The News
One of Silicon Valley’s top political fixers, Bradley Tusk, built a career helping startups like Uber and FanDuel brawl their way to legitimacy — in his words, “legalize disruptive sh*t.” His message to AI companies: You have no idea what’s coming.
“It is quite possible that this is the most permissive it’s ever gonna get for AI regulation,” he said on the latest episode of Compound Interest.
Tusk sees AI as unlike any industry where he’s fought and won. It’s too amorphous, too risky in its doomsday warnings, and too unpopular to run his tested playbook of mobilizing customers to pressure politicians. “I don’t think it’s inspiring people to say, ‘I must have access to Claude or ChatGPT or Grok, and I will fight to the death to keep it,’” he said.
He singled out Anthropic for smartly positioning itself as the “Luke Skywalker to OpenAI’s Darth Vader.” Anthropic has warned the public repeatedly about the risks of its AI advancements, while OpenAI has been more sanguine and has positioned itself as close to the Trump administration.
“I think OpenAI has very willingly embraced the Vader role,” he said. “I don’t quite understand why they’ve done that, but they have.”
And for all the focus on Washington’s relationship to AI, most of the fights coming for LLMs, data-center operators, application makers, power providers will be at the state and local levels, Tusk said. That means executive orders and White House positioning matters far less than the restrictions and limits on AI moving through state legislatures in Texas, Tennessee, Minnesota, and Indiana.
In this article:
Transcript
Bradley Tusk:
... but there are times where begging for forgiveness totally makes sense, and there are times where it doesn’t. And one of the biggest questions is what happens to you if you don’t listen? So with Uber, it’s like, “Okay, they’re going to impound the car. We can afford that.” But if it’s like a securities crime and you’re going to go to jail for 10 years, that’s not worth it, right?
Liz Hoffman:
Welcome back to Compound Interest from Semafor Business. I’m Liz Hoffman, Semafor’s business and finance editor, joined as always by my colleague, Rohan Goswami. Hey, Rohan.
Rohan Goswami:
Hi, Liz.
Liz Hoffman:
If I know one thing about you after working together for 18 months, it is that you are an Uber power user.
Rohan Goswami:
I would hope there is more than just that one thing, but yes. I do alternate, I use Uber and Lyft. I came to the office in one today.
Liz Hoffman:
Well, the reason you’re able to do that is because of our guest today. It’s Bradley Tusk, who’s a venture capitalist, but he came to that career through politics. He worked for Chuck Schumer, for Mike Bloomberg, for Rod Blagojevich as deputy governor of Illinois. And he got his start in venture by advising Uber in its early days on how to beat the taxi commission into submission.
Rohan Goswami:
Took a stake in them instead of getting paid, right?
Liz Hoffman:
That’s exactly right. Yeah. And he took equity for advising them on how to overcome local regulations here in New York, and ultimately, across the country. And then he did the same thing for other startups that were fighting entrenched interests, local opposition, Bird, the scooter company, FanDuel, an insurance startup called Lemonade. He tutored these companies in the art of regulatory hardball, and he won almost every fight he picked.
Rohan Goswami:
Yeah. And we are obviously not a politics show, we’re a business show, but his business proposition is basically VCs, and a lot of CEOs for a long time were mispricing political and regulatory risk because they just didn’t understand it. And so therefore, they couldn’t value it or value his services.
Liz Hoffman:
And he built this big career as Silicon Valley’s fixer. But he did it in a very Silicon Valley way, which was mostly to ask forgiveness, not permission. He’s sort of the city hall version of move fast and break things. And Silicon Valley has some big things it needs to fix right now. The two things I really want to talk to him about are AI and prediction markets, and what it actually takes to get disruptive technologies through the American political system right now.
Rohan Goswami:
We’ll get into all that with him. Let’s take a quick break, and when we come back, we’ll have Bradley.
Liz Hoffman:
Bradley, welcome to the show.
Bradley Tusk:
Hey, thanks for having me.
Liz Hoffman:
Let’s start here. How big a political problem do you think the AI companies have right now?
Bradley Tusk:
Significant, because fundamentally, they are incredibly disliked. It is across the country, it is across the political spectrum. You’ve seen legislation move and pass in red states, in blue states, on a range of issues, from frontier models, to consumer protection, to data centers. So it is quite possible that this is the most permissive it’s ever going to get for AI regulation, and for at least the foreseeable future, it’s going to get tougher and tougher.
Liz Hoffman:
So are you advising any of them?
Bradley Tusk:
I have a lot of companies in my portfolio that are AI companies, but I think usually people, when you ask that question, you mean the frontier models and the hyperscalers? No, none of them.
Rohan Goswami:
There’s a bunch of problems. Is there a clean single solution to any of this?
Bradley Tusk:
No, because the problem is that’s sort of like saying, is there a clean single solution to life? It’s too big. There’s no category that I’ve ever worked on from a political perspective that’s like this, because it encompasses every industry, every conceivable activity, everything else. And so at least when I think about how to deal with AI from a political regulatory perspective, I put it into four different buckets, which at least, for me, is a dichotomy that I think is useful. So the first would be consumer protection. So that’s anything that can touch a human being. So that could be things like data centers, and who bears the cost of electricity and water and everything else, to chatbots, healthcare, education, insurance, hiring, all of that. That’s one category. The second would be catastrophic risks. So that’s the frontier models, and Mythos is sort of the thing in the news the most these days.
And then the third category would be jobs, and that’s mainly around the harm caused by, potentially, AI job displacement. And look, I truly do believe that every major economic transformation in history, in 20 years, there will be all kinds of new industries and jobs that the three of us can’t conceive of today because of AI. But people during the steam engine, or the car, or whatever it is, had to live through those transformations, and that’s going to be the case here too. And then the fourth would be all the good things that AI could do. So that’s climate change, or drug formation, or education, or using AI to make government work a lot better. And so I think that you almost have to split it up among that to then try to analyze, “Okay, I’m an AI company in one of these four categories, how do I then go about my reputation, my regulatory status, everything else?”
Liz Hoffman:
Your playbook at Uber, which you’ve repeated a few times since then, I think was essentially to ask forgiveness, not permission to get startups to enlist their users to fight their political battles for them and just kind of brute force their way into ubiquity and legitimacy. Do you think that works with AI? And we can talk about prediction markets in a minute. These are things that actually people don’t naturally love, or it makes people queasy for some pretty fair reasons.
Bradley Tusk:
Yeah. Part of the reason that it worked for Uber is, think, now, Uber is such a sort of just normal fact of life, that you just get annoyed because it says pick up in four minutes and it takes nine, and you’re like, “What the fuck.” But back in the early days, if you guys think about the first couple times, you’re like, “This is amazing.” It was just magical. Because taxi, and I’m older than you guys, you have less probably recollection of that, was miserable. And so yes, we did beg for forgiveness, but we went in knowing that our customers were really, really going to like our product and want to fight for it. I think even now, if you were to ask people, “How do you feel about ChatGPT?” So forget about OpenAI as a company. I think they would say they use it, but it’s kind of like the phone company or something. You use it, but it doesn’t mean that you like it.
I don’t think it’s inspiring people to say, “I must have access to Claude, or ChatGPT, or Grok, and I will fight to the death to keep it.” And also, the second question around that is what’s the cost of begging for forgiveness? So even now, within the, let’s say, 100 companies I’ve invested in and worked on since Uber, there are times where begging for forgiveness totally makes sense, and there are times where it doesn’t. And one of the biggest questions is what happens to you if you don’t listen? So with Uber, it’s like, “Okay, they’re going to impound the car. We can afford that.” But if it’s like a securities crime and you’re going to go to jail for 10 years, that’s not worth it. And so with AI, because again, it covers all categories, an AI company begging for forgiveness in one space might say the risk is relatively minimal, it’s the right approach. And if you are one of the frontier models, my guess is the risk could be potentially extremely significant, and you wouldn’t take it.
Liz Hoffman:
The other thing that seems to be missing here is a natural villain. When you had Uber, you were fighting the Taxi and Limousine Commission. A lot of these other startups were fighting city hall basically, which is pretty easy for people to say, “I don’t like that.” It’s not obvious to me what the incumbents are here as you think about what you’re tilting against.
Bradley Tusk:
Yeah, or flipping around, I’m not sure there’s a hero. At least again, I think in this conversation, so far, what we’re really talking about are the hyperscalers. So we’re saying AI, but we mean Gemini, and Grok, and Anthropic, and OpenAI, and maybe one or two others. And in that case, you have something where people clearly feel... and their concerns very well might be wrong, overstated, whatever, but they feel like this thing’s going to take my job, it’s going to screw up my kids even worse than the internet already has, it’s going to unleash warfare and drones that the government can’t even control anymore. And so they’re inherently really anxious, and at least so far, the upside isn’t present enough. So in the equation, Liz, that you laid out, you need something good fighting against something evil. You have the thing that people already see as evil. Now, look, if there’s a moment where AI produces something that cures cancer, different story.
Rohan Goswami:
This raises an interesting point for me anyways, which is it feels like a San Francisco and a New York thing sometimes, this is a chattering class issue. Do you think that AI is actually a national issue and will be a key point and a point of contention in the election?
Bradley Tusk:
Yeah, I really do. And the way that I would answer that question is if you were just to go, ironically, one of the AI platforms and say, “Show me all of the legislation around AI that passed in states this year,” it’s really across the board. Yes, there’s obviously stuff on the coast, but there’s stuff in Texas, and there’s stuff in Tennessee, and there’s stuff in Minnesota, and there’s stuff in Indiana. And so when you cover all of the different categories, it really is all encompassing. And if you just take a half a step back, the reason why legislation would pass in any of those states is because politicians said, “My constituents don’t like this thing. And I need to show them that I’m fighting for their beliefs and their anxieties, so I’m going to move legislation.” I don’t think anything happens in politics if it is not reflective of an electoral outcome in some way.
And so that very, to me, clearly means that voters in... and I guess in this case, we can use voters as proxy for people, in all of these different parts of America clearly are anxious enough about AI that it is leading to specific legislative action.
Rohan Goswami:
As you know, there was a congressional primary this past week in New York, and I will begrudgingly can see that New York is not, unfortunately, the whole world, but one of the races was-
Bradley Tusk:
It is better.
Rohan Goswami:
It is better. Thank you. It is better. But one of the races was kind of a referendum on, I guess put it really simply, pro versus anti-AI and the sort of moratorium, let’s maybe put a kibosh on this thing, candidate didn’t win. And it got me thinking anyways, can you really be an anti-AI candidate today or is it just varying shades of gray?
Bradley Tusk:
No, I would actually flip the analysis completely, which is in the absence of Alex Bores getting attacked by OpenAI, he would’ve lost by a lot more. It propelled his campaign. It didn’t hurt it. So just again, now, this is probably for most of the people watching this, listening to this, going to feel really esoteric, but I’ve worked in New York politics for several decades now, and the reason I would argue that Bores, who was the East Side candidate, only lost by four points was because he was the anti-AI candidate. And so look, some disclosure, Lasher’s one of my closest friends and I supported him, but I know Alex actually helped him pass the RAISE Act in New York, and I think that he would’ve lost by a lot more.
Liz Hoffman:
What is your political advice to these big AI companies, and to politicians trying to figure out where the voters are?
Bradley Tusk:
If you are a startup in a highly regulated industry, you have to really understand the regulatory climate of your issue. And if you don’t do it well, your company could fail it, just like it could if you get any of those other things wrong. So the first thing is if you’re an AI company, regardless of whether you’re one of the hyperscalers or someone just in a particular application, again, if you’re doing automated workflows at business, you’re probably not going to have to worry too much about regulation. But if you are touching healthcare, or insurance, or education, or whatever it might be, you really have to understand both what’s happening right now and what could happen to you. And by the way, it’s not always all negative.
Liz, we talked about this a couple of weeks ago, if you are best in class... and I have portfolio companies like this, where if you’re best in class, sometimes what you want to do is lean into regulation and lean into the anxiety around AI and say, “Okay, let’s legislate the highest possible standards,” because if you know that you meet them and it’s really hard for others to do so, that actually builds you a regulatory moat. So that’s across the board. Now, if you are the hyperscaler specifically... Look, I think Anthropic is playing it really, really well. They kind of almost answered their version of-
Rohan Goswami:
Really?
Bradley Tusk:
Well, I think that if you think about Liz’s question about the heroes and the villains, I think they saw that. And what they’re at least trying to do is create a dynamic where they’re the good guys and OpenAI are the bad guys. So I actually think that they’re very much trying to position themselves as the Luke Skywalker, to OpenAI as Darth Vader. And I think for some reason, OpenAI has very willingly embraced the Vader role. I don’t quite understand why they’ve done that, but they have. And I think that of the two, Anthropic is both among the elites, and among... I don’t know how much the general public pays attention to Anthropic versus OpenAI as opposed to say Claude versus Chat, but I think that they have done that.
And I think that that’s smart, which is you’re building a tech that not only does government not really know how to regulate, you don’t seem to really exactly know what it is either. Because it just keeps evolving so fast that, given that situation, you can’t say, “Okay, here’s what’s going to happen over the next 36 months,” and step by step because it’s evolving too fast, and you’re learning things along the way too. So I think it’s more of almost a framework of we’re going to position ourselves as the responsible ones.
Liz Hoffman:
What are the lessons from the last set of battles? Because I would say, Uber was not the good guy, Lyft was kind of the good guy. And Uber won anyway. And I would say-
Bradley Tusk:
How’s Lyft doing now?
Liz Hoffman:
Correct. And isn’t the lesson from the fights that you were in that actually, it doesn’t really matter whether people like you?
Bradley Tusk:
No, because... No. Well, there’s a distinction there. When we say people, right now, we’re conflating customers and regulators. So regulators didn’t like Uber. Customers loved Uber, because if they didn’t, we wouldn’t exist today. By definition, they fought for us. And it’s not only just how we won the more... New York was the most famous fight, but that was the playbook everywhere. I didn’t work in other countries, but at least in the US, it was always the same thing, which was taxi was the sort of ultimate insiders. They were great about having lobbyists and political contributions, and we were the ultimate outsiders, at least in the beginning, because we were so tiny that we couldn’t play the normal game, and we were able to rally our customers, and ultimately, several million people across the US waiting with their elected officials and said, “I want Uber here. Don’t take it away from us.”
I do think that it matters, but ultimately, it is totally possible... Let’s say that you’re Sam Altman. You might be thinking, “You know what? All right, everyone hates me. Who cares? I’m about to have an IPO at a trillion dollars, and people use ChatGPT.” It’s like, I don’t know, 900 million people a week or something like that. “So I’m very happy to be the phone company. I’ll be one of the richest people in the world.” That may be true-
Liz Hoffman:
Do you think that Uber’s path would’ve been easier if Travis were more likable?
Bradley Tusk:
No, I think the thing that would have been easier would’ve been if we went public sooner. Ultimately, if you think about it, Travis’s departure from Uber was really because of Bill Gurley, and Bill Gurley wanted Uber. He and Travis obviously did not like each other at all, did not get along. And let me just be clear, I’m a Travis guy. I’m not objective in the slightest here. I think Travis is a genius, he is my friend, we work together on Atoms. I’m an... So I’m in no way objective here, but Bill and Travis hated each other, and Bill orchestrated a coup to get rid of Travis. And I think the only thing that might have changed that is if the company had gone public sooner. Throwing out a public company CEO is very different than throwing out a private company CEO.
Liz Hoffman:
Let’s talk about prediction markets. You worked with FanDuel during the fight to legalize online sports betting. And actually, tell us that story, because I’m not sure you’ve talked that publicly about it. How did you push that through?
Bradley Tusk:
So just for some context, my first exposure to tech was Uber. We’ve talked about that a lot. And what also made it slightly unique was Travis couldn’t afford my fee. I got paid partially in equity, and that obviously then became Uber. And so I was like, “Wow, this is pretty good.” So I did that again with Clear, and I took equity and helped me get to airports. And then I was launching a venture fund and I met someone at KKR who were the biggest investors in FanDuel. And they said, “Oh, you should just meet the FanDuel guys. They’re highly regulated.” So I met their general counsel, Christian Genetski, who’s now the president of FanDuel, in a nice meeting, but just kind of a meeting. And then the New York Times, a few months later, I see this alert on my phone, and they’re saying that FanDuel and DraftKings are guilty in their investigation of insider trading, which actually turned out to not be true, but didn’t matter, because once the Times said it, it was effectively true.
And all of a sudden, the walls caved in, and state attorney generals were sending cease and desist letters in droves, dozens of them. And I sent Christian a text and an email and I said, “Listen, I’m sure you’re already doing these things, but if it were me, I would do these five things.” And he wrote back and said, “Can you start doing them right now?”
Rohan Goswami:
What were the five things?
Bradley Tusk:
Mainly, it was customer mobilization, building the right immediate lobbying team, figuring out their media narrative. It was campaign 101, we’re not talking about anything particularly insightful or brilliant here. But that’s not what they did, that’s what I do. So started working on it. And at the same time, I had just raised my first venture fund, and so we both invested and started working on the underlying issues. And really, what it became was how do we just stay alive in every single state? Because effectively right now, they’re saying that we’re operating illegally. And it was very much the Uber playbook, which was people who did fantasy sports betting... So this was before the Supreme Court overturned PASPA, which was the federal prohibition on sports betting. So it was fantasy sports betting. It wasn’t nearly as voluminous as the number of people who use Uber, but they’re really passionate about it.
Because now, we’re talking about something that the people who do it... even Uber, which is a lot better than taxi, it’s still just getting you from point A to point B. For them, this was culturally who they were, that’s their thing. And so we went basically state by state, started rallying our customers, did what we did with Uber and a lot of other companies we’ve done, which is really have the big expensive public fight in New York, because you guys, meaning, the media, all see that. And then you think that’s just what it is. So then you reflect it, and that becomes sort of the perception nationally or globally. And so mega fight in New York, rallies, all this stuff, that created a perception of, “Shit, this is sort of the next Uber,” and it helped that I was the person doing each of them. And then we would go to any individual state and talk to a state rep, state senator, whoever it was, and say, “Listen, here in your district in Tennessee, between DraftKings and Uber, we have 4,142 customers.
Let’s be honest, they don’t know who you are. They probably don’t even know that the State Senate is even a thing, but they love fantasy sports betting. And if you take this away from them, we are going to make sure they know who you are, that you did it, that they’re registered, and we’re going to get them to the polls.” And in every single case, fundamentally, the politician in question just said, “You know what? I don’t care about fantasy sports over another. I’m not taking this thing on.” And we won pretty much everywhere. And then the company itself didn’t really explode in value until the Supreme Court’s decision.
Rohan Goswami:
Let’s take a quick break. We’ll be back with more from Bradley after this.
The View From
Rachel Oppenheim:
Welcome to What’s Working, the show within a show featuring Compound Interest Season partner, Amazon Business. I’m your host, Semafor CRO Rachel Oppenheim, joined as always by Amazon Business’s Doug Gray. Let me ask you Doug: as VP of Technology, how do you balance automation with human decision-making in procurement?
Doug Gray:
I think of automation as a helpful assistant, not a replacement for good judgment. It should take care of the repetitive stuff, reordering supplies, catching policy issues, routing approvals, so our procurement teams can focus on the things that require human thinking, supplier strategy, managing budgets, and thinking through risk.
The best systems keep humans in the loop for these high-stakes calls with clear guardrails and visibility — something we’ve built for an Amazon business. So automation speeds things up without reducing control.
That’s how I see AI working best. It keeps things running smoothly while the strategic thinking stays exactly where it belongs: with people.
Transcript
Rohan Goswami:
I want to ask you a slightly offbeat question. And Liz and I have this shtick that she’s the old head and I’m the young head. But it’s true, I am younger than Liz.
Bradley Tusk:
Liz, you look a lot younger. Just so we’re clear.
Rohan Goswami:
Liz does not age. This is true.
Liz Hoffman:
No, we have a shtick here where I’m a cranky millennial and he’s a sunny Gen Z. It’s a real click and clack vibe.
Rohan Goswami:
I have a famously sunny disposition. But the one thing that my proximity to young people, and I mean truly young people, people who can’t vote, my friend’s siblings, is they’re on Kalshi and Polymarket, they are using these apps. And we all know it to be true, their parents know it’s true. And I was telling Liz when we were talking about this interview, when we were in college, your bookies were your friends, and sometimes, your friends would get a really weird call-
Bradley Tusk:
That was your official motto.
Rohan Goswami:
Right. But your bookie is your friend. But the weird thing is you get a call sometimes from... not me personally, but if you were running a book, you’d get a call from someone’s parents, they’d say, “Look, I know my kids owes you thousand bucks. I’m going to cover them. Can you just not let them place any bets anymore? I don’t want them doing this.” And those were college-aged kids. I’ve yet to see that really mobilize with parents whose kids are spending, winning or losing, thousands of dollars. But it strikes me, we had the MAHA movement, are prediction markets and kids’ access to prediction markets going to be a midterm or an electoral issue?
Bradley Tusk:
It’s not, and I really wish it were. So I worked with Governor Hochul in New York earlier this year where she is going to make New York the first state to require biometric screening to log onto... Now, this would be things that the state regulates. So it’s sports betting apps, it’s not prediction markets. We can get into the regulatory distinctions, which might be worth talking about, but that’s federally regulated. She’s in control over that. And I did it because I have a 17-year-old son, and I was just watching him and his friends figure out ways effectively to get around the rules and-
Rohan Goswami:
Ripping on DraftKings. Yeah.
Bradley Tusk:
Right. So I was like, “Look...” his sister, my daughter, who turns 20, Sunday, came to me and said, Lyle, my son, asked her to set up a PrizePicks account and give her the login info. And Abby instead told me about it. And my thought was that even if she had done it and given him her login credentials, then he would still have to do, if it’s biometric screening, show his face. And that’s not her face, and therefore, he wouldn’t be able to log on. And so my thought was, well, if you require biometric screening, then you take away the different ability of people to use their parents’ account, or their sibling’s account, or things like that. And Governor Hochul agreed, and she now sit in her state of the state address, and it’s being worked on now. And I really hoped and thought that it would become something that we could then start driving nationally.
And it hasn’t really haven’t gotten any traction in other states. A lot of the advocates for the young men in crisis movement I thought would be really helpful here, and they have not been, quite frankly. Now, I will say, Congress is working generally on a regulatory bill for prediction markets. And my sense is, it’s funny, Kalshi and Poly have a similar dynamic to Anthropic and OpenAI, but I think that you’ll see different types of behavior, prohibited or regulated. For example, if you’re at the Department of Defense, you can’t bet on where the US is going to hit Iran and the drone strike, stuff like that. So I could see them taking up the issue of underage and that being part of it. So maybe that will happen. But even then, unless the Supreme Court, and we get into it if you want, sends regulation of prediction markets back to the state. Right now, you have a structure where the prediction markets are governed by the federal government, the sports betting apps are governed by states. And so what one does legislatively or regulatorily doesn’t apply to the other.
Liz Hoffman:
It sort of feels like this is more of an Uber situation where there is sort of an obvious villain, like big casino, state gaming commissions, that it is more of a traditional fight. I’m curious whether you think the playbook will work here. I guess the army of fans now is influencers, right?
Bradley Tusk:
Yeah, for sure. So the question is, will you need it? So let’s say your Polymarket, right now, the CFTC, they work well with you, you have your DCM, which is a license that you need. So it’s not necessary. Where that will come into play, if it does, is if the Supreme... So Supreme Court has not yet taken the case of who has jurisdiction, but just for context, for people listening and watching, there are court decisions that have said that states should have jurisdiction over prediction markets, or court decisions that said that the federal government should. The CFTC in various states are suing each other all over the place. So that all means the Supreme Court almost certainly will take this issue up, and then they will have to decide who has constitutionally the regulatory authority over production markets. If they side with the states, and it really feels kind of like a 50/50 question right now, then the Polymarkets of the world are going to have to go to the state gaming boards and try to get a gaming license.
I think that the gaming boards feel like, like you said, that they should be already regulating these companies. They should be capturing the tax revenue from these companies. And I think they’re going to be very, very slow to grant licenses to the Polymarkets of the world. And that’s when a campaign, where you turn on your customers, is going to become really important. Now, on the flip side, if the court rules with the CFTC, which is also very, very possible, then I think you could see a world where the sports betting companies just become prediction market companies and they leave state jurisdiction entirely, because right now, they’re paying higher taxes and dealing with more regulation. And then states are going to have this massive revenue gap, because the sports betting companies are some of the hot, biggest taxpayers in each state. And then they’re going to probably fill that by licensing new types of gaming, like casino gaming and things like that. And so the ripple effects on either side are going to be really, really significant here.
Rohan Goswami:
Let’s go with the scenario where it goes back to the states actually. Would their customers show up the way that Uber and Lyft’s customers did? I asked this, or even sports betting customers, but their customers don’t really trust them. They kind of know they’re getting screwed-
Bradley Tusk:
I think that they will, simply because it’s something they really enjoy doing. And if they’re no longer allowed to do it, they will be really upset about it. And keep in mind, these days, it doesn’t take... It’s great to have the physical rallies and the things that are great shots for TV, and YouTube, and TikTok, and everything else, but ultimately, like what we did with Uber, where if by pressing a button, you can let your relevant legislator know, the state senator or the governor, whoever it is, how you feel, it’s sort of low friction enough that... we’ll see, but my gut is that they will. If it comes to that, it might not.
Liz Hoffman:
Yeah. If you were running that today, what kind of influencer budget would Uber have?
Bradley Tusk:
A lot. It’s funny, the famous kind of Uber campaign was the one against de Blasio in New York, and we spent $5 million in a month. God bless Travis, I remember saying to him, “How much money I have?” And he’s like, “Whatever you need.” And I really spent what felt like an incredible amount of money in a short period of time. But this was 2015, at least 20X that now.
Liz Hoffman:
You once tortured Anthony Weiner out of running for mayor of New York when you were working for Bloomberg. I think you geo targeted his zip code with ads and some photographers out-
Bradley Tusk:
Yeah. Yeah, this is, keep in mind, 2009.
Liz Hoffman:
2009. Is there a version of that playbook that these companies should be running, or is that a little beyond the pale in today’s day and age?
Bradley Tusk:
Yeah. Well, let me give you a real world example right now. So we’ve been running a campaign for the solar and clean energy industry where they want to go after members of Congress who have been very anti-clean energy and taken away subsidies for it. And one of the campaigns we just ran was against a guy named Chip Roy, who is currently a congressman from Dallas, Ultra MAGA, Freedom Caucus, one of the people that led the fight against clean energy subsidies and tax credits, and he was running for Texas AG. And our goal was to take him out and make sure he lost to send a message that, “Look, if you mess with these guys, you’re going to pay a price for it.” And almost the entire campaign we ran was just ads on Truth Social, and Rumble, and a couple of other very, very right-wing platforms because we were opposing him in a Republican primary.
And our entire narrative was... Our first ad was literally called Not MAGA Enough. And the entire thing was that Chip Roy was insufficiently nice to Trump. And we didn’t bother with broadcast TV, we didn’t bother with direct mail, we didn’t bother even with broad-based digital advertising, we just honed in on the voters who would actually show up in a Texas Republican primary and played specifically to the thing that they care about, which is they want the most right-wing candidate possible, not unlike what we just saw in New York on the left the other night. So I do think that that’s where the confluence sits. The idea that someone who’s a looksmaxxer, you’re going to pay them a lot of money to say, “Here’s what the right policy should be on, prescription drug regulation-
Rohan Goswami:
Clavicular’s take, that’s how you say his name, on prescription drugs.
Bradley Tusk:
But there are definitely examples, because the Chip Roy race was May 26th or 27th, something like that. So that just happened. And we’ll have things... like right now, we have, to the geotargeted point, ads in airports like Reagan where we want DC types to see what we just did. So we have Don’t Mess With Solar billboards up in airports where we think that people who matter in those circles travel to reach them.
Rohan Goswami:
I always love flying into Reagan and seeing the hyper niche ads that are going after one senator’s staffers.
Bradley Tusk:
Yeah, this is exactly that.
Rohan Goswami:
I’ll keep an eye out for the solar ones next time I’m there. But you said something I want to go back to, which is in the Chip Roy race, you put ads on Truth Social and on Rumble. Our colleague, Max, had this great story the other day about the Daily Wire’s efforts to fundraise and build a business and get a nice valuation for that business. And part of their pitch is that blue chip advertisers, real companies, were finally okay advertising in conservative spaces again after sort of a backlash during COVID or what have you. Do you think that that’s a durable shift?
Bradley Tusk:
I do think so. But keep in mind, when the next January 6th happens, and for all we know, that’ll happen these midterms, then that’ll shift it again. So it might be true, but it’s definitely something where you’re constantly living right on the edge there. And you can also debate... I mean, it’s funny, I had this discussion yesterday with a friend who had a friend with a politics tech startup, and they’re like, “Oh, can you meet with them?” I’m like, “No.” I said, “I’ll talk if you want me to, but I’m not going to invest.” “Why not?” “Politics tech is not a real venture-backable category. I still believe that-
Rohan Goswami:
The left has their tech stack, the right has their tech stack, and there’s not really an expansion there.
Bradley Tusk:
Yeah. By the way, and this is for anyone listening or watching who is an entrepreneur, venture money is not always the right answer. Everyone likes to think that, “Oh, it’s super cool to be able to say that this fund or that fund invested in my company and I have this big valuation.” By the time all said and done, you may end up with less money in your pocket by the time there’s an exit than if you just build a really good business that there’s no equity that you’re giving away, or debt that you’re taking on and you’re just bootstrapping it. And ultimately, when P is greater than L, if P is significantly greater than L every year and you keep that money, that may make you a lot more than getting venture money.
Liz Hoffman:
You are winding down your venture funds and going back to where you all started giving advice for equity, not managing money. Talk about that a little bit. Is that just your personal circumstances or something else going on that you think is more broad?
Bradley Tusk:
Yeah, no, it’s a few things. Part of it is the somewhat, I’ll get into a set of unique opportunities that I think I might have compared to others, but some of it is also the state of venture capital broadly. So like I said, I did Uber, and I did Clear, I took equity, worked out great. I then raised and ran three funds, about $300 million AUM total. And it went really well. Companies like FanDuel, Lemonade, Bird, Circle, Ro, Coinbase, Latch, Alma. We did quite well. But as we were starting the process of raising fund four, I kind of took a half a step back and thought, “Okay, what do and don’t I like about this work?” And I like two things. I like legalizing disruptive shit, and I like betting on and working with founders who I think are really interesting. And what I didn’t like was raising money, dealing with LPs, sitting on boards, portfolio construction compliance, and I really didn’t like giving someone else 80% of my profits.
And that, combined with the fact that I felt like the venture game itself had changed where, look, if you can raise billions and billions and billions of dollars and the 2% management fee is so significant that none of the other economics really matter, sure, that makes sense. As long as you can get people to keep giving you money, keep doing it. But our third fund was 140 million. And in many ways, it was the worst of all worlds. So you get 2.8 million a year in management fees, which, by the time that you’re paying the team and everything means it’s pretty much all gone. So you’re not really making much money in management fees. But at the same time, you got to return 140 to the LPs before you even start to see a dollar in carry.
I kind of felt like we were never going to raise a $5 billion fund. That was just not something that I’m capable of doing, and my thesis is sort of too narrow for that. So I kind of created a different model instead, which is what I did with Uber and Clear where we take equity in return for our work. Again, still early stage, seed, Series A because I want to capture as much upside as possible.
Rohan Goswami:
What’s the illegal disruptive shit you’re waiting for someone to bring you? What’s the fight you’re looking to take on next that you want?
Bradley Tusk:
I think the one that I am really working on right now and excited about is around data centers. So sometimes, what I’ll work on is based on I’ll see a company and I’ll think, “Okay, this is really interesting. I like the upside potentially here, and we can win the regulatory fight, and therefore, we’ll take this thing on.” Sometimes, I start with a political thesis and then start looking into it. And if I believe, okay, the political thesis is pretty sound in my view, is there therefore an investible thesis that comes out of it? Data centers is one. So what was clear to me for a while now is the hyperscalers are not going to be able to just force regular people to pay for their energy costs at data centers. And we’re seeing that in places like Northern Virginia that have a lot of data centers where people’s electricity bills are going up 30, 40, 50% a month.
And the notion that everyone will be like, “Oh, it’s so important to me that Sam Altman become a trillionaire, that I’m just going to gladly pay an extra 30, $40 a month for my energy bill.” It’s crazy. And no politician’s going to be saying to those voters, “Oh no, no, you should have to bear the price of this thing. I don’t care if you want.
Rohan Goswami:
Some of them tried and they lost their races-
Bradley Tusk:
Right. Exactly. Specifically, in Georgia, the Public Service Commission. Yeah. So as a result, it’s unclear to me that data centers were going to be required to cover their own energy costs. And that could either come from one of two ways. They can do compute in a more energy efficient way. So that’s through alternate forms of compute, like reverse computing, or biological computing, or memory cylinders, or [inaudible 00:37:45] computing. Or it could be through behind-the-meter sources of power, like nuclear, or turbines, or things like that. And so I have set up a separate vertical within our venture business just focused on these types of companies, either energy or alternate compute, not to provide broad-based consumer electricity or things like that, but it’s specifically around meeting the needs of data centers, and then its needs are created because of politics. I have this theory that is grounded in nothing other than my head, but that you are either for or against nuclear depending on how old you were when Three Mile Island happened.
Rohan Goswami:
That is a good theory. I like that theory.
Bradley Tusk:
So if you were 10 years older or more, I think that you are terrified of nuclear. But if you’re not, and last I checked, it was like 82% of the US population was either not born yet or younger than 10. I think I was five when Three Mile Island happened.
Rohan Goswami:
I was negative five.
Bradley Tusk:
Yeah, right. Yeah, exactly. Right. Then you don’t care. So I think that means that if people could start to see, oh, here’s 20 different nuclear reactors that are powering individual data centers in these different parts of the country and everything’s going really well. Hopefully, that does lay the groundwork. Because it really is an incredibly clean, efficient source of power.
Liz Hoffman:
It’s funny, I’m on both sides of that theory you have, which is that I was not alive, but that I grew up about two miles from Three Mile Island. And so my parents have these stories of evacuating. But I remain perplexed that nuclear has the opposition that it does. It seems like a pretty magical technology.
Bradley Tusk:
And you’re like 0.00001% of the world-
Liz Hoffman:
Exactly. No, but it’s-
Rohan Goswami:
A proud daughter of Hershey, Pennsylvania.
Bradley Tusk:
That’s why Liz sometimes is really glowing.
Liz Hoffman:
Yes, exactly. Exactly. It’s that plutonium. No, but it’s an interesting investment thesis essentially, that nuclear is going to succeed because these guys do not want their heads to end up on pikes. That is a real political investment thesis.
Rohan Goswami:
I do see that. Yeah.
Liz Hoffman:
We’ve been talking about politics and regulation in the same breath, but one takeaway I have, particularly from the Trump era, is that they aren’t, that Biden and Democrats in general tend to be very process oriented and very regulation forward, and that Trump is all about political pressure and deal making. Does that change the advice you would give startups these days? Not that the politics are less salient and important to consider, but that-
Bradley Tusk:
No, because keep in mind, I think naturally, all of our heads, when we hear the word politics, we think of Washington. But when it comes to regulation, especially tech regulation, 80, 90% is not federal, it’s state and local. And so whatever Trump does or doesn’t do impacts a very, very small slice of the tech world. And so his behavior isn’t really that relevant for pretty much everything we work on.
Liz Hoffman:
Interesting. It’s such a counter narrative I think in the media how we think about stuff.
Bradley Tusk:
Yeah, I get it. And look, most reporters aren’t also trying to get permits for data centers or whatever it is.
Liz Hoffman:
Also true. Yeah, but you’re putting less stock in the EOs coming out of the White House and the Pentagon’s fighting all of that?
Bradley Tusk:
They’re almost all toothless, so they just don’t really matter all that much. The place where we’ve seen the least actual regulatory action on AI is Washington. They talk a lot about it. Look, when I worked for Chuck Schumer and I was his communications director, we had a press conference every single day. I don’t think after the two years that I was there, a single thing we had a press conference on ever turned into an actual piece of law. It seems to me to be exceptionally likely that a Democrat wins in 2028, and they’re going to actively reverse every single Trump policy, and every Trump favorite is going to become the enemy and vice versa. So I think the Anthropics play the long game.
Rohan Goswami:
But when you’re racing to build the most advanced thing, every little bit of leg up that you can get helps here. So yeah, point taken on 28, but would that be too late, do you think, the pace and rate of change that we’re moving in AI?
Bradley Tusk:
I haven’t seen any specific thing yet that Trump has done that benefits OpenAI at the expense of Anthropic. So I think that oftentimes, when we were doing Uber, Travis always was like, “Figure out a way to build a moat against Lyft.” And I couldn’t because the business models were so similar that any law I passed effectively was going to apply to both.
Liz Hoffman:
I want to ask you about one other project you have going, which is mobile voting. What’s going on there?
Bradley Tusk:
Everything I do operates on a pretty simple theory, which is every policy output is the result of a political input. Every politician makes every decision solely based on the next election and nothing else. And yes, there are exceptions, but by and large, that’s pretty much the way that they operate. And so the only way in my mind to align the public’s interest with the politician’s interests is to materially increase turnout, especially in primaries where 90 plus percent of elections are effectively determined. And you’re only going to do that by meeting the voters where they are and that’s on their phone. And so I started this thing called the mobile voting project. We tested the hypothesis by funding elections in seven different states. We either deployed military people with disabilities, were able to vote in real elections on their phones. The answer was yes, of course, when you put something on people’s phones to make it easier, people like it and they do it.
We then spent five years building mobile voting technology that is end-to-end verifiable, end-to-end encrypted, requires multifactor authentication, and biometric screening, and air gapping. It’s all open source code. It’s on GitHub, it is free, it’s in the public domain, I don’t own it, it is free for anyone to use. So why not take the money that I spent, the work that we’ve done, it’s already built, and use it to allow people to participate at much higher rates than a politician would make? So-
Liz Hoffman:
But there may be more people betting on an election than voting in it in a couple of years.
Bradley Tusk:
Maybe go, yeah. And that’s-
Liz Hoffman:
Some signals in that too.
Bradley Tusk:
[inaudible 00:43:49]. And so why not take the technology? Look, I’ve spent $22 million of my own money so far on this thing. Everyone in New York and the business community is freaking out about the election results on Tuesday night, but turnout was really low. And when only the people who are on the far left are the ones voting, of course, you’re going to lose. And either you got to mobilize your voters to the polls, which they’re not able to do, or you’ve got to make voting easier, and that’s mobile voting.
Liz Hoffman:
All right. Well, on that sunny note, it’s a good place to leave it.
Bradley Tusk:
Yeah. Sure.
Liz Hoffman:
Thank you so much for coming, Bradley.
Bradley Tusk:
No country in 20 years.
Liz Hoffman:
Yeah, I know.
Rohan Goswami:
This was really wonderful.
Bradley Tusk:
Yeah, thanks guys for having me. This was super fun. I appreciate it.
Liz Hoffman:
All right, Rohan, what’d you think?
Rohan Goswami:
Right off the bat, I was super surprised, at first, anyways, that he sort of not dismissed Trump, but was like, “Trump is not the main issue, the states are the issue.” And I think part of that is his seat. His clients are local businesses. They are regulated by states, the TLCs, or the gaming boards. But I also think there’s some truth to that, which is that we, in the media, and his comments were about New York City being all we care about were funny and fair, but we, in the media, really over-index on Trump, I think because he’s one guy. But AI is showing us, prediction markets are certainly showing us that the hot button issues that we care about as business reporters are really local issues. They’re really, really local neighborhood issues, whether that’s data centers in your backyard, or kids in your living room on prediction markets, when they shouldn’t be.
Liz Hoffman:
Yeah. And every cogent political thought I have, I’ve stolen this one from our colleague, Dave Weigel, but Dave was really early, he followed these campaigns around the country, these local county races, and he was the first to say to me and write about the way that these particularly data center backlash was bubbling up in totally politically unboxable ways. It is bipartisan, it’s sort of nonpartisan in some way. And one thing that I’ll be curious about is that it’s going to be an interesting political constituency for somebody, as these kids come out of college and are having trouble finding jobs, their parents who paid for this education, the local communities that just don’t like this stuff and don’t want it. And you started to see these companies wake up to this and start doing things in local communities. I just wonder if it’s too little, too late.
Rohan Goswami:
To the point that he made about Darth Vader and Luke Skywalker, you, and I don’t know if you stole this from Weigel, maybe you did, you made the point that OpenAI and Anthropic are Democrat, Republican coded, almost. It’s like choose your fighter. And each party is sort of aligned, and Dario’s a little bit more democratic, and Sam Altman is a little bit more Republican coded. And it’s just interesting, I take your point about nonpartisan, but these guys are also trying to make themselves partisan here.
Liz Hoffman:
I think the politics on this are going to be both at a very high level, the Pentagon is going to cut off your business at the knees and you got to go to city council meetings. And we’ve said a lot that I think the job of a CEO is just a lot harder than it used to be. And this is one of those situations where the swirling, competing political constituencies that you have to tend to who all want different things is just going to be incredibly problematic. And I don’t think that the playbook that Bradley spent a lot of time developing, but that really infused the culture of Silicon Valley in the 2000s and 2010s of move fast and break things. It turns out if you break these things, you cannot put them back together. And people are very wary of this technology. And so I think that’ll just be fascinating to watch.
Rohan Goswami:
The other problem is though, to your point about a town hall meeting, who goes? Is it Meta who’s showing up and trying to advocate for this? Is it the data center builders? Is it the power providers? There are so many different actual businesses caught up in here, any one of which can be the actual flashpoint that an organized, effective voice for the industry, the way that Brad was able to galvanize for Uber or did for FanDuel isn’t really possible because people are mad at so many different things.
Liz Hoffman:
I think the outrage doom loop around tech has just been sped up.
Rohan Goswami:
What do you mean by that?
Liz Hoffman:
His early clients were really popular. People really liked Uber. I don’t think he worked for Airbnb, but people really liked Airbnb, and they didn’t understand why local regulators were saying you can’t rent out your house. And then it took us 10 years before we realized that particularly a lot of the social media apps had done a lot of really negative things to all of our brains. This one is unpopular from the jump, people really do not like AI. Prediction markets obviously has a rabid fan base of people who use it, but it is not going to take 10 years for us to realize that this is really pernicious stuff. And so I think that it’s sort of happening in real time, and makes the job of a fixer like Bradley a lot more complicated.
Rohan Goswami:
I think because CEOs stupidly, and then smartly, and then stupidly again, have started blaming everything on AI, whether it’s layoffs or productivity gains. And so workers, whether they’re young or old, or somewhere in between, are freaking out. And I think it’s really hard for anyone to get their heads around what they feel other than... you put it really well, it was just queasy. Everyone’s just a little bit queasy about this and you can’t really fix that.
Liz Hoffman:
And without that fan base, that army of devoted users who will go to the mat, I do not know who those people are for this. Even the people closest to this technology are very wary of it because it is coming first for their jobs.
Rohan Goswami:
The other thing you said was interesting is that ChatGPT or Claude don’t have rabid fans. I would say, not yet. There is a whole generation of people who are in school right now who are using Claude, who are using ChatGPT natively, the way that you or I use the internet or use phones and we just don’t even think about it. And I think those are the people who are going to become the rabid fans, the ones who use Claude as their therapist or ChatGPT to plan their vacations for them and already can’t imagine existing life without them. I have friends who will literally say, “Let’s ask Chat,” to anything. The smallest thing, where should we eat? When should we go to the restaurant?
Liz Hoffman:
Yeah, but I regret to inform you that when I was your age, people said, “Well, just Google that.” This is a very rare technology where young people don’t like it either. And so I think that these companies are going to struggle to mobilize any kind of real enthusiasm among their users to call their local congressmen. And even power users of this stuff kind of wish they could unhook themselves, and it took us a decade to get to that place with social media. So I don’t know. There’ll always be a market for people to solve your political problems, but this is a doozy.
Rohan Goswami:
Thanks for listening to Compound Interest from Semafor Business. Our show this week is produced by the wonderful, Emily Gross and Manny Fidel, with limited thanks to Josh Billenson, who is currently tanning in Cannes.
Liz Hoffman:
With special thanks to Anna Bozzino, Katherine Bilgore, Claire Einstein, Rachel Oppenheim, Tori Core, Vilanna Wang, Garrett Wiley, Ben Smith, Stephanie Chang, and Daniel Haft. Our engineer is Bob Mallory, and our theme music is by Steve Bone.
Rohan Goswami:
If you like Compound Interest, please follow us wherever you get podcasts, and feel free to leave a review.
Liz Hoffman:
And if you want more Semafor Business, you can always sign up for our newsletters in your inbox @semafor.com.