a16z Podcast - Sovereign AI: Why Nations Are Building Their Own Models
Episode Date: May 24, 2025What happens when AI stops being just infrastructure—and becomes a matter of national identity and global power?In this episode, a16z’s Anjney Midha and Guido Appenzeller explore the rise of sover...eign AI—the idea that countries must own their own AI models, data centers, and value systems.From Saudi Arabia’s $100B+ AI ambitions to the cultural stakes of model alignment, we examine:Why nations are building local “AI factories” instead of relying on U.S. cloud providersHow foundation models are becoming instruments of soft powerWhat the DeepSeek release tells us about China’s AI strategyWhether the world needs a “Marshall Plan for AI”And how open-source models could reshape the balance of powerAI isn’t just a technology anymore - it’s geopolitical infrastructure. This conversation maps the new battleground.Resources:Find Anj on X: https://x.com/AnjneyMidhaFind Guido on X: https://x.com/appenzStay Updated: Let us know what you think: https://ratethispodcast.com/a16zFind a16z on Twitter: https://twitter.com/a16zFind a16z on LinkedIn: https://www.linkedin.com/company/a16zSubscribe on your favorite podcast app: https://a16z.simplecast.com/Follow our host: Follow our host: https://x.com/eriktorenbergPlease note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures.
Transcript
Discussion (0)
They're not being called AI data centers.
They're being called AI factories.
The Industrial Revolution, having oil was important.
And now having data centers is important.
These models aren't just compute infrastructure.
They're cultural infrastructure.
It's not just self-defining the culture,
but self-controlling the information space.
So if a model is trained by a country that's adversarial to you,
that's actually very hard to eval or your benchmark
when the models are released.
This is a massive vulnerability.
Is that the new age of LLM diplomacy that we're entering here?
Do we build? Do we partner? What do we do?
Today we're diving into a conversation that's just as much about geopolitics as it is about technology.
This week, the Kingdom of Saudi Arabia announced plans to build its own AI hyperscaler called Humane.
But they're not calling it a cloud provider. They're calling it an AI factory.
And that language alone suggests a shift. For decades, cloud infrastructure has
been concentrated in two places, the U.S. and China. But with the rise of AI, that model is breaking
down. Nations no longer want to outsource their most strategic compute. They're building
sovereign AI infrastructure, factories for cultural and computational independence. To unpack what
this means for the global AI stack, national sovereignty, and the new digital power dynamics,
I'm joined by Angine Mehta in Guido Appenzeller. We talk about what it takes to become an
hypercenter, why governments are spending billions to control inference pipelines, and whether we're
entering a new Marshall Plan moment for AI. Let's get into it. As a reminder, the content here is
for informational purposes only. Should not be taken as legal business, tax, or investment advice,
or be used to evaluate any investment or security, and is not directed at any investors or potential
investors in any A16Z fund. Please note that A16Z and its affiliates may also maintain investments in the
companies discussed in this podcast. For more details, including a link to our investments,
please see A16Z.com forward slash disclosures.
Ange Guido, want to talk about sovereign AI, AI in geopolitics, and let's start with the news.
Our partner, Ben, is in the Middle East right now to participate in his own way. What happened,
and why is its own point? What happened is the kingdom announced that they're going to build
their own local hyperscaler or AI platform called Humane.
And I think why it's notable is that as opposed to the status quo of the cloud era,
they're viewing the AI era as one where they'd like the vast majority of AI workloads to run locally.
If you think about the last 20 years, the way the cloud evolved was that the vast majority
of cloud infrastructure basically existed in two places, right?
China and the U.S.
And the U.S.
ended up being the home
for the vast majority
of cloud providers
to the rest of the world.
That doesn't seem to be the way
AI is playing out.
Because we have a number of frontier nations
who are basically raising their hands
and saying,
we'd like infrastructure independence.
The idea being that we'd like our own infrastructure
that runs our own models
that can decide
where we have the autonomy
to build the future of AI
independent of any other needs.
nation, which is quite a big shift.
And I think the headline numbers are somewhere in the range of 100 to 250 billion worth
of cluster buildout that they've announced, of which about 500 megawatt seems to be the atomic
unit of these clusters that they're building.
So a number of countries, with the kingdom being the one that's most recent, have been
announcing what we could think of as sovereign AI clusters.
And that's a pretty dramatic shift from the pre-AI era.
I don't know if you'd agree with that.
I think it's spot on.
I think many sort of geopolitical regions are reflecting back what happened in previous big tech cycles.
And wherever the technology is built and whoever controls the underlying assets has a tremendous amount of power of shaping regulation, shaping how this technology is being used.
And also puts themselves in a position then for the next wave that comes out of that.
And, you know, was the Industrial Revolution, having oil was important and now having data centers is important.
And so I think it's a very exciting development.
Yeah, in fact, you can often tell why something is important to somebody by the semantics that folks used to communicate a new infrastructure project, right?
In this case, if you look at how the cluster buildouts are being referenced, they're being called AI factories.
They're not being called AI data centers.
They're being called AI factories.
And I think there's two ways to respond to that.
One train of thought would be, hey, that's just branding.
That's just, you know, the marketing people doing their thing.
And under the hood, this is really just data centers
with slightly different components,
but everybody in the world, in every industry,
is looking for a way to be relevant in the age of AI,
and this is the compute infrastructure world's way of doing that.
An opposing view would be, actually, no, this is not just marketing.
If you look under the hood,
and if you x-ray the data center itself,
very little of it is the same as was the case 20 years ago.
The big difference in active components being GPUs, right?
about 20 years ago,
what would you say
the average number of GPUs
were in a...
What percentage of...
Pretty much, yeah.
It's a very recent phenomenon.
And today, I think if you look
at the average 500 megawatt data center
and you looked at what percentage
of the CAPEX that was required
to build that data center
or operated, rather, went to GPUs.
Massive, yeah.
That's a huge shift.
I think we're also seeing a specialization.
The kind of data center
you built for classic CPU-centric workload
and what you built for
a high-density AI data center
They look very different.
Right.
You need good cooling to the rack.
You need very different energy supply.
Right.
Close to a power plant.
You want to lock in that energy supply early on.
And then we also see that change, I think, in the consumer behavior where classically
you want a very full stack that has lots of services that helps enterprise build all these things.
We're seeing more enterprise that are actually comfortable with just building on top
of a simple Kubernetes abstraction or something.
Right, right.
And basically, you know, cherry pick a couple of snowflake or database type services on the side that
helped them complement that.
So I think there's a new world.
And so that's certainly true
that you could kind of look at the technical components
in an AI factory
are completely different from a traditional data center.
And then there's, what does it do?
And historically, a lot of the workloads
that traditional data centers were doing
were running one cloud-hosted workloads
for enterprises or developers, whoever it might be,
where most of that, the data sets
and their workloads
were actually not
particularly opinionated.
And when I say opinionated,
I mean,
they're not necessarily subject
to a ton of cultural oversight.
Yeah.
You could argue
that was not the case with China.
Right.
Where China wanted full
sort of oversight
over those workloads.
Right.
But for the better part of the 2000s,
until the rise of GDP, R, CCP, and so on,
we lived in an era of centralization
where having most of your cloud infrastructure
in northern Virginia,
was preferable for most of the world's developers' enterprises
because it gave them economies of scale.
That started to change, of course, with GDPR, CCP,
at the rise of data privacy laws
because then you had region-by-region compliance.
And that made the rise of something like Cloudflare critical,
right, where Cloudflare has this idea
of distributed infrastructure
where you can tie the workload policies
to wherever the user is.
But by and large, that was critical
especially for the rise of social media workloads,
but the vast majority of enterprise workloads
didn't need decentralized serving.
What's different about AI seems to be
that these models aren't just compute infrastructure,
they're cultural infrastructure.
They're trained on data
that has a ton of embedded values
and cultural norms in them.
And then more importantly, that's the training step.
And then when you have inference,
which is when the models are running,
you have all these post-training steps
you add that steer the models
to say something or not,
to refuse the user or not.
And that last mile is where things over the last, I would say, year have made it more and more clear that countries want the ability to control what the factories produce or not within their jurisdiction, whereas that urgency didn't quite exist as much.
Because of the cultural factors or because of certain independence or resilience?
It's a good question.
My sense is there's two things going on, but you should chime in if you think I'm being incomplete.
One I think would be the rise of the capabilities in these models being now well-bearing.
beyond what we'd consider sort of early toy stage of a technology.
I think our partner, Chris Dixon, has a great line,
which is that many of the world's most important technologies
start out looking like toys, right?
And four years ago, when the scaling loss paper was published,
and then GPT3 was published, most people looked at it and I said,
okay, it's cool, sure, it can produce the next word.
It's a nice party trick.
It's a nice party trick, right?
It's a stochastic parrot.
And now you have foundation models literally running in defense
in healthcare, in financial services industries.
ChatGPD has about 500 million monthly active users,
making real decisions in their daily lives.
I think using these AI models,
there was a paper that was recently published by Google
that showed the efficacy of Gemini,
their foundation model, at solving medical questions.
And one of the most interesting things you can see
when you look at the usage,
the types of prompts that people are using models for,
relative to two years ago, three years ago,
where it was a lot of helping write my essay.
It's turned into coding and helped me solve a whole host of medical problems
or personal life-related questions and so on,
where it's clear now that these capabilities can be used,
want to drive mission-critical industries like defense, health care, and so on,
and also then influence a number of your citizens' lives.
And so I think that makes a lot of governments go, wait a minute,
if we are dependent on some other country
for the underlying technology
that our military, our defense,
our health care, our financial services,
and our daily citizens' lives are driven on,
that seems like a critical point of failure
in our sovereignty.
So that's one. It's just that models have gotten good
and they seem to be good at a bunch of important things.
The second is, I think, an increasing belief
that if you don't have control over
the model's production pipeline,
then you're doomed or destined to use models
that reflect other people's cultural values.
We had a pretty in-depth debate about this with DeepSeek,
where the question was,
is DeepSeek fundamentally more biased or not
than open-source models trained in the U.S.
And I think there's early evidence to say that
you can actually see, certainly in the post-trained Deepseek,
that there's just a number of topics and types of tasks
that it's been told to avoid in answer.
differently from a model like llama. So that's the cultural piece. I think there's a critical
sort of national capability piece and then there's the cultural piece. And I think both are
combining to create this sort of huge rise and the demand for, you could call it sovereign AI,
which is the idea that you want control over what the models can, can't do. Or you could call it
infrastructure independence. I think everyone's got a different word for it. You could call it our local
AI factory ecosystem. But I think all these terms are trying to get at the same thing, which is
we've got to control our own stack. Yeah. I think I would make it even stronger. I think it's not
just sort of defining the culture, but sort of controlling the information space.
Right.
I mean, today, we're starting to see how, in many cases, models are replacing search.
Right.
I don't longer go to Google and go to chat GPT, and that comes back with an answer.
Right.
If there's historical fact and, say, in the Chinese model, does not show up,
and a U.S. model, it does show up, right?
That is the reality that people grow up with.
Right.
And if you write an essay in school, in the future, many of us essays will be graded by an LLM.
Right.
So in fact, in school, something that may be truthful, right,
maybe graded us wrong
because whoever controlled the model decided
that should not be part of the trading course.
So it has a very profound effect
on public opinion and stuff, you know, on values.
The downstream use is an interesting one
because it's very hard to measure for.
And certainly relative to two years ago
when the vast majority of products and applications,
like the ones Gita was talking about,
were basically pretty simple models, right?
Well, at the time, they were considered pretty complex,
but the frontier change is so fast.
Today, we'd look back at a model like GPT4
that was largely just a next word prediction model
and say that's pretty rudimentary.
Because if you x-rayed an app like chat GPT,
sure, on the surface, it looks like nothing much has changed,
right? It's still a chat box you type in what you need
and it outspits an answer relative two years ago,
but under the hood, there's been this insane evolution
where there's four or five different systems interacting with each other.
You've got a reasoning model that can produce a chain of thought
to think through what it should do next,
including then doing what we call tool usage, right,
calling out to third-party tools.
And then you have the idea that these models can start to self-learn,
go through a loop of taking your input
and reasoning about what it needs to do,
calling an action, and then evaluating its output,
and then updating that loop.
That starts to look more.
People use the word agent, right, to call it that.
But the idea is that it's going from being a pretty simple model
to being a system
and it's very hard
to measure
where the
adversarial cracks
are in this system
so if a model
is trained
by a country
that's adversarial
to you
to when you're
writing code
open up a port
or what we'd
call a call home
attack
right where it's
transmitting telemetry
that's actually
very hard to
eval or your
benchmark when the
models are released
right because
these models are
often tested
in very academic
or static settings
and so when
deep sea came
out, it was just such a great model. It was a phenomenal piece of engineering that suddenly
everybody was using it everywhere. And a number of CIOs and CTOs got pretty nervous because they
were like, wait a minute, if the model is being used in this agentic fashion, and I don't have visibility
on what it's doing adversarily until it's too late, this is a massive vulnerability. And so I think
the adversarial threat, as the systems go from being models to agents, is causing a lot of
governments to go, well, we'd rather have the whole thing running locally in a way that we can
lock down. Again, it comes back to a sort of independence and a supply chain question.
And is your expectation that this is going to play out? And to what extent is it going to play
out? On the cloud, as we mentioned, there's a Chinese internet and the sort of Western rest of
the world internet. How widespread is this sovereign AI thing? Yeah. I'm going to borrow an analogy
that Giro used, which is in the industrial revolution, you could look at where resources flowed, right?
I think you should talk about how viewing it from the lens of oil reserves, you know,
can kind of dictate which countries can and can't participate in the Industrial Revolution.
Go ahead.
So if you look at the industrial revolutions of oil was the foundation of a lot of the technologies,
right?
You needed all reserves in order to participate.
And I think it'll be a little bit the same thing, right?
If you want to build industry in a particular country, if you want to be able to export things,
if you want to be able to drive development, and if you want to harness the power that comes
with that, you need the corresponding reserves.
And I mean, I think AI data centers are a little bit like these oil reserves,
but the big difference being you can actually construct them themselves
if you have the necessary investment dollars and the willpower to do it.
But I think they will be the foundations for building all the layers on top
that ultimately, I think, determine who wins this race.
And in my mind, the countries that invest in building up the AI factories
or in this sense the oil reserves to borrow Gito's analogy,
I think of them as one body of countries.
Let's call them hypercenters, right?
the idea is they're centers that have enough compute capabilities to compete at the frontier
and run their own sovereign model, sovereign infrastructure.
And then there's everybody else who just doesn't have the resources to do that.
And if you look at after the Industrial Revolution, you could argue the next major technology
revolution was the advent of modern finance, the Bretton Woods and IMF regime,
where modern finance said we're going to all use this one measure of value called the dollar.
and you were either in a country that produced the dollars like America
or you were in a country that produced a lot of goods that acquired dollars like China
and then if you weren't in one of those too you really had to figure out whether you
aligned with one of these trade blocks or not and what happened is you had countries like
Singapore, Luxembourg, Ireland and Switzerland who'd realize well we just don't have the resources
to build out our own reserve system and there's not that much by way of local production
that we can do to acquire dollars, we can't really trade.
So we've got to find a way to insert ourselves in the flow, right?
And so Singapore, of course, famously became the entry point for dollar flows into Asia
because they invested a ton in rule of law and a great tax regime
and sort of stable government and low corruption and all of that.
Switzerland did something similar for European investments and in European capital flows.
So I think what we're watching right now is that buildout where there's U.S. and China,
which clearly have enough compute to be hypercenters.
And then you've got folks like the kingdom of Saudi Arabia
saying, we want to be a hypercenter.
And if that means we've got to trade our oil
to acquire large numbers of NVIDIA chips,
we do that right now.
And I think in that bucket,
there's probably the kingdom of Saudi Arabia,
there's Qatar or there's Kuwait,
there's Japan, Europe, clearly.
And then I think the question is,
everybody else, what do they do?
And it's not clear to me
what you have to do to become the Singapore of AI.
And maybe the Singapore VA ends up being Singapore
because actually now they have an enormous sovereign wealth fund
as a result of participating in modern capital flows.
But I think a bunch of other countries are sitting around wondering,
is this the time where we actually buy?
Do we build? Do we partner? What do we do?
Yeah. And talk more about the implications behind what this means.
Is this something that the U.S. should be excited about?
What does this mean as we think about foreign policies?
Are there now winners across the board and all these local environments?
Why don't you talk about some of the big implications here?
I think every big structural revolution is both a threat and opportunity.
I think the United States and AI right now has the world leadership.
Yeah.
That's an opportunity.
Hanging onto it won't be easy because it isn't every tech revolution?
Don't we want people to be dependent on us in the same way that they were in the cloud revolution?
Or do we benefit somehow from it being more decentralized?
The world is not one place.
So I think complete centralization won't happen.
I think the leader is good.
Having strong allies that also have, their technology is also very valuable.
So it's probably a balance of those that we're looking for.
Yeah.
To put a finer point on your last note there is that you could think about a balance.
Like we're clearly in an unstable equilibrium right now.
Yeah.
And so Gito's right that the arc of humanities and history such that things will shake out
until there's a stable equilibrium.
And so what is the stable equilibrium?
And I think one way to reason about it is you could look at historical analogy.
So post-World War II, when Europe was completely decimated,
there was a group of really enterprising folks in the private sector
and the public sector who got together and said,
hey, we can either choose to turn our backs on Europe
and adopt a posture of isolationism
where we mostly focus on a post-war American-only agenda
or we can try to adopt a policy where we know that
if we don't help out our allies, somebody else will.
And so they came up with this idea called a Marshall Plan,
where a number of leading enterprises in the U.S. got together like GE and General Motors
and literally subsidized the massive reconstruction of Europe that helped a lot of European economies
quickly get back on their feet. And at the time, there was a ton of criticism of the Marshall Plan
because it was viewed almost as a net export of capital and resources. But what it did end up
doing is then solidified this unbelievable trade corridor between the U.S. and Europe for the next 50 years,
which really kept China out of that equation
for the 70 years, yeah.
70 years, really.
And so I think we have a choice
either approach it the way we would
the Marshall Plan for AI, right,
and say, well, a stable equilibrium
is certainly not one
where we just turn our back
on a bunch of allies
because China definitely
has enough of the compute resources
to try to export great models
like Deep Seek to the rest of the world.
So what do we want our allies on,
deepseek or Lama?
That's what it comes down to
at the model level of the stack, right?
And I think that
the realities that a number of countries are not waiting around to find out.
That's why you have efforts like Mistral in the EU,
where they are being approached by a ton of, not just European nations,
but a ton of other allies of Europe to say,
hey, can you help us figure out how to build out our own sovereign AI?
And so I think we're about to see basically the single biggest buildout of AI infrastructure ever,
because most of the purchase orders
and the capital is being provided by governments
because they've realized this is a critical national need.
And their stable equilibrium is certainly not
to depend on somebody else
or depend on an uncertain ally.
And so the ones that certainly have the ability
to fund their own sovereign infrastructure
are rushing to do it right now.
And what does that mean for the sort of nationalization debate
or how you see that playing out?
Leopold, Dash and Brenner,
formerly of Open AI in his famous sort of report
talked about how, hey,
if this thing becomes as critical to national security,
as we think it will be, at some point,
the governments aren't just going to let private companies run it.
They're going to want to have a much more integrated approach with it.
Where do you stand with the likelihood of that?
And what does that mean just in terms of the feasibility of regulation
in a world where it's much more decentralized?
And we already had this with Deepseek.
I mean, that already changed the game in terms of we're in an arms race
and you can't control everything.
We're in the open source conversation as well.
We're backing some of these players.
Where are your thoughts on where this all nets out?
I think I have probably a strong opinion on that.
I mean, I grew up in Germany, right?
so benefiting from the Marshall Plan
and also seeing how that
pulled away Western Germany
towards the United States
and eventually East and Germany
also towards the United States
when everybody realized
the impact of that.
One lesson I took away from that
is that I think any kind of
centralized planned approach does not work.
Eastern Germany is Western Germany
is a nice A-B test.
Central planning versus a free market economy
works better, right?
And I think the results speak for themselves.
So I think basically having the government
drive all of AI strategy
the United Manhattan style project or Apollo project,
pick your favorite successful project there.
I can't see that working.
You probably need a highly dynamic ecosystem
of a large number of companies competing.
There's some areas I think
where the government can have a hugely positive effect,
right?
On the research side,
we've seen there again and again,
funding fundamental research,
which is not quite applied enough yet
for enterprises to pick up,
is very valuable.
I think it can help in terms of setting good regulation,
bad regulation, can easily torpedo AI, as we've seen.
And so I think there's a strong role for government to lead this and to direct this.
There's no master plan at the end of the day that you can make that basically has all the details.
That has to come from the market.
I don't agree with the Aschenbrenner point of view.
I agree strongly with Gito that the history of centralized planning at the frontier of technology is not great,
barring a few situations that were essentially brief sprints of war, right?
And arguably even the Manhattan Project, which is the analogy I think he uses in his piece,
we now know that there were leaks.
It was literally a cordoned off facility in Los Alamos or whatever,
and they were still spies.
And so if you're approaching this from the lens of the models
are what are the equivalent of nukes.
And we've got to regulate the development of these
by locking up our smartest researchers
in some facility in Los Alamos.
And that's what's going to prevent the best models
from getting exported.
I think that's great fiction, a very interesting novel.
Yeah.
But for anyone who has ever had both pleasure
and displeasure working in any,
large government system. It's a pipe dream. The good and the bad news is that in a sense,
it doesn't really matter where the model weights are. It matters where the infrastructure that
runs the models are. In a sense, inference is almost more important. And I think a year ago,
we went a pretty rough spot, I would say with the arc of regulation where there were a number of
proposals in the United States to try to regulate the research and development of models versus
the misuse of the models. I think that, luckily, we have moved on from that. Where we are now
is unfortunately still a state level
like batch work on whack-a-mole of regulation.
It's not consistent.
Hopefully, I think we've got a number of positive signals
from early administration executive orders
that they put out that hopefully means
there will be unified regulation around AI.
But I don't think that the answer is going to be
one single lab that has one God model
that then the country protects
as if it's a nuclear bomb.
I think we are now in a state
where partially because of the
build out of AI factories that we've discussed,
a number of countries have the capabilities
to train frontier models
and a number of them are quite willing
to export them openly.
China being a leading one.
DeepSeek has forced people to update their priors
where just a year before DeepC came out
a number of tech leaders in Washington
testifying that China was like five to six years
behind the US with confidence on the record.
And then DeepC comes at 20,
six days after opening I puts out the frontier.
I mean, just shattered all of those arguments.
So the calculus has changed.
I think it means that the only way to win
is build the best technology
and out-export anybody else.
Then if the question is,
whose math is the world using?
We'd love for it to be American math.
Right.
My view is that we are much better off embracing
the ability for other countries
to serve their own models.
And ideally, the best product wins,
which is the best models just come from the U.S. and our ally.
Is that the new age of LLM diplomacy
that we're entering here?
Actually, Ben had a great talking point
to this at FII Riyadh last year
and he said something to the effect of
because these models like we discussed earlier
are cultural infrastructure,
you don't want to be colonized
in the digital era
in cyberspace.
And I think that's pretty spot on.
Instead of colonization,
what we have is now,
I think foundation model diplomacy.
That's a good way to put it.
I think it's right.
It suits the U.S.'s relative skill sets
which is innovation and working with our allies
to China, which has been a bit more closed off as a country.
I want to talk about the bull case
for open source companies like Mistral
in a world where some of these bigger players
are open sourcing more or becoming more interested in that.
So there's a couple,
and we've talked about this increasingly
in a world where two years ago,
I think when we led the investment of Mistral,
we had a fairly clear hypothesis
for how open source wins
in an arc where foundation models end up
looking more and more like traditional compute infrastructure,
storage, networking, et cetera.
Because close source usually, if you look at databases
or operating systems, Windows,
close source usually leads the way in terms of opening up new use cases,
often captures a ton of value, certainly from consumers.
But when the enterprise starts really adopting that technology,
they usually want cheaper, faster, and more control.
And in the world of AI, you can't get the kind of control,
most enterprises want
without having access to the weights.
And at the time,
the only real comparable model
to the frontier close source was Lama
and then the creators of Lama
left to start Mistral.
So it was a pretty natural decision.
I think since then,
there's a different thing that's turned up,
which is the idea of sovereign AI infrastructure
that's not just models,
it's everything else down and up.
And I think something we've been debating
is, well, does that mean
the ideal provider of cloud infrastructure
is also the provider
of the best open source models?
Traditionally, cloud infrastructure is pretty well-dominated category owned by incumbents
whose core business was either in telecom or in commerce, like Amazon.
And it seems like now that's changing.
I think you put it more eloquently than I did, which is if you ask the wrong guys to design the data center,
they're going to design the wrong data center.
But I'm paraphrasing here.
I think it's exactly right.
I mean, each of the last big technological waves, if you look at the PC revolution or the internet boom,
We developed essentially a new building block for systems, right,
the CPU or the database or the network.
I think now with the process of building yet another building block,
which is the model or AI, whatever may be called in the end.
So it's a fourth pillar, in a sense.
Computer network storage has become a compute network storage model.
And in that kind of world, a cloud needs to provide all four.
And so I think you're exactly right.
This is just part of the infrastructure layer
that in the future you build all the software systems.
I think one way to think about that is there's two frontiers.
there's the capabilities frontier
and then there's the Pareto Efficiency Frontier
the capability frontiers
usually dominated by closed source
and then the Pareto Efficiency Frontier
because of all the goodness of open source
ecosystem flywheel effects right where in this case
you put out your model and the entire ecosystem
of developers can distill it, fine tune it,
ship better runtime improvements to the model,
quantize it and so on
that makes that family of technology
much more efficient to run than the close source
version. The second is more secure because you have the whole world red team in your model
versus just this limited group of people inside your company that if you're a closer provider.
So the business case is basically cheaper, faster, more efficient, more controllable. It's pretty
strong for the raw model abstraction. Then if you ask, okay, well, does the model provider have
the right to win? Is there a business case below the model stack at the data center at the chip level
at the cluster level
and is there
right to an above.
Let's start with the
topmost part of the stack
which increasingly
people would call
agents a less sexy
version would be to call
it a fully
end-to-end
automated workflow
where today you have
if you take the world's
largest shipping company
the Merks of the world
or the CMACGMs
right?
These are massive
logistics
and transportation companies
that have fairly
complex workflows
and if you think
about the power of these models
being turned into an AI agent,
the work required to customize that agent
for one of these mission-critical industries
is quite hard today.
An area where we're seeing a ton of progress
is reinforcement learning
where if you craft the right reward model,
the agent gets much better at accomplishing that task.
Well, it turns out crafting the right reward model
is really hard.
Even for sophisticated teams like Open AI,
I mean, they've literally rolled back an update
to chat GPT, I think three days ago.
They called it the sycophancy update
where they crafted the wrong reward model.
And so a traditional legacy industry company has no clue how to do this.
And the question is, would they rather invest that energy to customize a close source model
or an open source model where if the close source provider, for whatever reason, goes down, shuts shop, which happens, raises prices and so on.
Sears their customers, yeah.
Yeah, steal their customers.
We're essentially host.
And the natural arc of that as well for the agent layer seems to be to go to a deployment partner who has an underlying
open source base. I think the cloud infrastructure, the sovereign AI layer is a bit up for grabs.
And that might be a good topic for our next pod. Yeah, absolutely. Well, let's wrap on that.
Ange Guido, thank you so much. It's been great. Thank you.
Thanks for listening to the A16Z podcast. If you enjoyed the episode, let us know by leaving a review
at rate thispodcast.com slash a16Z. We've got more great conversations coming your way.
See you next time.
Thank you.