Ashutosh Desai: On the Shoulders of Capitalism | Intelligence at the Frontier 2026

Transcript

Um, so the name of this talk is on the shoulders of capitalism, which is kind of standing on the shoulders of giants. What happens next? Uh, this is going to be very quick. I apologize in advance for uh, speaking fast. Um, so I'm a Gujarati American. I grew up in Asia about till I was about 10, came here, uh, went to schools in the Bay Area, and I've had the privilege of working with some really awesome organizations, including Make School, the one that I founded, a university that we tried to build here in San Francisco, um, and have had a lot of window into technology in Silicon Valley over the years. Um, I want to start with a little bit of history from where my family is from. So, I'm a Naval Desai, and we all come from this like tiny little village right below Surat. Uh,

Surat in 1608, uh, the British East India Company landed, um, and had many, many years of, uh, taking control over the country, um, through capitalism. The British East India Company was the second shareholder traded company in the world. The first one was the Dutch East India Trading Company. And so, this inherently was very linked to, uh, what it means when you start investing in companies and promising forever returns and how that impacts populations over time. Um, the Great Bengal famine of 1770 was one example of this. Uh, the reason we couldn't grow food for a lot of people, uh, 10 million people died, is because instead we used that land to, uh, farm opium that we sold to the Chinese. And so, that was one big impact of capitalism and sort of this trying to

create more profits going very terribly wrong. Um, Regulation 9 in 1810, uh, India stopped being able to trade internally with each other. And so, the or the price of everything became really, really high because you just can't trade with your neighbors anymore. You have to trade through the British, send things away, have things come back. This decimated the entire economy of India. Um, prior to British colonialism, uh, there was, you know, 40% of GDP was held in China and India. Post-British colonialism is about 5% um, at the time of independence. Uh, the Bengal famine in 1943, uh, 4 million Indians died during World War II. Uh, Winston Churchill was responsible for this. Um, almost as many as Jews died in the Holocaust, but this is not something that we look at uh in

the same weight. Um, here is a picture of a lot of loom workers uh from uh Kerala. Uh, the loom was one of the inventions at the time uh that represented the Industrial Revolution. Um, this was something that created a technology that enabled us to produce our main good uh much more efficiently than before. Um, we're seeing that today again with AI. Software engineering and many other disciplines we are now able to do very cheaply and efficiently um in ways that weren't possible before. Um, what's really remarkable about these photos for me is that this shows that our bodies get shaped by the technology as much as we shape the technology. And as we think about AI, this is a very

important thing to think about is how we will be shaped by this technology. Um, what happened in my lens is that technological advantage enabled capital to out-leverage both labor and governance. Right? We've talked a lot about capital versus labor. Governance is another important thing. And there were many people in India who effectively allowed themselves to be succumbed to the capitalists, right? The British control of India was enabled by Indians. The Mughal control of India was enabled by Indians. My last name is Desai, landlord tax collector. So, we were the class in India who enabled the the Mughals to take over the society. And so, as we think about our roles here today, are we going to be the ones who enable

certain types of caste systems to be rebuilt with this new revolution? Um, I believe today, looking at this, similar to back then, the American military-industrial complex is the greatest risk to AI safety. We are at this moment in this precipice that if we do not put our military uh in control and actually use it as defense and not war, uh this is going to be the the greatest difficulty of this technology and it will last for many hundreds of years. Um the Silicon Valley rebranding of this is American dynamism.

We're already seeing it today. Why are we on track for this? Why are we on track for what I would call and what some of some of the folks that I've worked with in the past would call colonialism 2.0. Um I think we are all rational actors with kind hearts. I grew up here. I grew up with a lot of people um who are now in positions of power that I think are are making grave mistakes. Um but I don't think it's because they're acting irrationally or they're stupid or anything like that. I think we're in a prisoner's dilemma and in these systems it's very difficult to find the greater outcome for all of us. Um I think there's a few reasons for this. The penalty for dishonesty is lower than for uncomfortable truths. A lot of my friends are afraid to speak truth against power. A lot of

Republicans in Senate are afraid to speak truth truth against power because there's a great grave penalty. But the people who are lying to our faces are don't are not facing those penalties. Um sycophancy is something that is both a problem in AI systems but also with in AI systems it's when they give you the sugar of what you want to hear as opposed to pressing you to challenge your thinking. Um and ultimately this leads to a regression of the mean of outcomes. You know, most politicians are corrupt and if we keep being yes-men and allowing them to do so, this will continue to happen. Um when I looked at YC we funded you know uh 200 companies in a year. Uh a lot of it's experiments. You know, there's less of this founder brilliant idea. It's more of like let's run a lot of different experiments in a lot of

different directions and see what works. Um and as a result I kind of think that this sort of profit-maximizing capitalism without you know, regard for society, without really care about the impact of on the customers but really just has IRR generation. Um effectively this becomes a loss function uh just like in in machine learning algorithms for the recycle rate of revenue which is how quickly can you take in revenue then reinvest in the business for profits and continue to have that cycle. Um some data that I did after YC is I studied the the profits of for and market caps of Fortune 500s from 19 uh 50 50 to today um and I looked at how how that industries have changed over time, what the founder stories were, um, and the big thing that I noticed is is really what you're starting to see is that the

lowest marginal cost businesses are the ones that end up, effectively, oligopolizing most of the profits in the Fortune 500s. Finance and banking, obviously, very, very low marginal cost. Computing silicon is kind of the next. Um, pharma is another one that has very low marginal cost, but really that this profits sort of centralization into finance and banking, largely run by about eight uh, banks. Um, this sort of trend, I think, is actually more of like a this is the natural course of order that, uh, if you have low marginal costs and high recycle rates of revenue, this will kind of, uh, naturally happen. Um, so, effectively, a lot of the the US, as we've seen, has started to oligopolize profits in the entire economy to a small small thing. The other big effect here is you have, you know, a perfectly competitive market, each player has an

equal shot, whether you're a small or you're big. The big stack bully effect, imagine five people at a poker table, one person starts out with 10x the number of chips, you can push someone all in four times over the course of a game. He's a better player, he won three out of four times, but still, you do this eight times, you're going to knock out almost every other player out of the game. It's very, very difficult to compete when you're competing with, you know, this war chest, um, in your in your in your competitor. Um, we have a rapid acceleration towards a K-shaped economy here in the US. Uh, this is something that exists in India, you know, there's a class of people where everything is done, like life is good. If I were to move there, I would everyone would take care of me all the time. I don't think

that's a good thing, um, but this is pretty much what's been happening. Real wages have not been rising. Uh, we have this stark sort of separation between GDP growth in the US and actually, like, uh, people's lives being better. Um, the big things, you know, the cost of almost everything, CPI has come down except for education, uh, which is where I used to work, housing and healthcare. Um, those have skyrocketed up. So, as real wages for the bottom 60% percent of Americans have stayed flat, those three have shot up to a point, um, where it's really unsustainable and we're moving towards sort of a a more feudal system. Um one thing that I observed at YC as well is social signaling drives founder behavior a lot more than ethics and first principle. There is so much group think. Uh the number of times that VCs

actually look at a company's operating model or even like challenge a a founder on their sense of unit economics uh is very very low and this stuff happens within, you know, 3 days for the best deals. You don't have time to diligence those deals and the 3-day fund uh funding rounds happen through social signaling more than anything else. when people give up hope on a just world, which I think is starting to happen when I talk to people who are starting to have funded events companies and so on and so forth, they they effectively are a little pessimistic on the fact that we can save America. They kind of look at this as we're at we have a declining empire and this is going to decline, so I might as well take what I can for me and my family. Um when people give up hope, greed prevails and I think

that is what we are seeing in Silicon Valley today. I want to pause for a second and take a quote from MLK at the Berlin Jazz Festival in 1964. Jazz speaks for life. The blues tell the story of life's difficulties. They take the hardest realities of life and put them into music. Only to come out with some new hope. So, what are my hopes for post-capitalism? I want to build towards a Star Trek future and not a Snow Crash future. Snow Crash famously is this everything is commoditized and commercialized and you can't really go anywhere without being a member of paid member of some club. A Star Trek future is one that I think preserves more agency and autonomy and sovereignty of many different people living many different styles of lives.

Um we have many different examples in sci-fi that we can draw from. What we choose to build will get us there. Um avoid doublespeak, speak truth to power. This is probably the the most important thing that each of you can do today, tomorrow, for the next few weeks to help with this. Um we can explore third way ideas. Uh I helped pioneer income share agreements and did the first financing deal for income share agreements. They're now a regulatory approved instrument in the state of Illinois uh 10 years after or 15 years after we started working on that. PBC, public benefit corporations, it's kind of this middle ground between a nonprofit and a for-profit. Um, there's weird ideas that maybe don't make sense, but like in the '60s in India, there's progressive utilization theory from a class of thinkers who just really didn't like uh both um

both capitalism and and socialism. Um, look at approaches like Aadhar instead of uh Worldcoin. Aadhar is a very sovereign, open infrastructure that lets governments build their own uh infrastructure um for uh for uh payment rails, um and it's been basically the thing that has sort of digitized the entire economy. You you can use more digital payments in India than you can in the US uh for anything like a banana. You can even like scan a QR code for people begging sometimes. Like it's it's really a wonderful system um that's happened. Um, long-term funds, CAPEX and OPEX debt funds, like we try to do the ISAs. A lot of climate initiatives are really stuck right now because there is no sort of CAPEX fund and all the things like frontier carbon are basically only paying on start carbon sequestration delivery. Venture's not a good vehicle for CAPEX. Um, competition is for

winners, not for losers. Competition is for winners. We become better with competition. That's how we make better cars. That's how if we open up to more competition, we would have more companies like BYD and our car companies would have to compete with companies like BYD who are making frankly better cars than Tesla. Um, this is a very important one. We have to imagine BRICS as an ally and not an adversary. I grew up in China and India. There are good good people. The bulk of the people are good people. We may disagree with how they govern things, but for AI to be win-win-win, we can't just be fighting each other. We have as they are part we there are humans in our world that we are trying to help come out of poverty and succeed. Um, one of the people I respect most who is

working at Alphabet/Google, um, James Manyika, he was one of six kids bused to school in apartheid Rhodesia uh and with like police with rifles um to go to the the all-white school. He then got to Oxford, Geoffrey Hinton's lab, um worked at McKinsey for many years, now now at Google, and he he's quoted in a in a video saying, "My friends laugh because I was not the smartest kid in elementary school. Intelligence is distributed so much, and our system right now is only allowing a few people to to access it." Keep San Francisco weird. We had the Internet Archive here earlier, Monkey Brains, Exploratoriums. Like, these are the types of things that make San Francisco San Francisco, not just making money off of Silicon Valley.

And why I think this is important is like Monkey Brains is a great example of you can build a really great product and not try to conquer the world. And you can do it well, and you will outcompete the people who are trying to do it. There's so many basic internet services. We still don't have a good dating app that works for people. We still don't have a good blog. Every 3 years there's a new blog. There's not a one blog platform. We could have had this since the '90s. So, like, build something great and just like have it last last for a while. And the last thing I'll say is leverage AI to explore the vast expanse of latent space. So, the way these algorithms work is effectively it's doing next token prediction and predicting the most likely next token. And so, if we do

that, we sometimes get what's called a regression to the mean where we just kind of pull everything to what's the most likely outcome. The four-chord song is kind of what your If you If you just looked at what the most likely next chord is. And this is an example of a demo one of my team members built where they took all of latent space of music from one of these models, and they put it into three dimensions, and you're flying a spaceship through that latent space. And so, instead of you being limited to what the model tells you, you can drive anywhere you want in latent space and create absolutely new things that are really, really cool. And so, that's what I would say is like when you as you think about AI, don't get the average thing, don't get the first answer. How does this let you explore

the world and build branches of new things far outside of the mean. Yeah. That's it. >> Thank you.