a16z

Ben Horowitz on AI Anxiety, Big Tech Transitions & The Future of Startups | a16z

2079 summary words 9 min summary Watch video

Start with the signal

9 min read

Summary

At-a-Glance

  • Verdict: Watch fully
  • Core thesis: AI changes the economics of software, competitive moats, and industrial capacity so radically that incumbents must re-evaluate what they truly own, while the resulting trust, infrastructure, and agent-economy gaps create major startup and investment opportunities.
  • Why it matters: The conversation directly connects AI disruption to durable-moat design, cryptographic identity and authorization for agents, and the physical bottlenecks—power, memory, manufacturing, and grid hardware—that will constrain AI deployment.
  • Best use: Use it as a strategic framework for assessing AI-exposed software businesses and for identifying control-plane, provenance, payments, and infrastructure opportunities rather than as a tactical implementation guide.

Executive Summary

Horowitz argues that AI breaks two long-standing assumptions of software company building. First, money can now materially compress software development timelines: with capital, GPUs, and useful data, firms can solve problems that previously could not be accelerated merely by hiring. Second, conventional software lock-in is weakening because code and data are easier to replicate or move, and future users may be agents rather than humans tied to a familiar UI. A legacy company therefore cannot rely on its installed base or product surface; it must identify a more distinct source of value and price against that value.

He does not conclude that every incumbent SaaS company is doomed. Instead, he distinguishes companies merely losing investor confidence from companies whose customer demand has actually shifted away. Navan is his example of a business with defensible operational complexity: global travel requires supplier relationships, integrations into enterprise budgeting, and a specialized distribution channel to travel managers. The implication is that AI can rapidly commoditize features, but not necessarily a product or company built around hard-to-replicate relationships, workflows, channels, and real-world execution.

On the supply side, Horowitz sees AI as an industrial buildout rather than solely a software cycle. He argues that the United States lacks enough power, rare-earth inputs, manufacturing capacity, memory, and grid equipment to support demand. His central investment lens is to identify the active bottleneck at each layer of the AI supply chain; chips may become available before electricity or memory does. This motivates a16z's larger fund base and investments in physical infrastructure such as power transformers.

The most actionable systems argument is that AI-generated impersonation, spam, fraud, and autonomous economic activity will require cryptographic infrastructure. Horowitz expects proof of humanity, identity verification, signed content, hardware-rooted authorization, recipient addresses, and internet-native payments to become essential. He frames blockchain not as a speculative add-on but as a potential trust substrate for a world where neither messages, video, calls, nor agents can be assumed authentic.

Key Takeaways

  • Claim: AI invalidates the old belief that software execution cannot be bought with capital, making competitive catch-up and feature replication materially faster. | Evidence: Horowitz contrasts the traditional 'mythical man-month' principle with an AI environment in which a company with sufficient money, GPUs, good data, and talent can solve many software problems by scaling compute. | Implication: Ken should assume implementation velocity is a weak moat in AI markets and evaluate whether an agent or product has privileged data, workflow ownership, distribution, trust, or operational capabilities that cannot be purchased and copied as easily. | Caveat: Capital alone is not presented as sufficient; Horowitz explicitly includes good data, and he later stresses that quickly built features are not automatically viable products or companies.
  • Claim: Legacy software lock-ins—migration pain, customer data, and human UI familiarity—are weakening because agents can use interfaces flexibly and make switching easier. | Evidence: Horowitz says software code is easier to replicate, data is easier to move, and future software users may be AIs rather than people attached to an existing user interface. | Implication: For AI-exposed portfolio or operating-company assessments, separate superficial product/UI lock-in from embedded workflow, contractual, channel, network, and real-world operational moats. | Caveat: He does not claim all lock-in disappears; his Navan example shows that supplier relationships, enterprise integrations, and hard-to-reach buyer channels can remain meaningful defenses.
  • Claim: The relevant question for an AI-threatened incumbent is whether its underlying business is strengthening or degenerating, not whether public-market valuation has collapsed. | Evidence: Horowitz says a company whose customers have shifted spending to other products likely needs deep cuts and a pivot, whereas Navan's travel business retains value through global airline, hotel, rail, and budgeting-system relationships plus a channel to travel managers. | Implication: Use customer-demand movement and retention of hard operational assets as leading indicators; do not treat the broader 'SaaSpocalypse' narrative or multiples alone as a diagnosis. | Caveat: The durability of these defenses is time-dependent: Horowitz acknowledges that agentic travel may become easier than it is today.
  • Claim: AI investment opportunity will be constrained as much by physical infrastructure as by models, with bottlenecks likely rotating from chips to memory, electricity, and grid equipment. | Evidence: Horowitz cites shortages in U.S. electricity, rare earths, manufacturing capacity, and memory; he says a new DRAM factory could take five years and notes an investment in an actual power-transformer company because transformer technology and manufacturing need improvement. | Implication: Investment and strategic planning should map dependencies end-to-end—generation, transmission, transformers, data-center buildout, chips, memory, and inference location—rather than extrapolating from GPU demand alone. | Caveat: He compares the moment with the late-1990s fiber buildout but distinguishes it by saying current GPUs are actively used rather than analogous capacity being largely idle; the exact bottleneck sequence remains uncertain.
  • Claim: AI-generated impersonation makes cryptographic identity, provenance, and authorization foundational infrastructure rather than optional security features. | Evidence: Horowitz imagines an AI deepfake of himself on Zoom instructing a finance team to wire $500 million to Nigeria; his response is hardware-rooted access and refusing to trust purported messages without his cryptographic key. He also predicts a need to prove humanity, identity, and the origin of signed media. | Implication: For agent systems, high-value actions should require verifiable identity, signed instructions/content, hardware-backed credentials, and explicit authorization policies rather than trust in voice, video, email, or UI presence. | Caveat: He presents blockchain as the likely trust substrate based on mathematical and game-theoretic properties, but the transcript does not establish that it is the only technically or institutionally viable solution.
  • Claim: AI agents will need internet-native economic rails, creating a crypto-adjacent opportunity in identity, payments, fraud prevention, and machine commerce. | Evidence: Horowitz argues that CAPTCHAs are obsolete, notes estimates that roughly $450 billion of stimulus funds were stolen, and asks how an AI can receive money or operate as a merchant when it is not a human or conventional credit-card merchant. His answer is a bearer instrument and addresses on the internet. | Implication: Monitor and test primitives for agent wallets, scoped spending authority, verifiable recipient identity, payment settlement, and fraud controls—but design around unresolved compliance and accountability constraints. | Caveat: The discussion is directional rather than a product specification; it does not resolve legal liability, consumer protection, KYC, custody, or regulatory treatment of autonomous agents.
  • Claim: The venture and labor outcomes of AI remain path-dependent: concentration around giant labs and utilities is plausible, but so is broad entrepreneurial expansion enabled by cheap creation tools. | Evidence: Horowitz compares one scenario to the Industrial Revolution, where early industry consolidated and financiers evolved into banks, against another in which advanced intelligence becomes a utility that everyone builds upon. He also argues that AI lets billions of people express ideas in code, music, and film without prior capital or institutional gates. | Implication: Avoid basing strategy solely on either a centralized-lab thesis or a democratized-builder thesis; maintain exposure to both platform-dependent businesses and edge/distributed creation models. | Caveat: He explicitly declines to predict the outcome, citing uncertain electricity constraints, possible edge deployment of smaller models, and the unusually dynamic nature of the transition.

Detailed Brief

Feature, product, and company are becoming harder to distinguish

  • Claims: AI lowers the cost and time required to create functionality, increasing the frequency with which apparent startups are revealed to be replicable features.; The traditional hierarchy of feature, product, and company still matters, but the boundaries are unusually confusing in the current environment.
  • Evidence: Horowitz invokes comparative advantage: companies historically chose not to build every adjacent capability because building a feature took long enough to justify specialization.; He uses the phrase that the best companies have 'hostages, not customers' to emphasize the importance of durable dependence rather than mere adoption.
  • Caveats: The hostage framing is intentionally provocative and should not be interpreted as a substitute for measuring customer value, satisfaction, or ethical switching costs.
  • Implications: In diligence, force a concrete answer to what persists after a capable incumbent can recreate the visible feature set and import or reconstruct relevant context.; Treat the ability to produce a polished demo as weaker evidence of company formation than proof of repeatable acquisition, retention, embedded workflow ownership, and defensible unit economics.

A non-dystopian transition narrative, with real uncertainty

  • Claims: Horowitz argues that technological transitions are frightening largely because current jobs and needs are visible while future occupations and desires are not.; He expects AI to increase productive and creative access, potentially allowing people globally to create software, music, and film without traditional capital or institutional gates.; His long-run view is optimistic: he expects living standards, luxury, and access to information in roughly 15 years to exceed what even the best-off people experienced in 1980.
  • Evidence: He notes that roughly 93-94% of Americans were farmers around the founding era, whereas farming is now a tiny share of employment.; He criticizes Keynes's expectation that abundance would reduce work to about 15 hours weekly, arguing instead that humans quickly turn newly possible wants into perceived needs, products, and industries.
  • Caveats: The historical analogy supports an argument about eventual adaptation, not a guarantee that transition costs, labor displacement, or gains will be evenly distributed.; The transcript offers no policy mechanism for handling the transition, despite discussing UBI fraud and public anxiety.
  • Implications: For company and investment narratives, frame AI around expanded agency and new demand creation while retaining explicit plans for trust, affordability, and transition risks.; Do not assume existing job categories are the durable unit of analysis; look for emerging needs and services that AI-enabled abundance creates.

Notable Concepts & Terms

  • Mythical man-month: The classic principle that adding people to a late software project does not linearly accelerate it; Horowitz argues AI/compute partially overturns this constraint.
  • Feature → product → company: A diligence hierarchy that becomes harder to assess when AI makes feature creation cheap; durable companies require more than functionality.
  • SaaSpocalypse: The market narrative that AI erodes SaaS terminal value by weakening software moats and enabling rapid substitution.
  • Hardware root of access: Using hardware-backed cryptographic credentials as the basis for authorization, intended to prevent deepfake or impersonation-based instructions from being trusted.
  • Hashcash: An early proof-of-work anti-spam concept that Horowitz sees as newly relevant when AI makes personalized spam and automated outreach nearly free.
  • Proof of humanity / proof of identity: The need to distinguish humans from bots and verify a specific person's identity in social, communications, payments, and agent-mediated systems.
  • Bearer instrument / internet money: A payment primitive that could let autonomous AIs receive, hold, and send value without relying on human-centric credit-card or banking infrastructure.
  • Bottleneck mapping: Horowitz's infrastructure investment approach: locate the currently binding constraint across the AI stack rather than assuming GPUs are the sole limiting factor.

Operator Notes / Why Ken Should Care

  • Add cryptographic signing and hardware-backed credential verification to the approval path for any agent that can trigger payments, data exports, account changes, or external communications.
  • Define an explicit authorization model for agents: principal identity, delegated scope, transaction limits, expiration, audit trail, and revocation mechanism.
  • Create a moat-review rubric for AI products that scores workflow embedding, proprietary/reliably refreshable data, distribution, buyer relationships, contractual access, operational execution, and switching costs separately from model and UI quality.
  • For infrastructure or investment theses, maintain a live dependency map covering power availability, interconnection/transmission, transformer supply, memory, chips, and data-center deployment timelines.
  • Avoid using CAPTCHAs or unverified video/voice/email as sufficient evidence of identity for sensitive workflows; establish out-of-band cryptographic verification requirements.

Source/Metadata

  • Title: Ben Horowitz on AI Anxiety, Big Tech Transitions & The Future of Startups | a16z
  • Transcript words: 5651
  • Duration seconds: 1730
  • Timestamp note: No timestamps or chapter markers were present in the supplied transcript.
Search this video Resume transcript at 0%
Full transcript 5005 words · 25 min read
0:00

America's got to rebuild its entire infrastructure right now. We don't have enough rare earth minerals. We don't have enough electricity. We don't have enough manufacturing capacity. NVIDIA will make enough chips, but then we won't have enough memory. Almost everything is the bottleneck. The China graph is this, and the US graph is that. How do we make this seem less scary? The history of technology is things have always gotten better. Humans are unbelievable in their ability to come up with new things that they need. Now 8 billion people that might have an idea in their head can get it out of their head. I do think what's going to happen is...

0:39

So you've been doing this for a long time, and I thought maybe I'd start off... It's funny, we actually didn't rehearse this at all, because I thought that way would be more real, more unique. But let's talk about... You have this book where you talked about how hard it is to be a CEO and everything that you went through at Loud Cloud and Opsware. That was a giant shift where the market collapsed. The financial market collapsed. And you had to really pivot and just change the company. And what do you think... There are new age companies that are popping up right now. AI first. They hopefully have their shit together. They're off to the races building something new.

1:13

But a legacy company of five or ten years ago, where there's this great opportunity but also great challenge, what does a five- or ten-year-old CEO do? Where they're pre-AI. So they've got to figure out what they do. So the financial markets hate them. Yes. So the financial markets hate them. It doesn't matter who you are. Yes. So I don't know, maybe riff on that. I'd love to hear your thoughts. Yeah. Well, I think the first thing you have to recognize in a huge dislocation like this is some very basic axiomatic laws of physics are different.

1:37

And the two that are really different with AI compared to how companies have been built in technology forever is, one, it used to be very well known that you cannot throw money at the problem. So, for example, if I had a product and I was two years behind, I could not hire a thousand engineers and catch my competitor. It's a mythical man-month. Nine women can't have a baby in a month. Everybody knows that. It never works. No problem. That's no longer true. You can throw money at the problem. If you have enough money and some good data, you can buy enough GPUs and solve basically anything in software. So that's gone.

2:21

The second thing that we knew for sure is in software, possession is nine-tenths of the law. So if you have the customer, you have multiple lock-ins. You have the migration pain lock-in. You've got the data lock-in. You've got the user interface lock-in. Those are pretty much gone, right? So it's very easy to replicate the code. It's very easy to move the data. And then it's not even going to be a human talking to your software. It's going to be an AI. And AIs are really flexible on how they use user interfaces. So that mode is gone. So I think that's the first thing you have to recognize as a CEO, that, okay, that's going away. So then what is it? Where is your value?

3:02

What are you delivering? And there, it turns out there are many things that are of value. But if you're trying to get good pricing through any of those things, you're going to be under tremendous pressure. Your price has to be a function of some other value that's much more distinct that you provide. Got it. And the other thing is, we've talked about this a lot internally as a firm, that once upon a time, if you have a good product, you might have 10 years to run with that product. Maybe five years. And now it might be five weeks. Well, we also talk about this in terms of going public. So companies are staying private a lot longer, which probably is good.

3:30

If you're going through an existential crisis, you'd much rather do that as a private company than a public company. But also the reason why the SaaSpocalypse is happening is because there are doubts on terminal value. Yeah. Right. So everybody who starts a company, they're doing it because they want to create economic value. They're capitalists. They're trying to actually benefit from this equation financially. But if you wait too long, maybe your company is worth zero. That's scary. And that was always a risk, but it would play out over decades. Yeah, it's not as fast a risk. So I guess what do you, if you were, I mean, LoudCloud's around today, you're the CEO.

4:14

And again, bad example. Sorry to give you. It's very scary. I know. Although actually LoudCloud would be awesome. Actually, yeah, exactly. You would be very well positioned. But I guess what is it that a CEO should do potentially differently? I mean, obviously, move faster, cut faster, be more efficient, throw money at the... all these things that we've talked about, but it's like, shit, if I don't go public, if I go public and I get disrupted, then I have this terrible life of I'm going to be a penny stock. Yeah. If I just wait, there's this chance that I get eviscerated. And this roadkill-or-success equation is scary, right? I mean, it's always scary. Yeah.

4:57

You would have time. And now it feels like you don't. Yeah. I think you do have to be honest with yourself on what it is you have really. And there are companies that get thrown under the bus correctly and ones that don't. And then if you take a lot of these ideas to their logical conclusion, then nothing is worth anything because there are no people at companies. And if there are no people, who's going to buy your shitty software? But it is more subtle, and it just tends to take much longer than we think for some of these things to play out. So then the question is, are you getting stronger in that meanwhile, or are you degenerating?

5:21

So is what's happening nobody's buying, the money just shifted, the customers are buying other stuff, they're not buying yours? Because in that case, you have a huge problem. You probably have to cut deeply and pivot. On the other hand, look, there are companies that have been slaughtered in the valuation game but are pretty strong. So I'm on the board of this company, Navon, right? Their travel. So obviously the SaaSpocalypse, they're dead, no way you're doing travel. But then you look under the covers and you go, well, it actually is a little more complicated than that because on travel, you actually need explicit relationships.

5:34

If I'm providing your travel and you're any kind of company that's important at all, you need to travel globally. So now I need a relationship with every single airline in the world, every single hotel in the world, every train, everything. You've got to deal with that. You've got to connect back to their budgeting systems and all these things. And then the second thing is nobody wants to do, including OpenAI or Anthropic, is sell to the damn travel manager. Nobody has a channel to the travel manager. It's not something, and you can't even imagine that being a good idea. You want to keep advancing.

6:18

You want to do the things that Intuit is doing where, okay, turn ourselves into more of an AI company and then hold the customer. And by the way, the AI, the agentic travel experience, turns out to be much more complicated than one would think. And I don't know if it stays that way, but that's the way it is today. So I think it's very company-dependent. I don't think it's all one thing, but I do think brave new world. And if you keep looking at it like the old world and it's got completely different laws of physics, you are definitely going to die. Yeah. Well, maybe let's talk about venture capital. There's a lot of cope going on now, too.

6:42

So you got to be careful with that. Well, that's the thing. There are some things that really are features. And before, it would take a long time to build a feature, so you might as well, it's comparative advantage. David Ricardo. I could weld my own steel. I could grow my own food. But I'm just going to not do that because I can do things that actually produce more economic value for me. But now it's just becoming not that hard to go create features. But features are not products or companies. And we've always had this distinction.

7:36

I don't think it's all one thing, but I do think Brave New World, and if you keep looking at it like the old world and it's got completely different laws of physics, you are definitely going to die. Yeah. Well, maybe let's talk about venture capital. There's a lot of cope going on now, too. So you got to be careful with that. Well, that's the thing. There are some things that really are features.

8:14

And before, it would take a long time to build a feature, so you might as well, it's comparative advantage. David Ricardo. I could weld my own steel. I could grow my own food, but I'm just going to not do that because I can do things that actually produce more economic value for me. But now it's just becoming not that hard to go create features, but features are not products or companies. And we've always had this distinction. There's feature, product, company, but it's a little bit confusing figuring out which one is which right now just because the ability to create a feature and create a product and even get all the data.

8:19

My favorite saying: the best companies have hostages, not customers. Even get some of the data out of the hostage company. Yeah. So it's a very, very confusing world in terms of figuring out which one is which, which is maybe a good segue to venture capital land. Yeah. How do you think, when you started this firm in 2009, big financial crisis, actually very, very big financial crisis, global financial crisis going on, the biggest. The world has changed a lot since then. How much of what's happening today kind of fits within the mental model of back then, and how much is kind of brave new world? Maybe riff on that a little bit. Yes, it's really different.

8:48

So our first fund was $300 million, and we raised it from all the traditional kinds of LPs, endowments, charitable foundations, et cetera, fund of funds. We just raised $15 billion for four of the seven funds, four of the seven funds. Not even the whole complex, and we raised it from very, very different kinds of investors. Basically none of our LP base was international when we started, and we're at like 35 percent international money, and it's from all kinds of places. Tech has gotten more important. I think that we have to think in terms of the world in a way that we just didn't before.

9:06

So, for example, why'd you raise so much money? Which, by the way, I'm kind of mad at myself because I don't even think I articulated internally well enough, because we could have raised even more money. Next time, don't worry. We had more money on the table. But the way I was thinking about it is, look, America's got to rebuild its entire infrastructure right now because we don't have enough rare earth minerals. We don't have enough electricity. We don't have enough manufacturing capacity. We have the wrong chips. They take way too much damn power. They were built for games. We don't have enough anything to be in this future world. And somebody's got to fund it.

9:36

Clearly that's going to take a lot of money. So all that is brand new. And I would say it's fairly overwhelming in a sense, but it is really, really important. We're pretty much out of electricity now in the United States. Not 12 months from now, right now. The China graph is like this, and the U.S. graph is like that. Yeah, and the demand for these tokens is straight vertical. But the ability to build that capacity is absolutely not vertical. So we need new everything. We invested in a transformer company, not like an AI transformer, like an actual power transformer company, because you need better, easier to manufacture, more efficient transformers.

10:41

And the transformer hasn't changed since we invented electricity. So these kinds of things. Well, I guess how, there's an old saying, the cure for high prices is high prices. Yeah. But the problem is there's a lot of latency involved. So right now there are computers that show up with no RAM. If you go buy a server from Dell, they're like, sorry, we don't have any RAM to sell you. Because all of that has been gobbled up. Yeah, they could build a new factory. Or you and I could decide to go build a DRAM factory. That would take us five years. Yep. So how do you, I mean, and we don't believe. Got to start now. Yeah, you got to start now.

11:39

But this is actually, if you remember, which you obviously do, 1999. It's like, well, we have to build more fiber, right? We have to build more capacity. But it's obviously very different because all the GPUs are hot. They're all lit right now. Yeah. And back then, most of the fiber was dark. Yes. But how do you get, like. Yeah, well, there were bottlenecks. When we were building fiber, the bottlenecks were kind of in different places. The servers weren't capable of putting bits out even fast enough to do video, right? The software was really, we didn't have load balancers. We didn't have application servers. We didn't have anything.

12:24

And so you had all this fiber and all this bandwidth, but you couldn't actually build the applications. And then most of the end users weren't. It's a network, too. So people weren't connected on the other end. So it just didn't work. And then we had the dot-com crash and all these things. So now we're in a little different place because almost everything is a bottleneck. I do think what's going to happen is we'll probably have enough chips long before we have enough electricity. So NVIDIA will make enough chips, but then we won't have enough memory and we won't have enough electricity. So we're in that kind of situation now.

12:44

So I think you really have to study where we are at each point in the supply chain and figure out how to alleviate those bottlenecks. And, by the way, God bless Elon, the TerraFab, that's the idea. He's going to just go deal with all the bottlenecks himself, which is how he does things, which is why we need him. Indeed. So I feel like you're an expert in three things: hip-hop, AI, and crypto. And I don't know anything about hip-hop, but I've heard a lot from you. But let's talk about the other two, in particular crypto and AI. So I actually just wrote about this. You remember the origins of crypto was hashcash. Yeah.

13:01

And the scariest thing right now, from my perspective, is that everybody with Claude or with ChatGPT can actually go super deep and personalize a phone call, an email. It seems like all communication is going to be completely unusable. Yep. I don't know if you agree with me. I 100 percent agree. Because it's like, normally, I can just delete, delete. I get some email yesterday, dear Alan at Index Ventures. It's like, well, I'm not Alan. I don't work at Index Ventures. Delete. And I'm very grateful that this person messed up my name because I could just delete that. Yeah.

13:31

Whereas if I get 1,000 emails, the best way of thinking about an email inbox is it's a to-do list that has write access for the public. Yeah. Right? It's like, anybody can get in. And now anybody can personalize. Same thing for phone calls. What do we do? And then it seems like there's a lot behind crypto. And that's why I mentioned Hashcash, because it was originally intended to stop spam. Yeah. So do you think there's overlap between AI and crypto? I know you do. So tell us about that. Yeah. So I do think it starts with the problems that AI causes.

14:19

And actually, one of the first things, I woke up in the middle of the night one day, and I was like, oh my God, somebody's going to go on a Zoom. It's going to be AI me, and they're going to tell my finance team to wire $500 million to Nigeria. And that's going to be a problem. So, we're like, okay, everything's hardware root of access. Don't believe anything from me unless it's got my cryptographic key on it, all that kind of thing. So I knew these problems were coming. They're coming so fast now. And that's why I mentioned Hashcash, because it was originally intended to stop spam. Yeah. So do you think there's overlap between AI and crypto? I know you do.

14:45

So tell us about that. Yeah. So I do think it starts with the problems that AI causes. And actually, one of the first things, I woke up in the middle of the night one day, and I was like, oh my God, somebody's going to go on a Zoom. It's going to be AI me, and they're going to tell my finance team to wire $500 million to Nigeria. And that's going to be a problem. So we're like, okay, everything's hardware root of access. Don't believe anything from me unless it's got my cryptographic key on it, all that kind of thing. So I knew these problems were coming. They're coming so fast now. So I think there's several categories of things.

15:12

First is just, are you a human or are you a bot? I think everybody is going to really, really want to know that, be it social media, a dating app, a Zoom call. Anything, you want to know, am I talking to an actual human? Okay, can I prove that I'm a human being? And then can I prove that I'm me? And then can I sign content? How do I know it's true? There needs to be a distinction between, I get so many AI videos sent to me from my family that they think are not AI videos. And they're like, did this really happen? And I'm like, no, you could actually ask Grok, and it's pretty good at that right now. But I think Grok's getting to the point where it can barely figure it out.

15:36

And I think at some point it won't be able to figure it out, or AI will not be able to tell what's AI. So the only way is you're going to have to have something, some cryptographically strong indication, a signed piece of content that says, okay, yeah, I made this. Or this is really a video of me, Marco Rubio, giving a speech. This isn't something that somebody faked. And then there needs to be a source of that truth. And who are you going to trust for the truth? Are you going to trust Google? Are you going to trust Meta? Are you going to trust the U.S. government? I think you want to trust the mathematical, game theoretic properties of the blockchain.

16:10

So I think that's going to be just a very, very important part of the infrastructure. And then you get into fraud. And what is, how do you know somebody's a citizen to get them money? Everybody's talking about, well, let's do UBI. Well, great. But when we did the stimulus program, we found out that the government is very bad at getting money to people. I don't know what it's like. Depending on the numbers you read, it's somewhere around $450 billion got stolen. So what you really need is everybody needs an address where you can send them money. And so I think that's a crypto problem. And then finally, how does an AI become an economic actor? How do I make money as an AI?

17:02

How does somebody send me money? Can I be a merchant, a credit card merchant, if I'm not a human? I don't think so. I think that's actually hard. And it's probably not the right infrastructure anyway. And so you need a bearer instrument on the Internet. You need Internet money for these AIs to be economic actors. And I think that's very likely to be crypto. So I think there are many opportunities in the crypto space that have been generated by AI. Yeah, because it feels like it's this old Yogi Berra saying: it's so crowded nobody goes here anymore. We're entering that era. Because number one is, are you a real person?

17:49

But the problem is that co-work is so good right now that, or Open Claw, I just say, you are a real person. You were a real person. But now your addresses are being used by a machine. Yeah. Right. So CAPTCHAs don't make any sense. CAPTCHAs are an anachronism. Like, what is a CAPTCHA? Right. So it feels like the solution lies in economics somehow and game theory. So. Yes. Yeah. And that too. Right. Right. Right. Like, are you going to just have to, well, maybe, I think Hashcash is a relevant idea again. Yeah. No, totally.

19:02

So maybe, why don't we talk about where you think venture capital is going? And I mention this because Mark got some crap for saying all the jobs will go away except for one job of venture capital, which was seen as a self-serving comment. But in his defense, I will say it's partially because it's a non-deterministic problem. Yeah. Right. It's like, all right, you're betting on an entrepreneur first and foremost. And you want to know that this entrepreneur, as I like to say, can materialize labor, capital, and customers. And you can't just run an algorithm on it. I mean, maybe you can, but there's just not a lot of data out there. It's very, very hard to do. Yeah.

19:36

So that's the logic by which, and also just personal relationships in general, will probably survive AI. Yeah. But if there's a venture capitalist, then that kind of assumes there's an entrepreneur job now. Yes. Yes. That is true. It takes two to tango. I mean, it's hard to be a venture capitalist without somebody in the transaction. Yeah, if you're very bad, you could just raise money and never allocate it, I guess. But I guess, what do you think the world of venture capital looks like today? We've obviously done a lot of things internally as a firm to try to embrace AI very, very fully.

19:48

But now, five years, ten years from now, just given what's potentially going to happen to white-collar work. Yeah. Yeah. It's really tricky because you go back to the last transition like this, which was the transition to the Industrial Revolution. So the venture capitalists of the railroads and the automobiles and so forth ended up becoming JPMorgan Chase, Goldman Sachs, et cetera. So they ended up becoming banks. And some of the reason for that was just how fast that materialized. So I think in the '30s, 20% of American workers worked for the auto industry, which is spectacular compared to what it is today.

20:15

And so things in the Industrial Revolution started out very much like we are today in venture capital, where there were 300 auto companies and so forth. And then it consolidated very hard into, in the U.S., the Big Three and so forth. And then the venture capitalists went upstream with the companies. I think that's one scenario where, okay, there's going to be a small number of very gigantic companies, and they're going to own everything and so forth. There's another future where it's like, okay, they got really big, and then we've finally hit the asymptote on this intelligence idea. They're as smart as they're going to be or whatever.

20:29

And we're either going to nationalize the big labs, and they're utilities, they're electricity plus-plus, like, F you if you're going to think you're going to collect all the money. And then everybody's just going to build on this utility set of things. And then that's a very different venture capital world. So I would say, as I'll quote Yogi Berra, the problem with predictions, they're very hard, especially about the future. And I think this future is particularly hard because it's so dynamic. And it's really hard. And then, how does the electricity shortage play into it?

20:43

Does it make the big companies all powerful because they suck up all the electricity and nobody else can get it and nobody else can get any GPUs? Or does that push all the computing out to the edge, and then the models just get really good and small, and everybody's like, well, I got enough in my phone, and what they're going to charge me for their mega GPU farm is just outrageous, and I'm just going to do that? So there are many ways it could go. And I don't know. I guess I don't know, but I could see venture capital being much bigger and much more exciting because everybody in the world is an entrepreneur.

20:56

Or I could see it being more like what happened in the Industrial Revolution, and new companies are just harder. Yeah. And I think this future is particularly hard because it's so dynamic. And it's really hard. And then, how does the electricity shortage play into it? Does it make the big companies all-powerful because they suck up all the electricity and nobody else can get it and nobody else can get any GPUs? Or does that push all the computing out to the edge, and then the models just get really good and small, and everybody's like, well, I got enough in my phone. And what they're going to charge me for their mega GPU farm is just outrageous.

21:36

And I'm just going to do that. So there are many ways it could go. And I don't know. I guess I don't know, but I could see venture capital being much bigger and much more exciting because everybody in the world is an entrepreneur. Or I could see it being more like what happened in the Industrial Revolution, and new companies are just harder. Yeah. Well, it's a good follow-up or a good parallel question, which is, how do we make this seem less scary? Because I don't know if you saw it. It's a lot of change. It is.

22:49

Well, but yes and no. I mean, 98% of Americans were farmers. Yeah. In 1789. Yeah. I'm pretty sure they're not farmers right now. I mean, you made this interesting point where if you go to a third- or fourth-world, if there is such a thing, country, everybody's an entrepreneur. Yeah. 100%. Like the guy, I sell bananas by buying them here and selling them there.

23:46

Everybody's an entrepreneur. There are no organized companies. Yeah. And the cool thing is that now 8 billion people that might have an idea in their head can get it out of their head. And maybe it's a bad idea. It probably is a bad idea. But there is no longer a gate for them. There's no capital gate. There's no idea. It's just boom. And it's not just for code. It's, I can write music. Yeah. Right? I can make a movie. This is super exciting. So that feels like a very, if you're trying to make this not look dystopian, I don't know if you saw Bernie Sanders interviewing Claude. Yeah. This is literally old man yells at Claude. Yeah. Metaphor, no metaphor. Right?

24:43

It's just like he's yelling at Claude. Yeah. And that's the dystopian view. And it's wrong. I feel very passionate that that's wrong. But we need a better narrative. So the history, I would say, if you look at a macro standpoint, the history of technology is things have always gotten better. Would you like to live in the world before electricity? Probably not. It doesn't sound that you can if you want. But nobody seems to opt into that. And I think we're very much in a period like that. But the transition is always scary because it's a different world. The jobs. Everybody was a farmer. Everybody was a farmer in 1750. I think it was 93 or 94% of America was farmers.

25:09

And then almost all those jobs are gone. Just like the jobs that we think are jobs, that they would have thought were ridiculous. Ridiculous. If you were a farmer, you would think what I do is the dumbest thing in the world, or a product marketing manager, or any of this stuff. It's like, that's not a job. You're not making any food. You're not building a house. How could that be a job? So I do think it's very hard to see to the other side of that. But I think it's very, very likely to be way, way, way better for everybody, just like electricity ended up being way better for everybody. And to me, the biggest, the most salient wrong idea was from John Maynard Keene.

25:36

So he wrote a paper that wasn't that famous, but the great economist of the Depression, where he said, look, things are going to be so abundant and everybody's needs are going to be met. Everybody's going to have a house or a shelter, and everybody's going to have enough food to eat. And then once you have your needs met, you're going to work way less, like 15 hours a week max, because your needs are met. But what he didn't realize was, well, we're not just going to need one car. We're going to need a car for every person.

25:53

We're going to need computers and television sets and this and that and the other and awesome vacations and food that takes a chef 10 hours to prep and all this kind of thing, which did not exist then. There was no foodies and tasting menus and all that bullshit that we have now. But that's all a need. That want goes to a need very fast. And humans are unbelievable in their ability to come up with new things that they need. And then you have to make those and so forth.

26:11

And I think it's going to be, look, I think in 15 years, the truth is everybody is going to, in America and probably around the world, live better than the very best life, from just luxury, access to information, et cetera, et cetera, than anybody did in 1980. So that's the world that we almost certainly are going to get to. So you shouldn't be so mad about it, but it is disconcerting. All right. Well, especially if you're trying to teach little kids, they're like, what should I do? I don't know. That's a hard one. Well, on that note, Horowitz at Andrews and Horowitz, thank you very much. All right. Thanks. All right. Thank you. Thanks.

27:14

But like what he didn't realize was, well, we're not just going to need one car. We're going to need a car for every person. We're going to need, you know, computers and television sets and this and that and the other and and awesome vacations and like food that takes, you know, like a chef 10 hours to prep and all this kind of thing, which did not exist then. Like there was no like whatever foodies and tasting menus and all that bullshit that we have now. But like that's all a need like that that want goes to a need very fast. And, you know, humans are kind of unbelievable in their ability to come up with new things that they need.

27:50

And, you know, and then you have to make those and so forth. And and I think it's going to be, you know, look, I think in 15 years, the truth is everybody is going to in America and probably around the world is going to live better than, you know, the very best life, you know, from just luxury access to information, et cetera, et cetera than anybody did in 1980. So like that's that's a that's the world that we're we almost certainly are going to get to. So you shouldn't be so mad about it, but it is disconcerting. All right. Well, especially if you're trying to teach little kids, you know, they're like, what should I do? I don't know. That's a hard one.

28:32

Well, on that note Horowitz at Andrews and Horowitz, thank you very much. All right. Thanks. All right. Thank you. Thanks.

Reading tools

Type to find a passage

Appearance
Ask this transcript

Notebook

Export Markdown

Add a note