Reid Hoffman: Why The AI Investment Will Pay Off

0:00Reid Hoffman is here to talk about the promise and the business of AI. We'll be back with the legendary investor and entrepreneur right after this.

0:38Welcome to Big Technology Podcast, a show for cool-headed, nuanced conversation of the tech world and beyond. LinkedIn founder, investor, and author Reed Hoffman is here to talk with us about the state of AI and his new book, Super Agency, What Could Possibly Go Right With Our AI Future. Reed, I've been looking forward to speaking with you for a long time and great to finally get a chance to do it. Welcome to the show. It's great to be here. Thanks for having me. definitely I have a copy of the book with me now this is the the galley the preview and I think the book that debuts will look a little different kind of cooler I know you're doing some custom covers we can talk about that and the book is all about artificial intelligence your your belief in the promise of AI I love the what could go right subtitle and and you know I think that a lot of Silicon Valley is asking what could go right and they have to find things that go right because Because if you think about the way that these companies have been invested in, it's wild.

1:36And maybe it's like the true Silicon Valley story where you're going to lose a lot of money and then hope that you'll make some money on the other side. But Silicon Valley has never done it like this. And, you know, there's no one that knows it better than you in terms of the way that these work, the investing and the entrepreneurship side. Just a few numbers. OpenAI raised the largest VC round in history last year,$6.6 billion. Anthropic raised$4 billion and is on track to raise another$2 billion. Elon Musk has raised billions in his own right. NVIDIA,$3 plus trillion market cap. Don't these investments have to pay off in a way that is going to be difficult for the math to work out?

2:18I actually think that those investments will actually, in fact, be pretty straightforward. It's not to say that there won't be, just like the kind of early internet and others, there won't be a bunch of crazy investments that, you know, will essentially not make money. But when you think about what AI is going to bring to the transformation of human life and human work, the investments will, I think, seem small in retrospect. And obviously, that seems crazy from a viewpoint when you're saying billions of dollars in venture capital. But actually, when you think about when every single thing that has a compute unit in it, and it's not just phones and computers, which will be huge, but speakers and lights and thermostats and cars and planes, trains, and automobiles, everything is going to get smarter.

3:18And as part of that kind of intelligence adds a huge amount of value, a huge amount of efficiency. And so the question really isn't if it will pay out majorly, which I think it will, but which investments and over what timeframe. Because those have some variable. But part of what I encourage people, part of the reason why I wrote Super Agency, is anyone who's not actually engaged in using AI today. And not just for, hey, I opened my refrigerator. I have a set of ingredients. What should I make for dinner? That's a fun use case. Easy. Or a sonnet for my friend's birthday party or whatever else.

4:00That's fun. but for serious things, for important things that range from like, oh, I've got this health concern I want to talk about personally to, hey, I'm trying to do this kind of work. And how does it help me with that work? And I think already today, if you play with GBD4, Copilot, Sonnet, other things, you will find that it can add real value today. Look, I agree with you in all of this. And I think the thing that you mentioned that's most important is the timeframe because most VC funds, tell me if I'm wrong, usually try to get a return in seven years. And if you think about open AI, right, let's just take that as an example.

4:41You were an early investor in open AI. It's been a while at this point already, but even the most recent rounds, 6.6 billion, you know, in, in seven years, you want to be able to return 10X on, that money, so$60 billion, and they're losing a lot of money. They lost$5 billion last year, according to reports. So how does the math work? So all venture, as venture rounds, work, you know, start with the company, like frequently in series A, series B, there's zero revenue, right? So they're always losing money. And it's kind of the curve of the expectation of revenue, you know, kind of compounding over the cost curve.

5:21And so that's just business as usual. Now, it's a higher scale with OpenAI with kind of having both a massive footprint already, a massive kind of leadership position, and kind of revenue that you're expecting to grow in the substantive frame. But this is actually one of the reasons why Silicon Valley tends to create so many of the technology companies that dominate the globe. Because if you take an overly kind of classic DCF analysis, business school prof analysis, et cetera, which much of the world does in investing and doesn't say, no, actually, in fact, it's a feature that you're investing massively on something that at the moment looks like today's iBanking analysis is not pretty good.

6:12but two, three, five, seven. And by the way, venture investments can go out to 10, you know, kind of years. And plus, if it's compounding and growing the right way, can be, you know, kind of just epic investments. And it isn't just AI things. I mean, this is part of like, you know, Greylock's investment in Airbnb, which I led the series A of, had that similar characteristics of just bleeding money and then later being the amazing business we see today. And that judgment call on a risk-discounted basis is part of what you're doing. And the investment that investors made last year is kind of more of a growth venture round relative to it doesn't necessarily require a 10x.

6:53I mean, we have trillion-dollar companies now. So trillion-dollar companies are definitely within feasibility. but it doesn't require a trillion dollar company to make real money because part of what happens 10x is when you expect zero as a possibility lower multiples are fine when you know that you already have a baseline massive value that you're building on top of but there's still questions about the roi here so basically if the roi doesn't come through i think you believe it will but if it doesn't that could be a real problem again it's thinking about the size don't you think that that has some blowback and it's interesting to me one more thing about that But a lot of the big traditional VCs did not participate in the most recent rounds.

7:33I don't know if Graylike did. I know Andreessen Horowitz, I think, sat out after evaluating. And then you saw Josh Kushner's fund really led the OpenAI round. But the traditional names in the Valley were not involved. Well, there's kind of different risk profiles and different games. For example, in Valley history, DST led around in Facebook when the traditional funds were all like, oh, we'll only do a$3 to$5 billion pre with lots of caveats and everything else. And DST came in and said, we'll do a$10 billion post and no caveats. And it was an epic investment for DST and Facebook, just literally blazing amounts of money and establishment of a position.

8:21And so sometimes when the traditional VCs go, hey, we're only early stage or we'll only do this, like new firms come in and show that through a smart bet can make a ton of money. And so that is also a historical Silicon Valley position. And I think that the questions around – you say, well, when can you invest billion dollars as a venture firm? And the answer is, look, you're trying to hit a return multiple, but it's much easier when you know, when you believe with some vigor and evidence that your chance of going below 1x is very, very small. And, you know, when you look at OpenAI's position, you go, well, actually, in fact, is my position guaranteed on the downside seems, you know, like a good bet.

9:12And then do I believe in compounding with a risk discount that I can make a multiple on the upside? that's when you know investors make money and part of venture is there's you know it's traditional and a lot of venture investments that you know firm x goes i'm gonna i'm gonna bet on company a when firms y and z go oh that's nutty and sometimes you know firm x is right and sometimes they're not i mean this is you know whether it's you know you know like for example speaking for graylock You know, when Greylock invested at Facebook at a$500 million pre-money valuation, literally all of the companion VC firms were, oh, that's crazy, this college website, Greylock's demonstrating that it's past time, et cetera, et cetera.

10:01That's how you make potential, and they're risky, of course, but that's how you make epic investments. So then how does this change the AI field? I mean, you're an early investor in open AI. the idea was for it to be open it's not open anymore i mean you don't see a lot of open source releases from them they don't publish as much as they intended to at the beginning it's kind of a perpetual joke from elon that they're closed ai and you know this technology that you invested in i think because you thought it's going to be a fundamental building block of the future for humanity now has such pressure on it to deliver that profit for the investors that put the money in.

10:39So how does that change the field? Well, so first, in terms of being open, there's lots of different parameters of open and open AI has been, you know, working to have like open, you know, kind of, you know, with obviously identity and registration for safety, but open AI, API access. It wasn't called open source AI. It was called open AI, right? Or open model AI wasn't called that either. And so the question is, is the parameters of open? And sure, people would say, well, I think what you should be open is all your meetings should be Zoom broadcast to the world. And that's what I think open should be.

11:14I mean, it's not an outsider's ability to claim open, it's insider's. And for kind of the entertainment, Elon was part of that decision as kind of co-CEO in terms of making that decision. And so that's how OpenAI has always operated. Now, I think that OpenAI continues to stay on its mission, which is how do we make AGI for the maximum benefit of humanity? And there's various things that it has done for, you know, kind of catalyzing safety practices, catalyzed, you know, kind of knowledge about what kinds of things would be the right kinds of design for bringing it into, you know, kind of everyday usage.

11:59And I think part Part of what OpenAI also led the field in was doing this thing that I call my book super agency, Iterative Deployment, which is make it available to, you know, hundreds of millions of people around the world so they can play with it, give feedback, you know, kind of say, hey, these are the things that really work. These things don't work as well. And be able to get familiarized and participate in the conversation of it evolving. And I think that is also part of the openness. And so I think that all fits within the OpenAI mission for what they're doing. It doesn't mean that there aren't footfalls.

12:33It doesn't mean there aren't corrections. You know, in early days, OpenAI was publishing a lot and then went well. From a safety perspective, we think we should slow down publishing some because part of being, you know, for the benefit of humanity is paying a lot of attention to what are called alignment issues within AI. And OpenAI has continued to be one of the major leaders in that. And so I think it's, you know, I'm still very, I think that OpenAI has stayed true to its mission and continues to do so with vigor. But the money does change things. And you sort of tend to see it as a consumer when you're in an app and you're like, oh, they're growth hacking me.

13:12And that's because they have to return some value to investors and they need you to get back on the app, either to see an ad or they just want to show Wall Street that their daily active or monthly active user count went up. it changes the product and so i'm curious what you think uh all this investment does to the product like do we get a different product uh because of this need to show returns well one of the things look there's a kind of a little bit of a trope within a bunch of media that money somehow is is anti the interest of individuals or anti the interest of society it's actually how we power everything in society and so you know having i'm not anti money I get it.

13:55I'm just, you know. I hear what you're saying. So money itself is not necessarily a problem. It's a question of how the money plays in. So, for example, money playing in as within the kind of functioning of society, products and services, you know, integration in industry, et cetera. Broadly, most of those things are mostly very positive. Now, of course, in a pure mission standpoint, you'd say, hey, we could make our service, you know, kind of free for everybody. And that would be a really great thing. an open AI is actually, you know, spending a bunch of money making it free or subsidized to get the widest possible exposure, engagement in the, and learning about these products and being able to use them.

14:36You know, a lot of open AI products are offered at below what their cost right now in order to do that. But of course, you know, they will, as an invested company, need to begin, you know, So kind of, you know, over time showing returns, it means that they need to also focus as intensely on their money-making capabilities as their, you know, engage the world, you know, kind of possibilities. But ultimately, I think, just like many commercial products, that's not antithetical to its mission because succeeding in raising capital, succeeding in building these products, and succeeding in offering them to the world is part of the, you know, pro -humanity mission.

15:19And, you know, we live in a capital world where these things are decided by capital. So I think that's a good thing. Now, might it say, hey, we're going to focus on some real money-making products for enterprise before we get to the mass market, you know, kind of global south, you know, things that are, you know, for people who are in less wealthy economies. You could have impacts like that. That's a natural part of how the capitalist system works. But I know that OpenAI's mission for humanity won't allow it to be strategic, but not distracted from its focus. So you think that we can still get AI technology that's going to benefit humanity as the North Star, even if there's an imperative to return to investors?

16:10100%. And I think that's already what we're seeing in motion right now. You know, as long as we're talking about OpenAI, I have one more question for you because you are on the Microsoft board and it's been pretty obvious to everybody watching from the sidelines that, you know, there was this really close knit partnership. Then some governance weirdness at OpenAI. Sam is fired. Sam comes back. And then next thing you know, Microsoft hires Mustafa Soliman, who you co-founded Inflection with. and it seems like there's some healthy competition, but competition where Microsoft is trying to start to build some of the same products that you might see within OpenAI and maybe insulate itself from some of the risks that it became painfully aware of during the incident with Altman.

16:58So what's the status of that relationship and can Microsoft and OpenAI still form a strong partnership and still have a tight bond given all that's transpired? and where things seem to be heading? Well, I think they're both partnering and competing. And actually I think that's healthy for both companies and healthy for the ecosystem. So they're partnering because there's a lot of ways in which OpenAI uses Azure infrastructure, which has been building amazing new capabilities through being on this journey with OpenAI. I think that OpenAI sells things to enterprises and consumers that are, you know, it's like, you know, you can pay OpenAI or you can pay Microsoft.

17:42And so they have a, there's competing in that. And I think that's healthy for the ecosystem and for also both companies. And so, but, you know, they also have ways that, you know, they, by partnering, they also potentially both, you know, both get a lot of benefits from that. And so I think it's a good ecosystem. I think the, I think when the history books are written, I think this will be one of the kind of epic partnerships that, you know, kind of plays out into the field. And, you know, there's now lots of people who, you know, it's a growing number of competitors. You know, so not just Google, not just Anthropic.

18:24You know, Amazon's beginning to do stuff directly itself as well. There's, you know, DeepSeek and others within China. And so the competitive sphere, you know, as anticipated is growing. Yeah. And you're setting me up perfectly for the next question, Reed. So thank you for that, which is you started off by saying the question really is time frame and who's going to win. Who do you think is going to win? Well, the good news is I think there's going to be multiple winners. And, you know, maybe that, you know, one company wins at one thing, one company wins at another. You know, the incumbents are obviously trying to bring AI into the areas that they have strong positions.

19:08So, you know, Microsoft's very focused on AI for the enterprise. Google's very focused on AI for search. And obviously, that doesn't mean that the only thing that any of these companies are doing, you know, part of what I think, you know, OpenAI is doing is saying, hey, we need to have a kind of a third strong position on with kind of an AGI for humanity mission. And so I think there are going to be multiple winners. And, you know, I and Greylock have continued to invest in a variety of startups because we think there's going to be a massive number of startup positions that will be really interesting as well.

19:44And so I think that there's going to be lots of winners. And so the kind of who's the winner is I think actually, in fact, it's like who's the winner of the Internet, right? Well, obviously, Google had a really strong position, as did Amazon, as did Facebook. but there were a bunch of others as well. And so I think that's similar. There's going to be many winners. But do you think there's room for all these foundational model companies to succeed and pay off? Like one of my perspectives is that like there's going to be some company that's just going to sort of collapse under the weight of what it needs to deliver, whether that's an anthropic, I don't know, or open AI.

20:23It just doesn't seem to me like they can all succeed, but maybe I'm wrong. What do you think? Well, I do think that there will be efforts at foundational models that won't work. And especially, I think, late entrants. Like late entrants have a harder, you know, catch up routine and a lot of kind of expense to go to it, especially given that the early folks like OpenAI have a very broad base, like to some degree, if you ask the average, you know, worker or person in the street, what is AI? They'll say, chat GPT, right? And so that gives OpenAI kind of a strong initial position. But I think that the later entrance to frontier models as startups may have particular challenges.

21:08Does Elon Musk fall into that category? I mean, he's a late entrant with XAI, but he's building a million GPU cluster out of Memphis. TBD. Now, it's a huge expense line prior to kind of a revenue position. but you know that as we talked about earlier can be a classic silicon valley position so i think it's a he's a he's a you know the the the grok and xai is a later entrant but you know obviously a very capable player okay um let's just talk quickly about deep seek since you brought it up uh it's one of those things that no one saw coming at least in the u.s and now if i talk about ai and don't mention deep seek i get a lot of comments uh it seems to indicate to me that china so for folks who are listening who don't know about it it is a large language model out of china according to its own benchmark so take this with you know for what it's worth it outperforms metas llama 3.1 and open ai's gpt 4.0 it seems like china's in the game do you agree uh absolutely And it was one of the conversations that I frequently have the last couple of years where they'd say, oh, you're just bringing up China competition as kind of the bugbear.

22:22We're like, no, no. Part of my job as an investor, as an entrepreneur, as a theorist is to predict where the puck is moving to. And the Chinese are strong and vigorous. There's lots of very capable, very hardworking entrepreneurial folks with technological depth. And so it doesn't surprise me that DeepSeek has entered the field with vigor. And I think that part of when you look at what is going to be playing out with AI, China and multiple companies within it, because I haven't evaluated yet, Minimax, which also just came out recently, but will have a number of very strong contenders for this in the world technology landscape.

23:12And so I think it's a, you know, the game is on. Our game is afoot if we want to be Sherlock Holmes. Every little thing that the U.S. and the West have done to hamper China's ability to play in cutting-edge technology doesn't seem like it's working. There's been restrictions to the silicon that they can use in China. And now DeepSeek has shown there in the game. Now, look, maybe it started with a llama model and they've just kind of Frankensteined it. But anyway, they've done it. It exists. Then you think about phones, right? There have been efforts to make sure that the processors within phones that are the most advanced don't get into the hands of Chinese phone manufacturers.

23:54I think the Huawei Mate 70 is a real advance and maybe on par with the iPhone in some ways, or maybe U.S. smartphones in some ways. Anyway, it's definitely cutting into the margins of Apple. Apple's struggling within China. U.S. handset makers are struggling there. And it's almost like a point of national pride there. I was there recently just for a very brief stop. But it seems to me like the government ban of iPhones within Chinese government offices is legit. And part of the way that they can do that is because they've gotten around these restrictions. So talk a little bit about why the West has failed on that front.

24:35I wouldn't necessarily say it's failed because most of the restrictions are kind of call it headwinds or kind of slowing, not stopping. Right. And so the question is, what degree has it actually successfully slowed? Now, part of the reason why I'm a supporter of the CHIPS Act and other kind of efforts is because I think part of the thing that we as the world, including America, have to kind of demand from China is kind of a level playing field. So if they say, hey, we can sell our stuff everywhere in the world, you know, and we can subsidize it in various ways that make it, you know, less, you know, kind of free market competitive.

25:16And we are restricted from how you can sell things here. That's a kind of a mercantilist policy that you have to essentially respond to. And I think that part of the CHIPS Act is to say, well, if you're going to be playing this merchantless policy, we're also going to, you know, put some brakes on, you know, kind of your ability to develop the kind of the large size, very effective clusters that we think the next generation of AL models are going to be very dependent upon. because it doesn't surprise me. I mean, GBD-4 was built on A100s. It doesn't surprise me that with the current kind of A100s and other things, you could build another GBD-4 model.

25:59So that's not as surprising. And we have H100s going to Blackwells. And the question is, what is the speed at which we're going to be getting into that? And I think that's part of the game that's still afoot. And so I don't think it's a failure. And I do think that it's the right thing for the U.S. to be doing, given that China is generally taking a kind of a mercantilist competitive position. And I think that's something that the U.S. and other, you know, kind of countries, Europe and everything else, should be opposed to. You mentioned the Blackwell, which is NVIDIA's cutting-edge chips. all the headlines have been showing that they've been having some trouble getting them to production or getting the yields they want and there have been some delays.

26:50I feel like, Reid, you'd probably have a pretty good window into what's going on there. Maybe not like what's happening in the halls of NVIDIA, although you might, but certainly what type of delays we're seeing given that you work with Microsoft and you invest in startups. Is this a real concern? Is this going to be a bigger story over time if they can't get their yields up and produce the Blackwell chip the way that everybody is anticipating? Well, it's definitely if they can't get their yields up and produce it, that will be a concerning thing for NVIDIA and will slow down some of AI's development because the faster chips in a dense environment is actually, in fact, instrumental and helpful.

27:31And is that a legit concern? It slows it down. It doesn't stop it. But on the other hand, and look, I think this is, again, one of the things where the actual timing matters. It's actually not atypical for either NVIDIA or chip manufacturers to sometimes be, you know, to have six to 12-month error bars in the stuff they're doing. And even though, you know, public markets tend to be the, oh my God, this quarter, and it's like, well, the real thing is what does the next two to three years look like? And so I, you know, maybe this is my longer-term venture perspective. I tend to go, well, a quarter here, a quarter there, it doesn't really matter in the three to 10 year kind of time horizon.

28:13And so maybe I'm a little bit too casual on this, but time will tell. Yeah. You can be chill given your position, but public market investors will freak out. But I guess we'll cross that bridge when we come to it. Okay. I want to speak more about your book, which I think is sort of like in some ways a manifesto to the world talking about the positive uses of AI and why that book needed to be written. So we know that there's some issues with trust in AI and maybe that's the reason. I'm going to ask you a bunch of questions about it when we come back right after this. Did you know your credit card points and miles can lose value to inflation?

28:52Credit card companies often reduce the redemption value of your points and miles. Now, imagine a credit card with rewards that can grow in value. With the Gemini credit card, you can earn Bitcoin or one of over 50 other cryptos instantly with no annual fee. Every swipe at the store or gas pump earns you instant rewards deposited straight to your account. Plus, sign up now for a$200 Bitcoin bonus to kickstart your rewards. Visit Gemini.com slash card today. Check out the link in the description for more information on rates. Again, if you're looking to invest in Bitcoin but don't know where to start, the Gemini credit card makes it easy.