was impossible to keep up with the demand. But now uh we we uh uh we
00:48:46
disclose our road map to all of our partners uh a year in advance. Gives everybody a chance to plan with us. They
00:48:53
decide how much power and how much data center space and how much capex they want to allocate. We plan together. We
00:48:59
work on transitions. It's really quite oily these days. What's the lifespan now? You, you know,
00:49:04
I was looking into how they're amateurizing, you know, these units four, five years. What happens to this
00:49:09
massive buildout in your six, seven, and eight? What will be the use of those computers if you keep building such
00:49:15
great products that replace them at 2, three, four times? What do we do with all that? Concepts are happening right now. The
00:49:20
first thing first thing is every generation we increase the performance by X factors. Yeah. Um if the perf per dollar perf per
00:49:28
watt goes up by x factors whatever your data center power is we just increase your revenues
00:49:35
by x factors right so perf per watt is equal to revenues per per dollar equals
00:49:41
the cost and so we increase your perf per dollar by x factors we reduce your cost by x factors does make sense that's
00:49:47
the first idea and so every single the reason why we're moving so fast is we're trying to increase everybody's revenues
00:49:53
we're trying to decrease everybody's cost so that we have the benefit of driving AI cost down as far as possible
00:50:00
so that we can have thinking AI, right? It's not that we're trying to make, you know, AI so that it generates a thousand
00:50:07
tokens and that's it. In the future, you're going to be generating millions of tokens and it generates an answer as
00:50:13
result of that. You got to think a long time and so you got to get that cost down. The second idea is if you look at
00:50:19
the residual value of Nvidia gear right now, Hopper for example, one year one
00:50:26
year later it's probably about 80% 75 to 80% of the value of the original value
00:50:33
and then one year later is another kind of like 65% and then one year later is like 50%. The reason and right now if
00:50:39
you try to get hoppers in the cloud it's all sold out. The reason for that is because CUDA is so programmable and
00:50:46
we're constantly the whole world not just us the whole world is doing open source development improving its
00:50:52
effectiveness. And so what's amazing is the performance of Hopper increases over time because
00:50:59
we're improving the software stack it Hopper improved in performance by us and others by a factor of four
00:51:06
right in the time that we shipped it. Now, you can't get that out of a CPU, right?
00:51:11
Jensen, can you explain to us um Elon's tweet and the impact to to your
00:51:17
industry? He said, "We're going to have 50 million H100 equivalents by in 5
00:51:23
years from now." And everybody started to feverishly do the math because if he
00:51:29
has 50 million H100 equivalents, then OpenAI will have that much or more. Meta will have that much or more, Google,
00:51:36
etc., etc., etc. Can you just explain to us layman what that means what he just said and how it impacts your business?
00:51:43
Um, one of the biggest observations about AI
00:51:49
is that there's there's the industry of applications that AI has created. It's a
00:51:55
revolutionary technology. Every industry would will be revolutionized. New applications will be created. so on so
00:52:01
forth that all the things that we know agentic AI reasoning AI robotics AI so
00:52:07
on so forth we we know all those things now every industry healthcare education transportation you name manufacturing
00:52:14
all revolutionized the one part that that that we observed and and made a
00:52:21
great contribution to is that in order to sustain those applications you need
00:52:27
factories of AI you have to produce produce AI unlike unlike software you
00:52:33
write the software and that's it in the case of AI you have to continuously produce it generate the tokens
00:52:38
right in a lot of the same ways that energy production was a large part of the
00:52:43
economy a couple two 300 years ago I think it actually peaked out at 30%.
00:52:49
Yep. There's a whole there's going to be a whole industry of just producing tokens
00:52:55
and this is going to be the new infrastructure just as we have the energy production infrastructure, we
00:53:00
have the internet infrastructure and we got to build out that plumbing and now we got to we have to build out the AI
00:53:06
infrastructure. My sense is that we're probably, you know, a couple of hundred billion dollars, maybe a few hundred
00:53:11
billion dollars into a multi-t trillion dollar infrastructure buildout
00:53:16
per year. Yeah. What about manufacturing? And the reason for that is because you
00:53:23
want the new infrastructure which increases revenue driving your cost down. Right. That's right. What about manufacturing in the US? So
00:53:29
where are we? We um you know we've seen stories of TSMC in Arizona. We asked
00:53:34
this question earlier about how it's going. Is the US equipped? Uh what is it going to take for us to get there to
00:53:40
have onshore fabs? First of all, you guys know you're talking about the United States.
00:53:45
uh the the uh I know that there's there lots of concerns and and everybody's you
00:53:51
know worried about competition and things like that but we are talking about America here. This is this is
00:53:57
unquestionably the most technology rich country in the world
00:54:02
and this is the most innovative countries in the world and the computer industry I have the I have the honor to
00:54:08
serve is the single greatest industry our country has ever produced. I think
00:54:14
we could acknowledge that. Yep. The the level of leadership of the computer industry, the technology
00:54:20
industry is just unimaginable worldwide. And so this is our national treasure.
00:54:26
This is one of our country's assets. We have to make sure that we continue to to to advance it. Um onshoring
00:54:33
next generation manufacturing is going to be insanely technologydriven. Uh robotics technology, AI technology.
00:54:39
You're going to have factories that are going to be orchestrated by AI orchestrating a whole bunch of robots that are AI building products that are
00:54:46
effectively AIS, right? So, you're going to have this layers of inception. And the amount of technology necessary to
00:54:53
create that is is really insane. We've I I love President Trump's vision, bold
00:55:00
vision of reindustrializing the United States. That entire band of industry that's missing, we out we outsource too
00:55:07
much of it. Frankly, we don't need to insource all of it, but we ought to bring onshore the most advanced, the
00:55:14
most economy sustaining, driving, national
00:55:20
security enhancing parts of the industry. You know, people always degrade down to tennis shoes. We don't
00:55:26
have to go there. We just manufacture chips and AI supercomputers. In Arizona and Texas, we will in the next four
00:55:32
years probably produce about half a trillion dollars worth of AI supercomputers. that half a trillion
00:55:38
dollars worth of AI supercomputers will probably drive a few trillion dollars worth of AI industry, right? And so that's only in the next several
00:55:44
years and and uh they're doing great. Arizona is doing great. And so there's um there's a lot of talk
00:55:49
about American competitiveness today and the White House ruled out its AI action plan and Nvidia is making very big bets
00:55:56
on the United States. And so as a CEO of a global company, what do you see are America's unique advantages that other
00:56:03
countries don't have? America's unique advantage that no
00:56:10
country possibly have is President Trump.
00:56:16
And let let me let me explain why. One uh on
00:56:21
the first day of his administration, he realized the importance of AI and he realized the importance of energy.
00:56:29
For the last I don't know how many years energy production was was vilified if
00:56:36
you guys remember. Yeah. We can't create new industries without
00:56:42
energy. You can't reshore manufacturing without energy. You can't sustain a brand new industry like artificial
00:56:47
intelligence without energy. If we decide as a country the only thing we want is IP to be an IP only a serviceson
00:56:55
country then we don't need much energy. But if we want to produce things, something as vital as artificial
00:57:03
intelligence and we need energy. And so I'm just delighted to see pro to
00:57:08
accelerate AI innovation to accelerate the growth of energy so that we can sustain this this new industry and um
00:57:17
you know go after the the new industrial revolution. Big big deal. Can can you talk about physical AI
00:57:22
versus data center AI? We tal we talked a little bit about this today. Is there a threshold where you see physical AI
00:57:28
accelerating and ultimately the deployment of chips outpaces the deployment of chips in data centers? Is
00:57:34
that where the world evolves to or what do you think of the world looks like? Yeah. Excellent. Everything in the world
00:57:40
that moves will be autonomous someday and that someday is probably around the corner. So everything that moves, we
00:57:46
already know that your lawn mower is going to, you know, who's going to be pushing a lawnmower around? That's craziness. Unless you want to. I mean
00:57:52
it's you know and so so I think everything that moves will be autonomous and every machine
00:57:59
every company that builds machines will have two factories. There's the machine factory for example
00:58:05
cars and then there's the AI factory to create the AI for the cars.
00:58:10
And so maybe you're uh a machine factory to build human or robots. You need an AI factory to build a brain
00:58:17
for the human or robot. Right? And so every company in the future, in fact, the the future of industry is really two
00:58:23
factories. Yeah. Uh Tesla already has two factories, right? Elon has a giant AI factory. He's
00:58:30
he was very early in recognizing that he needs to have an AI factory to sustain the the cars that he has now. Now he's
00:58:37
got AIs in the car, but in the future instead of, you know, I imagine that in the future instead of a whole whole lot
00:58:43
of people remote remotely monitoring air traffic control, it'll be a giant AI that's doing the remote
00:58:49
control and then only in the case of the the giant AI um can handle it with a
00:58:54
person come in to to uh intercept. And so so I think you see that that these
00:58:59
industries in the future, every industrial company will be an AI company or you're not going to be an industrial
00:59:04
company. There was a couple of moments throughout the course of this year where
00:59:10
people almost threw in the towel and said, "Oh, we lost to China." Right. There was the Deep Seek moment, then
00:59:15
maybe this week, last week, there was this Kimmy model moment. Um, but then it
00:59:20
kind of fizzled out. Can you just uh explain to us how big of a threat they
00:59:26
really are in terms of getting to supremacy, getting there first, whether it's AGI or, you know, super
00:59:33
intelligence? Yeah, excellent question. Um, the Chinese AI labs are the world
00:59:39
world's leading open open model companies. They they offer
00:59:45
the most advanced open models. Open source is fantastic. If not for open source, we know startups won't exist.
00:59:53
And to the extent that we believe that the future is going to be the future industry is going to be today's
00:59:58
startups, they're going to need open open source models. And Deep Seek when it came out, it was a great win for the
01:00:05
United States. It was an incredible win. What people didn't and two two reasons. First, imagine if Deep Seek came out and
01:00:12
it only ran on Huawei. I just want us to pretend. Use that thought experiment.
01:00:18
Totally. You got two parallel universes. Exactly. Could you imagine if Q came out
01:00:23
and it only worked on non-American tech stack? Could you imagine if KI came out and it only worked on non-American tech
01:00:30
and these are the top three open models in the world today? It is downloaded hundreds of millions of times. So the
01:00:36
fact of the matter is American tech stack all over the world being the world's standard is vital to the future
01:00:44
of winning the AI race. You can't do it any other way. We've got to be you know as you know any computing platform wins
01:00:51
because of developers. Yeah. And half of the world's developers are in China. So speaking of developers,
01:00:57
the second the second I'm sorry, please go ahead. The second thing and it's really a big deal. When DeepC came out,
01:01:02
we were thrilled for the second reason which is we now have a super efficient
01:01:08
reasoning model. And the reason for that is because the old models are one shot.
01:01:13
Give it a question. Everything was memorized. You know the pre pre-training is basically memorization and
01:01:18
generalization. Two concepts. post training is teaching you how to think. And so now with Deep Seek R1, Kimmy
01:01:26
Kimmy K2, um, uh, Q13, you now have reasoning models that can allow that
01:01:33
help you think. And so the reason why I was so excited is if each pass of a
01:01:38
thought is energy efficient, then you can think for a long time, right? Yeah.
01:01:43
The last question from for me is that we see uh this capital being applied to
01:01:51
human capital in a way that we never thought was possible. It used to be NBA players signing $300 million contracts.
01:01:57
Now it's you know uh model researchers and then there was a there was a post this weekend that that said that there
01:02:03
was a person that was offered a billion dollars over four years by Meta. Now if that's happening at this layer,
01:02:09
why hasn't it happened at your layer? because you are the enabler of all of that. And how do you think all of this
01:02:15
human capital is going to actually play out? First of all, I've created more billionaires on my management team than
01:02:21
any CEO in the world. They're doing just fine.
01:02:28
Okay. And so, and and they're doing don't feel sad for anybody at my layer.
01:02:34
Yeah, everybody's doing okay. Yeah, my layer is doing just fine. I I tell I but but the important the big
01:02:39
idea though is that you're highlighting is that the impact of a
01:02:45
150 or so AI researchers can probably create with enough funding behind them
01:02:51
create an open AI it's a it's not a 150 people yeah it's not a it's not well deepseek
01:02:58
150 people moonshots 150 people right right and so I mean look at the original uh
01:03:03
open AI was about 150 people uh deep mind you know They're all about that size. I think I think um you know
01:03:10
there's something about the elegance of small teams and that's not a small team. That's a good goodiz team with the right
01:03:16
infrastructure. And so that kind of tells you something. 150 people if you're willing to pay say $20 billion
01:03:23
$30 billion to buy a startup with 150 AI researchers. Why wouldn't you pay one? Right. Uh speaking of By the way, we
01:03:31
told we need to wrap because I'm going to do this one question. Somebody who was inside your organization told me
01:03:36
with the options that uh you have a secret pool of options and that you will
01:03:41
randomly just if somebody does a great job dropped a bunch of RSUs on top of
01:03:47
them and that you have this like little bag of options you carry around and that you leave them out. That's nuts.
01:03:53
Is that true? Yeah, I'm carrying in my pocket right now. So listen, so this is what happens.
01:03:59
I review I review everybody's compensation up to this day. Yeah. at the end of every cycle of when
01:04:05
they present it and they send they send me everybody's everybody's recommended comp. I go through the whole company.
01:04:12
I've got my methods of doing that and I use machine learning. I do all kinds of technology and I sort through all 42,000
01:04:18
employees and 100% of the time I increase the company's spend on opex and
01:04:24
the reason for that is because you take care of people. Everything else take care takes care of itself. All right. Well done. Thank you. Thank you Jess. Great to see you.
01:04:30
Great to see you. We have an event in LA. We'd love to continue the conversation. So, we'll send you a note. The world's number one podcast.
01:04:36
There you go. Thank you.