Robert F. Smith: The Future of AI and Investment
It's my distinct pleasure to welcome Robert F. Smith,
Founder, Chairman and CEO of Vista Equity Partners,
one of the world's leading technology-focused investment firms
with over $100 billion in assets under management.
Since founding Vista in 2000,
Robert has led more than 600 transactions
representing over $320 billion in aggregate value.
A testament to his visionary leadership and deep conviction
in the transformative power of enterprise software.
His pioneering approach to investing has earned him recognition
on Forbes' list of the 100 Greatest Living Business Minds
and Times 100's Most Influential People.
Robert, thank you for joining us today,
and it's an honor, truly.
It is my pleasure to be here,
and thank you all for welcoming us so warmly to Korea.
Again, we are here quite frequently,
have a number of companies operating here,
and it's always been a wonderful place to visit and do business,
so thank you, Jay, for today.
Korea needs a lot of help on software.
They're very renowned for hardware,
but they need to get a lot of help from you on software.
We have found just very strong technical capabilities here,
and I think you know we're members of COSA
and have had a chance to meet with a number of
very talented software engineers
as well as companies that are doing some pretty interesting things
in not only Korea but throughout the region and globally,
so quite interesting and fun to be here.
Thanks for all your help.
So to set the stage, we'll kick off.
So could you give us a quick snapshot of Vista,
what inspired you to focus on software,
and how your strategic approach shifted
as both the market and Vista itself has grown?
Great. Thank you.
We have now been at this 25 years.
My career, as we talked about, I started as a chemical engineer,
and my generation of chemical engineers
focused a lot on process controls.
Those process controls were leading,
one of the leading adopters of technology,
and we were adopting in those cases
programmable logic controllers or compute
that was able to bring to bear a tighter process control paradigm.
I like to explain it this way.
In the past, you would manage an industrial process.
You would have an episodic observation by the operator.
They would observe the reaction or transact that was occurring.
They would then intervene by changing certain states
of heat, temperature, catalyst, or some condition,
and it would change the reaction dynamic,
and they'd monitor it and maintain it that way.
And so you had that sort of a control paradigm
that was present in operating in those processes.
Well, when you introduce compute,
compute will be able to read and evaluate
the state of the transaction much more frequently,
like to say 60 times a second as opposed to episodic observation.
And so you could intervene at a higher frequency.
So you went from control paradigms that looked like this
to control paradigms that basically looked like that.
And so for all the wonderful mathematicians in the room,
all the area under the curve is waste.
And so in introducing compute into process industries,
it created a massive amount of productivity
throughout first process industries,
and then we introduced compute into the office environment
as it became cheaper and the ability to proliferate
what were then on-premise solutions and applications into offices.
That same sort of productivity was then driven
throughout the global economy
as we distributed more compute into every environment.
I would argue that over the last 50 years,
enterprise software, enterprise software as business software,
has been the most productive tool introduced
in the business economy in the last 50 years.
And it has hit every single segment, every single market,
and now every single geography as we've distributed compute globally.
One of the insights that I had
as I was first an investment banker at Goldman Sachs
was that there were various ways that companies were approaching
bringing enterprise software into the industries that they served.
And the idea and the insight was,
well, why don't we bring together a group of standard insights
or best practices, as we called them,
and drive them into the operations of enterprise software companies.
And the reason that became quite an interesting opportunity at that time
was there was now a separation of hardware and software.
In the early days, hardware and software were sold as solutions.
They then got separated, and you can now actually develop
application software for specific industries
at a much more rapid rate with a higher utility
because it was focused on the actual dynamic in that industry
and the workloads in that industry
and had the ability to improve without having to make hardware improvements.
And so Vista was founded on the principle
that what we will do is identify enterprise software companies,
bring a set of best practices to how you operate those businesses,
the way you write the code, the way you support your customers,
your go-to-market strategies, your back-office administration,
and all of those components,
and make it a much more efficient operating business.
And so that was the thought.
We have now brought to bear a tremendous amount of capital.
We now have over $100 billion in assets under management.
We today hold 90 software companies across 70 industries,
which if you add it all up, it's a little over $32 billion in revenue,
over 120,000 employees globally, operating in 180 countries,
which makes us about the fourth or fifth largest
enterprise software company on the planet.
And so we have brought to bear the system of best practices.
And over time, that's evolved.
In the early days, it was focusing on legacy,
on-premise enterprise software companies.
Then there was the emergence of hosted.
Many of you know it as SaaS or a multi-tenant hosted environment
or in the cloud.
And so we built a factory at that time to migrate companies
from legacy enterprise software companies to cloud,
and I think have now converted more businesses from legacy to cloud
than any institution on the planet with tremendous success.
The beauty of technology is it is always changing, constantly changing.
And so now we have this new technology,
which we're going to spend a bit of time on today,
called artificial intelligence or generative AI,
and in our case, moving that towards agentic AI.
And we've now built another factory to make those same sort of conversions.
So that's the beauty.
I call it the interesting part of our work.
It is constantly evolving, and you have to constantly evolve your team
and your infrastructure and your approach to adopting new technologies
into your individual portfolio companies at scale.
And the ability to do that globally is, I think,
something we are positioning ourselves to be the best at.
That's great. Thanks for introducing the next question.
So as we see development in artificial intelligence, AI,
and many are calling this the most significant technological shift
of our lifetime, some even suggest that AI could eventually
replace traditional software altogether.
How do you view this evolution, and do you see AI as a complement software,
a disruptor, or something entirely new?
It's a great question, and it's one that the world is asking.
The world is asking very large software companies as well.
I think to really understand where we are, we have to look a bit at the past
to see how technology, in essence, starts to be adopted in various industries.
The first thing to think about is every time there's a new technological innovation,
for the most part, there are groups that capture economic value, economic rent,
and typically the first wave often goes to the hardware vendors.
And so you're experiencing that today.
You're experiencing these hardware vendors, which are capturing a massive amount
of not only attention, but economic rent and value.
I think the estimates now are about $1.8 trillion, with the T, dollars,
that are being poured into not only the hardware and the semiconductor
layers of this stack, but also now the infrastructure.
This group of hyperscalers that have taken advantage of the last wave
of Internet capabilities and mobile capabilities are best positioned now
to be the hyperscalers of the future, to deliver compute on demand,
not just CPU compute, but now GPU compute.
And so these waves were in the middle of this build-out of infrastructure
where they would look to deliver artificial intelligence, I call it capabilities.
In many cases, it will be inference capabilities into the marketplace
that will be utilized ultimately by the software vendors
and ultimately by companies that are going to use it
to enable their infrastructure to be more efficient.
This is a very important aspect of what is the future of software,
what is the future of AI, what's the future of AI in software.
The way to think about it, the way we think about it is,
AI won't eat software, it will actually feed software.
And it will feed software so software can actually displace and augment services.
A big part of what's important is, there was a report by IBM two weeks ago,
less than 1% of the enterprise data actually exists in the public domain.
So these large-scale models have not absorbed that data and don't have access to it
because that is proprietary data that sits within the enterprise.
Much of that sits within enterprise software infrastructure.
And so the important part of being an investor in this space is understanding a few things.
One, do the enterprise vendors, enterprise software companies and companies
have sovereignty and dominion over the workflows and data sets?
If you do not, frankly, those companies don't really have a right to exist
because these very capable models, large-language models in this case,
will absorb this data and information and in some cases render less utility
to those workflows and data sets.
But if you actually have maintained that as a function of the business that you built,
the industries that you serve, the way that you have brought specific use cases
to those industries, that should give you a unique opportunity as an incumbent
to take advantage of this new technology.
So that's kind of point one.
The important aspect of this is how do you really think about how AI is going to evolve?
We've been using various forms of artificial intelligence across our infrastructure
for well over a dozen years.
The early states of artificial intelligence and machine learning,
focusing on specific algorithms, sets of rules, and actually leveraging that
to deliver predictive-type outcomes and predictive information or data sets
in the environment.
Well, the introduction of this generative AI, which actually came from some
of the early work that they were doing at Google, in essence,
starts to generate predictive next, call it next states.
So rather than as you're typing something and it's trying to predict
what the next letter is, it will now put it in context and say,
well, maybe let's start predicting what the next word is
or what the image should be, et cetera.
And as they started to put more capability, compute capability,
into that neural fabric, in essence, they found that it became much more productive
from a generative perspective.
We're now in the next phase of this, which is what we call agentic AI.
And basically, agentic AI has the ability to not only do the predictive
but actually use this and reason what should be an action
and actually take the action.
And so if you think about enterprise software,
in the construct of enterprise software, if you have agentic systems,
they are following and reasoning about how a customer uses a technology,
how a customer's actions should move to the next state.
If you identify an infrastructure, a workflow, and a data set,
utilizing specific data sets, you will actually be able to create working,
call it workers, who can actually do work, take action,
and you have to think about what kind of action you want them to take.
You also have to think about what actions you don't want them to take
that they may reasonably think is the right action,
but may not be the proper action for the enterprise.
And, Jay, it's important to separate what I call the consumer part
of this action versus the enterprise.
You and I today, after this, we say, hey, let's go have dinner.
We're going to invite some friends to dinner.
And we say, you know, let's go eat.
We're in Korea. Let's go eat Thai food.
So we say, book a Thai food restaurant.
Ninety-five percent of the chance that we will come out of there
will have a Thai food restaurant booked.
There's a five percent chance it will go,
and the restaurant actually has just a Thai dish, you see,
which in a consumer context isn't really a big problem.
And we call that those outputs.
I call it the fuzziness of the output.
They use the term hallucinations, in essence.
It may not give you – sometimes they're a little more dramatic than
Thai food on the restaurant.
It may have, you know, something in there that, you know,
there may be a Thai flavoring or something.
But that is a reasonable outcome that may come from a consumer output.
You, as you know, coming as a CIO of a very large corporation,
that five percent is unacceptable.
So in the enterprise, you have to have a different standard,
a different standard of output and a different standard of precision.
And so designing enterprise-agentic systems is, A, non-trivial,
but as you are able to develop it,
it actually creates a massive amount of utility for your customer base.
So that's the journey that we are on, A, as an industry
and one that we are leading as an organization at Vista
in building capabilities to create enterprise-agentic solutions
that will power our enterprise software companies,
those that we own and those that we will underwrite.
And we believe that's going to create a massive amount of economic impact
for our investors.
And actually, looking at the prior slide,
if you look at the projection of the industry going from hardware to cloud
to where we're headed into the future,
to me, when I come to Korea, many Korean institutions worry about
Korea's hyper-growth economy for the last 30, 40 years
based on their success in providing supply chain to the growing technology field.
As they look forward, as we move forward,
they need to figure out, they need your help,
figure out how to be dominant supply chain provider for this future of AI.
I really strongly believe that.
I agree with you.
I think there's an important part of what you said.
Korea has unique capabilities, unique insights
because of the great companies that you've built over 100-plus years.
Those companies actually have some unique workflows,
data sets, and insights of what their customers want
and that they uniquely provide.
And the question is, how do you identify those ecosystems that already exist
and also protect those ecosystems from being exposed to very large-scale models
which can potentially absorb that knowledge
and render it less useful for the companies that have built,
in some cases, spent tens of decades developing those insights.
Those are important parts of this process.
So one has to be thoughtful.
One has to be intentional.
But you also have to be measured in the way that you deliver it
so that you, I would say, bring the models to your data
and not your data to the model.
And that's an important aspect of how you organize yourself as an institution,
as an organization that is looking to take advantage of the capabilities of AI.
You and I use productivity AI today, and it improves personal productivity.
The real question is how we now agentify it in an enterprise
so that the corporations, so that the institutions that we are part of
capture the benefit of these new hyper-effective workforces called agents.
Productivity is not one of my next decade worries, but we'll do that on another day.
We'll keep going with your investment approach
and try to educate our audience more of it.
You've just given us a very clear view of agentic AI, what it is,
and its potential to reshape enterprise software.
So now I'd love to hear how Vista is leaning into this shift.
What's your strategy to position Vista for success
and how you're ensuring long-term value creation for your LPs in the next wave?
Great question.
The way that we have always approached our investment market is,
I like to tell people, we are not stock pickers.
We don't pick a company, hope the management team can figure out
what to do to continue to grow it.
What we look to do is garner sets of best practices and insights
from groups of management teams, codify them,
and then deliver those to management teams that may not necessarily have that capability
or need augmentation in their management teams and capability to enhance their strategy.
That's what we've been doing for the last 25 years to success.
We have a group called our value creation team.
That value creation team constantly evolves.
I'll give you an example.
Back when we were building our first factory to convert from on-premise to cloud,
I probably had 35%, 40% of that team focused on building the infrastructure,
the capabilities to do that at scale.
And I remember the first time we did it, it took about six months,
really more like nine months to actually make that conversion.
This was a new technology.
It was a new platform.
It was a new thing.
We had partnered with this young man by the name of Andy Jassy,
who was building out an infrastructure for AWS for a company called Amazon.
And he and I talked and said, listen, we'd like to use some of your capability
because they had compute available.
And so we were one of the first to actually adopt a strategic partnership with them.
And as far as I know, we have converted more businesses from on-prem to cloud
than any institution on the planet.
But we build a system to do that, which requires, I'll call it an ecosystem approach
to this work.
Well, as you think about today, we drove that capability through our value creation team
across all of our portfolio companies.
So even our smallest portfolio companies had access to this capability, this knowledge,
these best practices, the resources.
We're taking the same approach to converting businesses to agentic.
And so we started building a factory.
Well, let's back up about as this was first introduced,
we built strategic partnerships with OpenAI and Anthropic.
I think many of you know Anthropic has focused more on the enterprise than OpenAI.
But we built strategic partnerships and relationships with both of the executives,
the senior executives, both Sam and Dario, to also understand the capabilities.
And we typically get some early releases of the code so that we can test it
in our environments for the enterprise.
So that's an important thing, building out that ecosystem.
The next thing we did was we formed a Gen AI council to now,
which is made up of a number of our CEOs and outside advisors,
to actively think about the implementation of Gen AI solutions
and now agentic solutions across our portfolio.
And they meet quite frequently once a month to talk about, well, what have we learned?
How do we use this shared knowledge to enable all of the companies to progress?
You then have to set up what I'll call a series of internal KPIs.
One of the KPIs over a year ago was 100% of the company,
the early benefits of this was code generation.
These agents were actually able to generate code much more efficiently
than I'll call it in the early days, a standard entry-level,
kind of level one, level two type programmer.
They're now kind of level four or five type programming.
And so ensuring that every one of the companies is adopting code generation
at a minimum to enhance their capabilities to write code much more quickly,
much more effectively.
Another area that was important was using it for things like customer service.
Okay, how do we use these bots to actually deflect cost in customer service?
So that's another area.
So we, in essence, have driven all the companies to have 100%,
I'll call it, adoption of enablement for code generation, et cetera.
One of the other things that we found to be very effective
was to create a whole series of hackathons.
We've now had six in the last 24 months.
And these are just our internal companies.
Often we bring the senior development talent from some of the partners that I mentioned
to work with the companies.
In the early days, it was on code generation.
Now it's on agentic systems and agentic orchestration.
The last two of these hackathons were all about agentic development
and agentic orchestration.
How do you actually put together a series of agents to do a series of tasks
and organize them?
And the advantage is, of course, when you have breakthroughs
in certain of the companies, you can now share those with other companies
at a much more rapid pace with a higher utility associated with it.
So all that is a function of organizing ourselves to create AI enablement capabilities.
As we then have gone into that next state of this work,
we said it's now time for us to really build the next factory.
So as we started building that factory, we engaged with Microsoft.
And from that engagement, we now have had a number of our people work with them
to build and enable our factory to now convert all of the businesses
that can become agentic.
The first one, actually, Satya featured at his build conference a few months ago,
a company called Gainsight.
And now all of the companies are now moving through that process
of creating agentic solutions where we are now able to build orchestration capabilities
with agents to do actual work.
In many cases, those agents will have a human orchestrator.
In some cases, it is an agentic orchestrator.
And what we have learned, of course, I think is quite profound
because as you go through development processes, as you know,
and when you're the first one, still you will find and discover things,
new problems you now have to solve.
And we think by being a leader in that regard, we will be able to drive
the agentic enablement across our portfolio at a faster rate than others
and capture what I call more the economic opportunity than others will sooner.
So that's really the point of how we build our organization.
Wonderful.
You've answered some of the questions I prepared, so I was just trying to filter.
I read ahead.
I'm going to try to pick off a few things that I still have questions on.
So do you see potential in how AI can enable new revenue or driving cost efficiencies?
Maybe we can talk about that.
Absolutely.
I think there are a couple of ways to really think about this.
I think that enterprise software is going to go through a massive transformational change.
First of all, I like to think enterprise software, again,
many of these companies can have the potential to become agentic.
Some will be able to become what I call rule of 70, which I'll get to in a second.
And then others, to the extent you don't have sovereignty and dominion
the workflows and data sets don't really have a right to exist in a broadly defined agentic world.
So that's that.
Our portfolio today across our companies, we have about 750 million users of our software.
These are companies that are paying for subscriptions, et cetera, to use our software.
And those software subscriptions enable them to use the workflows
that generate specific data for their companies.
The way I like to think about it, most of the people in this room are knowledge workers in essence.
What a knowledge worker does is you're collecting data and information out of some system.
You wake up in the morning.
There's some system that you may log on to or frankly you're reading.
And then you reason that and you put that back into another system or another person
who may use it for another set of systems.
And so each of those activities is kind of a knowledge worker dynamic that occurs.
Well, if you actually follow you during the day and you start to create what I call discrete tasks
that make up your knowledge worker pathways, you can actually create specific agents to do those tasks.
And so there's a proliferation.
So every user, the work that they do can be broken up by a number of agents.
And so one user will equal a certain number of agents in essence.
Then you have to think about things we've discovered in building out our factory,
you know, what sort of administrative agents you actually need to have as well
as part of the construct of agents to complete those tasks to a high precision that an enterprise needs and or wants.
Our rough math today, which what we're experiencing is, you know,
of our 750 million end users of our software today will make up in this translation somewhere around the order of,
you know, four to eight billion agents plus a series of administrative agents
that also have to work alongside those agents.
Those agents perform specific work.
And you can actually value the work that it's done because you can compare it against what's the work that was done in the past
because these agents work 24-7, 365, have limited coffee breaks.
You know, they don't complain as much as other workers in some respects.
You have to monitor, maintain, and control them in certain paradigms in certain ways.
But they have a massive amount of utility.
And so what we are now finding is that if you look at the math on the left side, the revenue impact,
you can improve, call it the go-to-market productivity,
kind of the personal productivity of, oh, how do you, you know, what's happening with your customer base
and how do you call them at the right time and efficiently evaluate what it is,
the experience they're having so you can enhance that.
So there's that.
But there's another piece, what we call the new agentic offerings, which we're just now starting to size.
And let me give you an example.
When we did our earlier conversions from on-premise to cloud,
we found about a two-and-a-half times pickup of value that we could capture from our customers
in moving from on-prem to cloud.
Part of that was because they no longer had to go through hardware refresh cycles.
So the CapEx that they were spending for hardware recycle then became OpEx, in essence,
in the way that they're utilizing cloud.
So that's kind of point one.
Point two, they often needed yes, less, call it, you know, human oversight and utility
to manage each of these instances of hardware and hardware instances to support the application.
So we had about a two-and-a-half times pickup from on-prem to cloud.
But what we're seeing now, as we're now introducing these products into the marketplace,
agentic products with existing customers, is we're seeing 10x that in some cases.
So that utility is much higher because these workers can actually eliminate broad cost elements
of delivering a solution and expand the TAM, the Total Addressable Market, through these new offerings.
Again, what we're doing work today, 4x, 5x, some are 10x, some are 20x,
the utility that these new agentic solutions can deliver.
On the cost side, what we are seeing, and this is for enterprise software companies.
So to be clear, this is for software companies.
The largest cost component of software companies is typically R&D.
Makes up about 30% or so, in essence, of the total cost stack.
We are seeing, for the generation of new code, 30% and 40% productivity in writing new code,
much lower for existing code, but we have standards around that,
that over time should be able to reduce your R&D costs pretty significantly.
And so we're starting to experience that.
We're also finding that same sort of dynamic with customer support.
The deflection of calls, for instance.
If you have a call center that is taking a call, using these agentic solutions can deflect that,
often with a higher net satisfaction with the customer at a much lower cost.
And then, of course, back-office administration, go-to-market, etc.
So what we're experiencing now is the paradigm that companies can actually go from
what we call rule of 40, which is a combination of EBITDA, margin, and growth rates, to rule of 70.
They can grow faster with a higher profitability, utilizing these agents to augment
and enable these companies to be more efficient in the work that they do for the customers that they support.
That's the economic paradigm that we are now experiencing and witnessing,
and the key is to do that at scale in a way to capture more of that economic value for our customers.
We think about it with our own portfolio.
We think that the revenue potential is a multiple of our current revenue potential of our existing portfolio companies.
Okay. So maybe last question on your factories and what you've learned.
So how are you thinking about which companies can benefit from AI
and which stand to be disrupted in their characteristics?
Yeah. Again, we have a first-of-a-kind partnership announced by Satya at first of our AGM
and then ultimately at his Build Conference, and building that capability enables us to drive all of our software companies
through this process, the key learnings, the capabilities.
When you're in the early stages of development, you learn things that others don't,
which should enable you to be more effective long-term.
One of our companies, like I said, Gainsight, we now have co-developed an agentic offering for a customer.
Again, they focus on customer success and customer success representatives.
In that one in particular, we're seeing the potential of 10 to 20X times the revenue
from individual customers utilizing this solution set.
The other thing that came out of this approach was beyond developing the products for that particular company,
we have now developed and understood the componentry that we need for the rest of the software companies to make them factory ready.
And so now we've gone through a whole factory readiness dynamic and paradigm with all the rest of the portfolio companies
as they're on this journey as well.
I like to say stay tuned for the results, but what we're seeing is today the average enterprise software company that we own,
well, the enterprise software companies we own, the average ROI for the products that they sell to their customers,
so this is our portfolio companies, and we measure this pretty much every year, is about 625%.
So I like to say irrespective of what you're doing in your businesses,
typically the next best use of your next dollar is buy more software.
As we actually go from an enterprise software environment to an agentic enterprise software environment,
we believe it's going to be multiples of that.
And the early cases that we're finding in actuality that we are implementing in our customers is supporting this sort of an analysis.
So it's quite remarkable, and it's a quite remarkable technology, and it truly is a general purpose technology
when utilized effectively in an enterprise should deliver a massive amount of economic rent benefit
to the enterprise software companies that are positioned with workflows and data sets that are highly unique for the industries that they serve.
And that's how we're now looking at the underwriting of our companies going forward.
What are the companies that have that capability, that have those sorts of attributes,
that we can bring a set of best practices, which is now augmented on the best practice of converting from,
in some cases, again, SaaS or buy beyond prem, but SaaS, posted to agentic enterprise software companies?
And what's the value that we now then can convey to our stakeholders as a result of the conversion processes
that we're going to undergo with those businesses?
Okay, now we're going to take a step back and go macro.
So throughout history, transformative technologies like Steam Engine, the assembly line, the Internet,
have reshaped industries and propelled global economic growth.
How does AI compare to those technological advancements?
I believe it's interesting.
It is a general purpose technology like each of those.
You know, we've now gone through a very interesting state, you know, really as a planet and as a world.
The last 15, 20 years, we've gone through a digitization of infrastructure,
and that digitization of infrastructure has provided a massive amount of productivity.
That productivity, when operated in a globalization environment,
lifted a massive number of people in their economic standard of living, which we've all been beneficiaries of.
This newer technology has productivity capabilities of orders of magnitude above that.
And it will create some societal questions and challenges that we're going to have to deal with as humans on this planet.
But it also will probably solve a number of issues and give us capabilities that we didn't have in the past.
You know, our job as stewards of capital is to bring the greatest amount of utility of this new technology
into the companies that we own and into the companies that we're looking to underwrite
to create value for the pensioners, the investors who are our stakeholders.
And that's an important part of this.
I am more optimistic about what this technology can provide today than I was even, you know, six months ago,
because I can now see the utility benefits of what it will do.
We, of course, have to understand and figure out how these agentic systems, agentic workers will work,
how we will work with them, how we will enable them to be effective to deliver the values that we want,
how we can ensure in many cases that it delivers the precision in the environments that we care about.
But there are some bigger issues that, you know, our leaders and politicians are going to have to think about and figure out.
But I do believe that humanity and society can progress utilizing these tools as they had with steam and mechanization,
and there will be a reformation of new tools.
And I always think about these tools will enable us to, if we do this right, be more human,
focus on the human interactions and not just focus on, call it the data extraction and analysis,
and utilize this to actually create a greater benefit for society as a whole.
Yeah, I hope so.
So as AI continues to reshape industries and there's geographical aspects that you talked about,
so there's very specific pointed questions maybe.
Which countries or region are best positioned to lead innovation and adoption,
and how might differences in infrastructure policy, talent, ecosystem influence the global balance of power over next decade?
I think it's going to be interesting to see how this evolves in different places.
The U.S. today seems, at least our largest tech companies are focused on achieving states of AGI, SGI, et cetera.
Other places that I go in the world seem a little more focused on the utility and application of this technology
into existing experiences or for the benefit of their health care or their citizenry, et cetera.
The fact that we have distributed compute enables many, if not all, many communities to benefit from this new technology,
because we've been distributing compute for some period of time.
We now are seeing, like I said, $1.8 trillion, at least that's our last measurement,
being driven into the build-out of the infrastructure for GPU capacity that ultimately will be accessible at some price.
I think that will enable many policymakers to think about how to utilize this, I call it, in a beneficial way for the economies that are in.
But it is interesting, you see different priority sets geographically and politically about the utilization of this technology.
I just hope that there will be a little more thoughtful engagement about,
okay, what are the impacts that we have to think about in humanity?
I know there are certain leaders who are thinking about it and talking about it, et cetera.
I just hope we advance that at the same rate that the technology is advancing,
because I think we as humans are going to benefit if that happens effectively.
Well, let me also, again, reiterate and add that as Korea moves to, Korean economy moves to,
has huge success in being part of the U.S. supply chain of technology.
And as we move to AI, please keep an eye on the Korean companies.
I will tell you, so here's the interesting thing.
Being now, again, part of the COSA for a number of years, you have a wonderful, I'll call it, engineering dividend.
Maybe not as large as other countries, but as a percentage, you actually have a fairly high number of engineers,
very capable engineers who can help and be a part of the transformation of existing systems to agentic systems.
Again, each of the companies will have to think about how they utilize it.
Institutions and the government have to think about how they utilize it.
But the good news is you actually have a very capable group of people who have been trained
and have built infrastructure to continue to train and take advantage of this engineer dividend.
As engineers, we love more engineers, right?
They'll have to probably learn and study a little differently.
But I think that's a great advantage that Korea has, that they should obviously look to take advantage of
and utilizing this for not only development of commercial products,
but also to enhance quality of living in things like agriculture, in things like healthcare,
societal benefits in relating to civic opportunities.
And I think the good news is you have people who are capable of doing that.
There's other countries that don't have that engineer dividend.
That's going to take them a little longer.
They're going to have to rely on other parties and other groups and other countries,
in many respects, to capture that knowledge and capability.
So Korea's in a good place to take advantage of that.
I fully agree.
Very adaptive people.
5,000 years of history has proven that.
But we are at a crossroad of trying to get on a different conveyor belt, too.
Yeah.
But your leaders in the past have done a good job at least getting you to a place
where you now have the institutions that develop great engineers and not only the universities,
but actually the companies.
The company has been very engineering-centric.
And I think that's a very important part of this dynamic of progress, utilizing this new technology.
So I think you should absolutely embrace that wholeheartedly.
Right, right, right.
Yeah.
Thanks for your leadership on that issue.
Sure.
I'm cheering for Korea.
So let's get into much more lighter topics as we close down for the last couple of minutes here.
Okay.
On a lighter note, amid all the disruptions and uncertainty, what signals are giving you hope?
Where do you see optimism emerging?
And whether in markets, technology, or human spirit, you kind of touched on it already.
Yeah.
The thing that gives me the most hope, I always tell my team, you know,
honor the opportunity for the work that we get to do.
Part of the work that I get to do are things like this.
And what I mean, not just be on stage with you, and I do honor that.
Okay.
Thank you.
But it's to come and interact with, you know, the people in the Korean government,
in the companies that we work with, the institutions, organizations,
because I see that optimistic spirit of let's embrace this technology to make things better.
And I get to do that all around the world.
And by and large, most of the people have that spirit.
So I'm actually encouraged by that.
And I think, you know, I'm chairman of Carnegie Hall.
And part of what we like to do is bring our National Youth Orchestra around the world.
And this year we were in Asia and we were in Seoul just recently.
And bringing the youth and to see that youthful exuberance of interacting with each other,
you know, the orchestras playing with each other, playing for audiences,
that's the fabric I think that's important that will keep humanity bound together
and moving forward in a positive way.
So we have to continue to encourage that, continue to encourage, you know,
student exchanges, continue to encourage our students interacting
and having opportunities to work together, develop together, build together,
and get to know each other and get to know culture.
I think that's an important part of what we need to do.
Right. I love that.
So my last question is along the similar line
and for maybe some of the younger people in the audience.
You have a beautiful family.
Oh, thank you.
You have kids, children who are growing up in this rapidly evolving industry,
AI power world, right?
What guidance would you offer them in terms of how they begin their career
and are there any mindset skills that you would believe would be helpful for them?
The most important thing I tell my children is get off the payroll, your dad's payroll.
Sorry, touches part of me.
Yeah, in all sincerity, I tell them, you know, embrace the tools that are available to them.
Become expert in utilizing them to deliver the value of the things that you think
will improve the human condition for the communities that you care about.
I have three rules for my children.
The first one, I tell them that they are enough.
They are enough to actually go and embrace, you know, their life
and make changes and change the condition of not only their family
but the communities they care about.
The second is I want them to discover the joy of figuring things out.
There's a great joy in being an engineer to solve complex problems,
to figure out those problems and learn how to do that.
And the third thing I tell them is love is all that matters.
And so I hope that they adopt those three rules to chart their course in their life
to be great contributors to the human society.
Right. Thank you.
And just for the audience, I think, Robert, you've done a wonderful job
committing to philosophical endeavors.
And I won't read it all out. It's a long list.
Well, thank you.
But I really respect that.
Thank you very much.
I appreciate it.
Thank you for welcoming me and my team to Korea.
It's been an absolute delight.
Thank you.