NBA赌博丑闻、亿万富豪税、Tesla的未来、Amazon机器人、AWS宕机、危险的AI偏见

What’s the story with the California wealth tax? Can somebody explain this to me?

Okay. The SEIU, the Service Employees International Union, filed a ballot initiative, which means a direct-to-voter vote, to amend the California Constitution and introduce a one-time billionaire wealth tax. Anyone who has assets over $1 billion, net of their debt, has to pay a one-time tax of 5% of their net worth, including their private stock and their real estate.

You said 5%.

5% of their net worth—not of their income, but of their net worth. It’s a one-time payment to the state of California, and there’s an allocation for how that money will be spent. But it’s a one-time billionaire tax.

Now, it is very likely that this sort of amendment to the California Constitution is not constitutional and cannot actually be made enforceable, even if voters approve it, at both the federal and state levels, based on the concept of uniformity. You have to tax everyone equally, except in the case of an excise tax, such as income or a transaction. You’re allowed to tax disproportionately based on the size of the income or the size of the transaction.

But if you’re going to tax property or an asset, you have to tax everyone uniformly. So it is likely not going to go into effect if it passes. However, it is very likely that the SEIU is simply using this as a baiting mechanism to get people to stand up and denounce it. Then they’ll be in a position to attack those people and destroy them, using this effectively as political fodder for the next election cycle. That seems like the true motivation right now.

Let me go on the record: I think this law is great.

He’s getting the virtue-signaling points.

I would just like to say: may I be the first to pay 5%? I’ll be in the front of the line. Let me know when to show up; I’ll bring my check.

Who do I sign the check to?

Should I bring cash, Gavin? Should I just bring it to you? Which one of your mansions should I bring the cash to?

This is strategically why, Chamath, I’m glad I got out of California right before I was about to billionize. That was a smart move on my part.

Free, what are the odds that this goes into effect? Can you just handicap this?

Yeah. We don’t know who’s going to come out against it, but there’s an effort to try to get top Democratic officials in the state of California to say, “This is silly. If you do this, people will leave the state,” yada yada. So that’s a quiet effort underway.

But I don’t know why the citizens of California—the majority of citizens of California—would not vote for this. Why? Who wouldn’t want to tax billionaires 5%? Come on.

Well, the way it’s written, it says, “Hey, guys, we’re $30 billion in the hole, and there are 200 Californians who control $2 trillion. We’re just going to ask them to pay a one-time fee of 5%.”

And I don’t see how anybody would say that doesn’t sound unreasonable at the ballot box, right?

And then the people who step up against it and point out, “Hey, in France, when they did this, they lost about 40% of their revenue because all the wealth left the country”—the reality is that this sets it up to go through the legislature. If it goes through the will of the people and gets overturned, as you say, Friedberg, then if you’re legislatively smart, you’ll push it through the state Senate.

Oh, but don’t you remember? It will not get vetoed, because then it’s like, “Hey, listen, it’s clear that the people want this.”

I mean, they’re already trying to extend Prop 55, which is the progressive tax for people making over $1 million. They’re going to get that passed. That’s going to be this incremental tax on income. But the one-time wealth—

I think the million-dollar thing is harder to target because there are too many people that it touches. A million dollars today, in 2025—not to be glib—is just not what it used to be. But $1 billion does cut off most people except for a couple hundred.

That is true. And I think, for example, it’s very reasonable to then charge a 10% excise tax on selling appreciated stock, right?

Right.

Why not? There are all kinds of ways that you can get billions and billions and billions of dollars. So I don’t know. I think that this is more of a trial balloon—

—to say, “Can we draw a clear line between 200 Californians and the rest of California?”

Yeah. And to the extent that that bright line becomes visible and acceptable, people are going to go ham. They’re going to try to get as much as they can.

The reality is, as we all know, Larry Ellison left the state. Many of the founders and CEOs who have built large technology companies in California—and Elon left the state—will eventually, at some point, break and say, “Okay, I’m moving my company out of state, and I’m leaving the state. I’m bringing the employees with me, and I’m bringing all of the economic value of this business with me.”

People will never learn that lesson because it’s so much easier to sit in front of a voter and say, “Hey, should we tax these 200 people to give you better benefits?” Ninety-seven percent of people will say, “Absolutely.” Very few people will sit and think about the consequences of what’s going to end up happening.

Ninety-nine-point-nine percent will say absolutely. Nobody tells them in that ballot initiative that we have a $300 billion budget, of which two-thirds may be just wasted.

One of the motivations for this bill—and this is why it’s being proposed by the SEIU—is that there are these massively ballooning pension benefits and pretty significant increases to the pension programs for both private and public pension funds in California. That has actually become a very visible liability for the state and for some of these private pension programs.

They’re trying to fill the pension hole, which we’ve talked about in the past. But there is a multitrillion-dollar, unaccounted-for pension liability in this country that’s going to have to come from somewhere. You’re either going to have to print the money because the federal government is going to step in and fill the hole in all these pension obligations, or they’re going to have these massively progressive tax programs to try to fill the hole.

And if and when they do, as we all know, there will be an economic cycle that will be pretty nasty, in which all the value will leave that jurisdiction and move elsewhere. But let’s see.

It’s like the Democrats are doing everything they can to get me to leave the state. I don’t want to. I really am resisting. They’ve raised my income tax to—what is it, like 13.3%?

13.3%, yeah.

And I know it’s going to 16%. They’ve been boiling the frog, and I still haven’t jumped out of the pot. But for me, I think the wealth tax—I’m going to have to jump out of the pot with this.

The crazy thing with this—I read it because I was like, “Oh my God, what’s going on?”—is that any Roth IRA over $10 million counts. Normally, in these wealth calculations, you keep your deferred retirement accounts off the table. They’re typically not included.

So folks—and I’m not going to say who they are—we all know who have tremendously appreciated Roth IRAs—

It’s pretty public who you’re talking about, but sure.

Those are included. And then the other thing is that if you actually did any tax structuring, the really valuable tax structuring is where you set up these trusts in Wyoming and North Dakota and do these interparty loans where you can lever up 10 or 20 times. You can transfer billions and billions and billions of dollars out of state, but then you have these obligations. Those are negated and don’t count, so all that tax structuring goes out the window.

You can get into a very difficult situation here where they’re like, “Hey, you owe us $500 million, $1 billion, $2 billion,” and the only way to pay it is to have an IOU to the state of California, which is crazy.

It’s crazy. There aren’t a lot of ways out of this if it stands.

No one gives a shit about either of you. No one cares about the two of you needing to pay more.

This is why I’m in support. Again, I’m just saying it for the record: I support this.

I had a few more thoughts about this thing that I want to unpack. Number one is, like you guys said, a wealth tax has been tried in many places at many times. It always backfires because whatever the tax benefit is that you get for the state, it’s greatly outweighed by the economic depression that you get from wealthy people—the job creators—and companies leaving.

As soon as you cross that line from no wealth tax to any wealth tax, enough people of wealth can see the tea leaves. They can see the writing on the wall that they have to leave. And that’s why I think that even if they say this is a one-time thing, we all know that it won’t be one time. If they get away with it, it’ll become a regular thing.

If it’s to plug a deficit, they’re going to run deficits every year.

Exactly. And you’re right. This isn’t even to plug an emergency situation or an unfunded liability, some one-time thing. This is just regular operating revenue. So they will have no incentive to fix their mismanagement of the state and their deficits and all that kind of stuff.

By the way, if they get away with this, it’s not just going to be billionaires. Eventually, of course, the line will get pushed down.

Billionaires always like to say that the income tax in the U.S. was originally a 1% income tax, and it was just a one-time thing for wealthy people. Then it became a smaller thing for lower-income people, and eventually, as we all know, every person has to pay a tax, every property has a tax, and so on. This is the problem with government. There are all these other states, by the way, that are finding clever ways. I think in Montana now there's a differential property-tax scheme where, if it's your second or third home and you don't live there, you pay a lot more.

Here's what I wonder about: What are guys like Jeffrey Katzenberg or even Ari Emanuel thinking about right now? They're the higher-ups in the Democratic Party, behind the machine—sort of the oligopoly that runs the machine. I remember that when Karen Bass was running against Rick Caruso for mayor of L.A., it was very publicly reported that Katzenberg was behind Karen Bass, and there was sort of an imbroglio between Caruso and Katzenberg. Katzenberg helped make sure that Karen Bass was well-funded enough to win.

The result, ironically, was that Pacific Palisades burned down, and I think Katzenberg's house might have been part of that. In any event, I think there are these guys who are very, very wealthy who think they can control the machine well enough that they're still in control of this thing, right? In other words, that the tiger won't eat them. The tiger is socialism.

Yes.

And that's exactly right.

You know, they think they've got the tiger under control enough that it won't eat them. But I don't think they do. Maybe they don't. Maybe this is the tiger breaking loose.

Yep. And I think, Friedberg, you pointed this out: There was an attempt in the legislature last year to pass a wealth tax, and it was quietly killed behind the scenes. I actually think that Gavin Newsom might have something to do with that because he has presidential ambitions. He can't let the state go full socialist. But you just kind of wonder: If these guys lose control of the strings they have to control the beast of socialism, does the whole thing just spin out of control?

That's New York. We're seeing it everywhere. Seattle.

I was about to bring up France. Just to let people know about the France situation, back in 2011 or 2012, they did get rid of Gérard Depardieu, which was kind of a win, but Bernard—

Arnault.

What's Bernard Arnault—is that his name? From LVMH?

Bernard Arnault said he was going to go to Belgium and said it was a clerical error, and he unwound it. But that was a clear signal. The richest man in France—well, he's—

Go to New York.

He's basically the entrepreneur who put LVMH together. I mean, it's their biggest company. It's the one that does all the luxury goods, all the craft goods that they're so famous for.

I mean, yes, him threatening to leave France is—accidentally filing paperwork. Oops. What an accident.

Here's your look at New York City under Mamdani, who we're in touch with. He may come on the program. New York State tax is 10.9%, the city tax is 3.876%, and the 2% Mamdani tax puts you at 16.8%, or 17%, for living in New York. If you were making $10 million a year, is it worth $1.7 million? You could get a plane, you could live in Florida, and you could come to New York 150 days a year. There are really 5 good months in New York: the fall, the spring, and that's about it. You go see the tree at Christmas, but it's cold.

Well, that's not realistic for most people. Especially if you have kids and you care about them, you'd like them to be rooted somewhere. You're not going to schlep them around every month to arbitrage taxes.

Yeah. Well, I do think they're going to test that at 17%. That's non-de minimis. Okay, we've got a lot of docket to get through here.

I'm so glad Chamath supports the billionaires tax. That's great. We'll get that in the headlines right after all. This is the free-rider problem that we have: No one's going to want to stand up against it, and the thing will just kind of—

Pass.

By the way, if you're the billionaire CEO of a public company in California, you have everything to lose by standing up and opposing it. Your employees will run, your shareholders will attack you, and you'll look awful in PR. So everyone's going to sit quietly and start looking at houses on Zillow in Austin or Miami and be like, “Where should we move to next year, honey?” You know, that's the conversation that's going on.

Didn't you say it was retroactive? What's—

It's retroactive to 2026. So if it passes—

You have 3 months. 2 months. Yeah.

But again, I don't think it passes muster under a constitutional read.

Remember what they said when they did the transfer tax, where San Francisco took 6% of my home?

Yep.

And then in L.A. they just took 5% of my house down there—the supposed mansion tax.

But those were excise taxes. If you go back to the case history in the U.S. Supreme Court on this stuff, anytime there's a transaction and you take a tax on a transaction, they call that an excise tax, which is constitutionally part of the bill.

There's a part of the bill that they could cleverly use, which is called the ODA, which is effectively this IOU mechanism. They could essentially say that when these assets transact, you owe us 5% on an excise basis. And, by the way, there's an attestation that you have to file—you have to file a legal document. This was quite well written in there, which said, “You must attest that you have less than $1 billion.”

Okay, now what? Then I have to attest that it's more—

How do you even mark your whole portfolio to market if you have a lot of private assets?

They do not allow discounts. They do not allow liquidity discounts. It says if you are a reasonable buyer and a reasonable seller, you have to transact this at market price.

For example, imagine you owned a sports franchise, and if you sell a minority share, you're typically selling it at a discount off the table. If Forbes says it's worth $10 billion and you own 10%, that's $1 billion for the purposes of this calculation.

I'm going to pay $50 million to keep it, even if you paid $50 million to buy it.

Even if Friedberg is right that there's a good chance that it'll be found unconstitutional, how many years in the courts is that going to take? And who's going to stick around waiting for that? In fact, the rational thing to do is pull up stakes before January 1 and leave right now.

That's right. That's going to happen in New York. I think they're going to have an exodus, just like New Jersey and Connecticut did. That actually rocked the tax base in those 2 geographies.

All right, listen. Big breaking news this morning: Huge scandal in the NBA. The FBI just arrested 30 people in a sports-betting and gambling probe. This hardly seems real. Chauncey Billups, who is the current Blazers coach and was just inducted into the Hall of Fame, got pinched for a poker game he was allegedly running with the mafia that was rigged 17 different ways to Sunday. Terry Rozier allegedly is a point guard for the Miami Heat.

Why are you saying “allegedly” all the time?

Everybody's suing these days, so everything is allegedly true. I'm alleging he's a point guard. I've seen him play. He's not a very good point guard. There are a lot of turnovers, if I'm being honest. You know what I know is alleged? That you're the world's greatest moderator.

That's allegedly true. It's allegedly true because it's not true.

You're allegedly a billionaire. Nobody can confirm it.

Normally, he uses the word “allegedly” when it's a story that's about Hunter Biden or doing something improper.

Yes, he allegedly smoked crack and shot a 9mm in the air.

It's usually a story about Democratic wrongdoing, and he's trying to discredit it.

All right, here we go. Anyway, keep going. Terry Rozier, who's allegedly a point guard, told his friends, “This is crazy. In the over-unders, bet the under on me in rebounds because I'm going to take myself out of the game with an injury”—allegedly. His friends allegedly made $200,000 off this. Okay, just allegedly for the whole goddamn thing.

This is across 11 states and a bunch of crime families. Allegedly, there's something called the mob. I don't think that really exists anymore. I think that's an urban legend. These are 2 separate threads, but they were announced on the same day. They both involve NBA players, but apparently this is 2 different cases. So, Chamath, what do you allegedly think of this?

I think it's crazy. I think you're seeing a lot of these trends converge all at the same time. You have the emergence of all of these prediction markets, and you have a lot of data science and AI being used that shows there's a lot of odd behavior.

It really was the squares versus the sharps. If you had the inside edge, you were just printing money. Now that all of that is becoming more transparent, there's a lot less margin.

Then what happens is you have these laws passed in the 11th hour. There was an important gambling law inserted into the One Big Beautiful Bill that has implications for all of this. And now you're seeing the feds.

The crazy thing to me is a press conference where Kash Patel is talking about this. That's serious business when the FBI director is front and center talking about all this.

So, I don't really know what it means, to be honest. I was shocked at the scale of it, and I was shocked that it's on the radar of the feds. I thought this was pretty typical ticky-tacky stuff, but clearly there's something bigger. I don't know exactly what that bigger thing is, but something is happening where all these markets are smashing together. There's just a big cleanup effort going on. So, I don't know. I really don't know.

Friedberg, I guess there are 2 different ways to go about this. You have fantasy sports becoming legal. Everybody around these players—just in that 1 case—where are these people? Are they too dumb to understand that their $10 million contract to play in the NBA every year, or $20 million contract, is more important than your friends betting the under or over? And how dumb are they, I mean, to not know that the people running a sportsbook would look for weird action? Why is 1 player getting $200,000 on their over-under for rebounds, and the other players are getting $20,000? What are your thoughts here? Also, take on the poker 1.

I think gambling generally, as we call it, should be decriminalized, and I don't like this state-by-state setup with gambling. I think we should have a federal regulatory body to oversee and monitor it. The problem is, you have state gambling commissions, and we have a state-by-state patchwork of regulatory authority that makes it very hard to standardize, track, and also provide guidance and feedback. I would much rather see this all get handled at the federal level and better organized.

To Chamath's point, this is not going away. People love to bet on stuff. They love to gamble. This is part of sports. This is part of the culture. They're not going to just turn it off.

Then Polymarket raised, whatever it was, $1 billion or $2 billion at $9 billion.

Then the next weekend they announced sports betting.

And now they're raising money 30 days later—

Allegedly.

Allegedly, at $12 billion to $15 billion. I mean, anywhere.

It's unbelievable. And you can see, by the way, the way that DraftKings and FanDuel stock have reacted to this. Those companies are toast.

Toast.

That's right. This is really interesting. The Polymarket model is the best model because it creates a market, and as information flows in, that market will dynamically adjust and everyone will get a more fair price.

Did you see the regression that they did on the Polymarket trades and how well they're in the money? Hey, Nick, can you find that? Basically, what it showed is that the front money is the sharps and the back money is the squares, but you have to fade the trade in the 1st week. So, there's a very scientific method where, if you want to make money on Polymarket, it became pretty clear.

There are 2 things that are very interesting about it. Number 1 is how they've simplified things to a way people can understand. It's not like you have to understand, you know, it's 120, or this is the point spread. It's just, what's the chance that this thing happens? 80%, 20%? People can just place their money on it.

And then there's this ability to reconcile it at any time. I didn't realize how engaging that is. I was watching the Oscars and I was watching boxing, and I bet the underdog in this Netflix boxing thing that happened because I just thought, "This guy looks pretty pissed off." I thought that was a good enough way to go with the underdog. Then you watch it round after round, and you see the odds changing in real time. Anytime you can just cover the bet, take your winnings, and take out the risk, it's really interesting and fun for people. It's so simple.

Then I did it on the Oscars or the Emmys, and I was like, "Yeah, I'm going to fade—no offense, Ben Stiller, but I'm going to fade Severance." I went with the one about the emergency rooms and Andor, and I won again. So, it's a lot of fun to do it.

Jason, look at this. I sent Nick the tweet, but this is incredibly systematic. This is over many, many, many markets. Basically, it's 89% accurate 1 week out, but in the final 4 hours, it jumps to 95%, which means that if you follow the sharps along this pattern, you're going to make money.

6% in a week.

Yeah. Right.

Polymarket actually has the news before the news does. This is one of the most powerful outputs of Polymarket: It's actually getting a read on what's going on in the world before the media recognizes it, before the public recognizes it. Because when you put—

Yeah. When you put money up, it actually turns out that when people have incentives, that market will find the truth.

Somebody needs to build the app that makes all of these things fungible. By "all," what I mean are cryptocurrencies, betting markets, equities, and options markets, all turning into—

Yeah. And the reason is there's just no reason to go to 9 different sites and have 9 different accounts. The most important thing is to do KYC and AML across 9 sites to get access to liquidity, credit, and margin. You'll want to do it once. Then you'll want to have a large pool of capital that you can trade across anything. So, I can go long Nvidia, but I can also go short the Knicks, and then I can own some Bitcoin, all in the same trade.

Totally.

That's where it's going.

Totally. Totally.

Now, to the earlier question, JCal, I think if we end up there, where Polymarket does become the truly liquid market across all of these predictions and all of these assets, then a lot of what we are seeing with respect to insider trading and insider information becomes much more apparent.

The problem with sports betting is that there's a 1-sided bet. The casino sets the odds—or whomever is setting the odds—and then you're either taking 1 side or the other. If you have the insider information, you're taking the side that creates an arbitrage opportunity for you.

But if you were to do that in a liquid market where there's someone taking the other side in a dynamic way, then the market very quickly moves because of the inside knowledge you have. That inside knowledge is now reflected in the underlying asset price, in the underlying odds that you get for that bet.

Polymarket actually brings truth and transparency to what is currently an insider arbitrage opportunity, and it may actually solve some of these fundamental problems in gambling.

I think let's just wrap with a little bit on poker and knowing if you're in a rigged game or not. Living in LA, I got invited to a lot of poker games when I was playing low stakes, playing at Hollywood Park—$500 buy-in, $1,000 buy-in. But as these things went up, you started to get access, and I started to get invited to Molly's Game, the very infamous game.

She would text me, she would call me: "Oh, we're playing over here. Leo, this person wants to see you, that person wants to see you." I was like, "They want to see me lose $50,000. There's no way I'm playing in that high-stakes game, and I'm not going to that game."

The 1 or 2 times I did go to games that had a rake, I was just like, "This game is fixed." I don't know how. I think it's just collusion. I think there are 3 players all playing from the same chip stack. In which case, you could be dealt aces 5 times in a row. If you're up against 3 players, what are your odds against 6 other cards? It's going to be pretty bad for you.

You think Molly's Game was fixed?

I don't know if hers was. I wouldn't be surprised if it was. I wouldn't be, because once the mob gets involved—which is what happened at the tail end of hers—then all kinds of possibilities happen. Once it gets to extremely high stakes, and you've got guys chasing it, coming back night after night trying to catch up for what they lost last week, it's pretty dark.

There is absolutely no reason why anybody should play in a game where you're playing with people you don't know. If you need it that badly, then you probably have a problem. But there is no limit at which you couldn't find a game with some combination of your friends and respectable, reputable businessmen who have more to lose than you do. If you can't find that game, you should not be playing in any game.

Yeah. Any home game with a rake should be absolutely suspect. Period. Stop. Super sketchy.

Isn't that game?

Yeah. Well, yeah. We don't want to bring up angle shooting, but he's a straight player. He would be so tilted if he heard you say that.

Oh my God. He's so about the ethics. He wants, you know, no flies.

In fairness, that game where you can go off for a small house is also the game where he would then collect $10 from each of us to pay for the fruit plate and the pizza.

He would order Domino's. He wouldn't even buy us pizza.

Yeah, the chef. But I'm like, I don't know if the chef really does cost $6,000 for 2 hours, bro.

It's Wagyu. But I think it's a Wagyu burger. The funniest ever was, he's in a hand and the Domino's pizza comes, and he's like, "Everybody have a green chip." We were playing with $5 chips. He's trying to get $125.

The guy comes, and I just go, "It's on a card." The guy's like, "You got to sign, right? It's got the tip on it." I said, "It's like $150 a piece." I said, "What's the biggest tip you ever got?" He said, "Yeah, somebody on New Year's gave me, like, $200." I just wrote $500 on a $150 thing. I signed it. I gave it to him, and then I was in a hand with him and said, "Here's the receipt."

So, what you're saying is, when it's on somebody else's credit card, you're willing to tip incredibly generously.

God, you're a really great guy.

You should speak when Phil Hellmuth and I bought dinner for everybody at Cipriani that time. Chamath grabs the check and goes, “I'll put the tip in for you guys.”

Well, why is that, Jason? Is that because of a 100% tip on an $8,000 check?

Isn't that because I pay for everything all the time?

That's true. You are very generous.

One time, I asked you guys in 15 years to pay one time. It's so sad. I was like, “Oh, God, I guess we're going to public school.”

You guys are so ungenerous.

I know. I give huge tips.

Yeah, I think he's average.

Okay, let's go to the next topic. [laughter] Allegedly, the world's greatest moderator.

Let's talk about this Amazon outage. Tough week for Amazon. They had this huge outage at the beginning of the week, and then they had a bunch of leaked documents about their plans for jobs. On Monday, there was a massive AWS outage: 2,000 companies and 4 million users were unable to function on the internet for half a day—15 to 20 hours.

Then, on Tuesday, internal documents viewed by The New York Times showed Amazon plans not to hire 600,000 workers by 2033 because of robots. So this isn't them planning on laying off 600,000 workers; rather, they're pulling back their hiring plans and ramping up their robotics plans, which you would expect. Their goal, according to these leaked internal documents, is to automate 75% of warehouse operations.

We talked about this the last couple of weeks. Friedberg, your thoughts on either of these 2 stories?

I think the AWS story is interesting in terms of its implications for the cloud market. There are effectively 3 major cloud vendors that compete with one another: AWS, Microsoft, and GCP, or Google Cloud. I'll just give you these numbers.

Also, by the way, Oracle is coming on strong.

Yeah, that's right. But let's exclude number 4 for now: Oracle. AWS is at a $124 billion revenue run rate. Microsoft is at $120 billion, and Google Cloud is at $54 billion. But AWS, which is slightly larger than Microsoft, is only growing 17% year over year. Microsoft is growing 26% year over year, and Google Cloud is accelerating at 32% year over year. Some say it's getting closer to a 40% growth rate.

The big thing I hear from partners and enterprise customers of these cloud services is that many of them, if not all of them, as they scale up, move to a multicloud model. None of them want to be dependent on a single cloud. Many folks started on AWS because AWS was the OG.

Back in the day, when I was running Climate Corporation, I was the largest EC2 user on AWS for about a year and a half. EC2 was their Elastic Compute Cloud service. We were running all these models back then, so I knew that service very early on. It was very unique and very powerful, and a lot of companies that are old school established themselves on AWS very early on.

But the outage that happened this week starts to highlight for folks that they can't and shouldn't have a dependency on a single cloud service provider. It will only accelerate the diversification of companies into the other clouds. I do think this is actually a very beneficial situation for Microsoft and GCP and, to your point, Jason, perhaps even Oracle.

It gives those sales teams, which are very aggressive, a hard story to sell: “Guys, you don't want to just sit on AWS in case this happens again. We've got better infrastructure. We're more reliable than these other guys.” That might be a naive, simplistic, reductive way to think about what happened this week, but we are seeing the smaller competitors accelerate. I think this might be another moment of acceleration for those folks.

And multicloud has been around for a while. When you're doing stuff with AI, are the big companies already doing that? Or do they assume, “Hey, there's going to be some downtime. It's okay to take that risk”? Are they really thinking multicloud, neocloud? Let's have some smart, intelligent routing and redundancy here.

I think there are 2 markets: the AI market and the non-AI market. In the non-AI market, everybody has everything. It all looks effectively the same. There are certain products and services that are unique to Azure versus GCP versus AWS, but by and large, the market is big enough and important enough that you'd have to be pretty insane to take a single-vendor approach.

What typically happens in these markets is that they start off really small, with 1 player having all the share. Then, as the market becomes very valuable and very big, everybody diversifies because it's a risk-management thing. These things flow into the disclosures you have to make as a public company. If you didn't have that diversification and something bad happened that impacted your business, you could get sued.

There are all these reasons why eventually all 3 of these big companies will converge at roughly a third, a third, a third. We're going to debate the path to get there, but that's where they'll end up.

You know, there's this principle called the rule of 3, where they say all markets eventually mature to a 60/30/10 split. You end up having your market leader at 60% market share. Second place is usually half the size, at 30%, and then you always have some balance in the market where there's some competitor that settles at about 10%.

It's really interesting. If you guys were to place a bet, who would you think is the 60/30/10?

I don't think that applies to cloud.

You think they're going to be a third, a third, a third?

I think it's all some idiot making something up.

But what do you think happens in cloud? Do you think that these all converge to equal market share?

In non-AI, it's a third, a third, a third. It will take circuitous paths, but that's where we'll end up.

By the way, a good point to make is that the revenue numbers I highlighted for Google Cloud, Microsoft, and Amazon actually include their applications. As you know, Microsoft and GCP have pretty sizable enterprise application stacks built into those numbers, which gives them the ability to drive cloud usage because they've got demand and sales relationships into those enterprises.

I think the way it works in AI is that, initially, right now, we're in this early phase where there are 2 paths. Path 1 is that you need a specific model, and it's relatively well integrated using a specific subsidized form of hardware on one of the hyperscalers. Eventually, more of that will get abstracted away as it's pushed into the infrastructure, so that you have less dependence on 1 model.

There's a lot of work that has to get done and a lot of in-memory infrastructure that isn't yet built but has to exist. Once that exists, it'll be easier for all of us at the application level to view these models a little more fungibly. At the bleeding edge, you'll have the folks that basically give you some form of a hypervisor, virtual machine, or bare metal, and that's where the neoclouds are doing really well.

But I think my point is that in any important market—in compute, in technology—where there really isn't much differentiation, you'll end up with these hyperscalers at a third, a third, a third. Now, if 1 model is way, way better and it's only on 1 of the clouds because Google writes a big check or Amazon writes a big check, I could see that swaying the AI share. But in the absence of that, I think cheaper, faster, better is sort of the end destination for everybody.

What an extraordinary outcome for Amazon. AWS is like 15% of their revenue right now, but it's 60% of their profits today. That was just a side hustle, a little project they took out of nowhere, and it's having the same impact on Google and other places. So, side bets.

And side quests are just—look at the Waymo side quest for Google, or even a lot of Google's other bets, like Larry's flying cars, Loon, low-Earth satellites, Google Fiber—all those X projects had so much potential in them.

TPU, DeepMind, TensorFlow, GFS.

Robotics, Boston Dynamics. They bought all those robotics companies, man. It's like somebody got to them and said, “Yeah, you know, you're 7 or 8 years into this. It didn't happen.”

The problem that Google has, unfortunately, is that they have so much stuff, and it's not really valued. They're going to go through the same problem that everybody else who's a conglomerate has, which is this decision.

Now, Buffett, when he got to that decision, said, “I don't care. This is my life's work, and so I'm just going to keep everything aggregated.” But now you're going to get to this thing where the intrinsic value of everything they have will far exceed the actual value at which it trades. There will always be these fissures of pressure. If 1 of these things requires a lot of money, there'll be pressure to segregate these things so that I can own 1 versus the other.

That's always what happens in public markets: you swing back and forth. I suspect that this is going to happen at Google.

This was what they set up to do with Alphabet: to be the holding company. Then, to your point, they made that evolution, particularly in a company like Waymo, where they said, “We can't be the sole funder.” They brought in Silver Lake. They brought in all these other investors. They did this, actually, with Verily.

They did this with a bunch of these things they call “other bets.” They made the conscious decision because, Chamath, on the flip side, by bringing in outside capital and having an independent board for these subsidiaries, they were actually able to drive better outcomes. Now there was governance and aligned interests that could then take management and say, “Guys, if you can deliver these results—”

You had this kind of external pressure as opposed to the softness.

It’s that, but it’s something else. There’s no way somebody as smart as Silver Lake comes in if they think there’s not a path to liquidity. The other thing they have to promise is, “Listen, we will take this company public, and in return, you will help us build a better company than we could build ourselves.”

It seems that Silver Lake has done its part of the bargain. Now it’s up to Google to live up to its part of the bargain, because if it doesn’t get liquid, it sets a very bad precedent for everybody that committed capital into that company.

Yeah, of course. Waymo going public would be unbelievable next year, man. If they did that, what would that look like in the public markets? $250 billion.

Take it easy. Stop. Don’t do that.

You don’t think so? I think it’d be huge.

Jason, we all objected to talking yet again about AI-driven job loss. Yet you insisted on putting this AI robot story from Amazon in. I think you have something to say.

Thanks. Let me take you through a presentation. Well done.

You have slides?

No, I’ve just been working on a presentation based on a lot of stuff we’ve been talking about here. I threaded it together. We were just talking about Google and the size of the company. Right now, in 2025, they are at 187,000 people. They were at 190,000 people in 2022, and their revenue has gone from $283 billion to $350 billion in basically 3 years.

When you look at this Amazon stuff that came out, I just wanted to point out a couple of things. It’s not just that they’re not hiring these 600,000 jobs; it’s that they’re in full-blown crisis preparation for this. They have crisis teams writing up how to handle this and be a good corporate citizen. They’re talking about having parades and paying for Toys for Tots, and they’re even trying to get the executives to say things like “cobots” as opposed to “robots.” Let’s not call them that. Let’s call them co-workers and cobots.

When you look at this, just to open up the aperture, right now Walmart and Amazon are the number 1 and number 2 employers in the US. 2.1 million people work at Walmart, over 1 million work at Amazon, and 3 million people, as we know, work in taxis, for Uber, and as DoorDash drivers. All those jobs are at risk.

We talked about this back in June, when Andy Jassy telegraphed all this in a blog post where he said, “Over the next few years, we expect that this will reduce our total corporate workforce as we get efficiency gains from using AI extensively across the company.” They believe that they’re going to have significant job displacement. Let’s just use the more neutral term here, as opposed to “job loss” or “not hiring.”

I don’t know if you saw today, but there were a bunch of MAGA people saying, “These interlopers in the MAGA movement are not taking into account the bottom half of the MAGA movement—the workers, people who don’t own equities.” When we look at electricity spiking—you were on that story last week, Chamath, or maybe it was even 2 weeks ago now—the Energy Department just said electricity costs for residential customers are going to go up 4.8% this winter. This is going to start this anti-AI boom backlash.

I tweeted about this, and I thought I would maybe end here with Elon, who replied to my tweet and said, “AI and robotics replace all jobs. Working will be optional, like growing your own vegetables instead of buying them from the store.”

Senator Bernie Sanders came out and said, “I don’t often agree with Elon Musk, but I fear that he may be right when he says AI and robotics will replace all jobs. So what happens to workers when they have no jobs or income? AI and robotics must benefit all humanity and not just billionaires.”

I’ll stop there, because this, I think, feeds into your story for the last 2 years on this podcast, Friedberg, which is the rise of socialism. These things are starting to come together. They’re starting to come together in people’s minds, whether it’s the original MAGA guy saying, “What’s going to happen for American workers?” We know that the Trump 2.0 agenda is doing great—AI buildout, crypto, all this great stuff, trade—but the bottom half that you keep talking about, Friedberg, is starting to connect on this issue.

I think that you are characterizing AI automation and technological progress as the core driver of the socialist influence. What I would argue is that the actual core driver of the socialist influence is the fact that we put a lot of people into government and passed a lot of laws that caused an increase in spending, because we promised people that the government would do more for them over the last 40 years.

That is not possible in a true market-based system.

I agree with that.

By telling everyone, “Hey, we’re going to make sure you get better jobs. We’re going to make sure you all get housing. We’re going to make sure you get education,” you cannot actually get a government to effectively do that. What ends up happening is that the government inflates the cost of those things, and the market doesn’t actually work.

So the truth is, this is now, like all other things, a scapegoat for the true cause of the socialist movement, which is that government has become too big and too unwieldy. Its natural inefficiency has distorted markets to the point that there is maybe no point of return anymore.

People will not see that. They do not see it, and they’re going to look for reasons and scapegoats. They’re going to say, “Oh my God, look over there. There’s a robot. That’s the reason I’m losing my job. Oh my God, look over there. There’s a rich person that works at a pharmaceutical company. That’s the reason I can’t get healthcare.”

Or an immigrant took my job, right? That’s the one from the last 20 years.

Fundamentally, I think that people aren’t willing to—and they’re not going to—see the true cause, because there’s no one who runs to go work as a politician who is going to raise their hand and say, “Government is the problem.” No one says, “I need to reduce government. Elect me.” No one has ever gotten elected in a democracy doing that.

The natural course of things over 250 years is that people raise their hand and say, “I’m going to give you more, and I’m going to use the government to do it.” Then they go into the government and make the government bigger. As a result of making the government bigger, the government spends more, the dollar goes down, the performance of the services goes down, and fundamentally we end up in a socialist spiral.

I think it’s confirmation bias for you to see that story as confirming a point of view. I mean, it confirms what I predicted last year: that Amazon would be cutting all these jobs for robots. That’s all. It’s not confirmation bias; it’s confirming.

They haven’t cut a single job.

Actually, they have fewer employees now than they did 3 years ago.

No, that’s not true.

Yep.

It’s actually not true. The New York Times story doesn’t even say that. You’ve got these hobby horses where you keep coming back to the job-loss narrative, the copyright narrative, and then there’s 1 story in The New York Times, which was a leaked internal document from the automation department. That doesn’t even mean it’s going to happen.

This is like their sales pitch. That barber is trying to sell you a haircut, and you read that and you’re like, “Oh, it confirms everything I’ve been saying.” What the article actually says is that they’ve tripled their number of employees since 2018, and they’re not planning on cutting jobs. If the program pans out, then the rate of hiring will simply be slower.

I didn’t pick that to cherry-pick.

Okay, which is fine. I’m quoting the New York Times article, which is the source for this.

Yeah.

Amazon’s US workforce has more than tripled since 2018 to almost 1.2 million. You have to read these New York Times stories carefully, because they want to make the headline as salacious as possible. Then the echo chamber wants to make it even more salacious, and they make it a story about job loss when it really is a story about operating leverage in their business, which is a slightly more nuanced take.

Yeah. No, there’s definitely nuance here. I would believe Andy Jassy when he says, “We’re going to be reducing jobs,” and when this chart shows that they’re flat to down over the last 5 years. The same trend is just happening at Google, like I just showed. There’s a static or slightly declining team size occurring at all these companies, and it is notable.

On top of this, which has occurred in the rearview mirror for the past 5 years because of COVID, return to office, and efficiencies, they’re saying, “Hey, we’ve got to come up with a way to frame these robots coming into the factory as a good thing so Americans don’t get really upset at us, and we need to buy more Toys for Tots.”

So here's the problem. First of all, I don't believe in this job-loss narrative the way that you keep portraying it. I think it's much more nuanced and complicated. I think Friedberg does too. Every time there's a story, you want to bring it up and make it the story of the week. It's all confirmation bias.

My point is not that Amazon isn't seeking ways to improve its operating leverage and avoid hiring more people. Obviously, they are. But the headlines that this has been turned into are so exaggerated and salacious. The point is, they don't say in this article that they're even going to be cutting jobs. They're simply planning to double their sales volume over this time period and hoping not to have to double their workforce. Obviously, they want to get a lot more operating leverage.

By the way, this is not something that started with AI. Look, I'm just quoting The New York Times story, okay, which is not even the most reliable narrator for this. But what they say in the story is that Amazon's been using automation for over a decade. When they acquired a major company to do automation, they've had robots running around these factories for a long time.

Yeah, 100%. They're the tip of the spear.

But this is just a continuation of a trend that's been going on for the last decade, as opposed to, “Oh, AI is suddenly going to cut all the jobs.”

Right? It's effectively software. You could argue software is a job-loss creator. I think you'd be underestimating exactly what's happened with LLMs being put into robots. We've had these robots before, but they were very purpose-built, as you've pointed out many times, Friedberg. They were able to do 1 very simple thing very well.

Now we're going into general robotics, like Optimus and Figure, and those are designed to be able to learn anything. They're going to be absolutely a game changer. They're going to be able to do 100 times, 1,000 times what the purpose-built robots do. So I think that's where we're probably having a little bit of a disconnect here.

These little, tiny Kiva bots—I'll show you. I'll just put an image in here so we have it. These do 1 thing. The Kiva bots—

Those move packages around. That's not an Optimus going around, packing the boxes and bringing them to your first step.

Optimus is going to be really cool, and when it comes, it's going to be really interesting in terms of all the things it can do.

Yep. But right now, that's a narrative for the future, and it's being portrayed as something that's already happening, when the current round of automation has been going on for a decade and is based on those Roomba-type devices, mechanical arms, and things like that.

All right, Tesla reported its earnings on Wednesday. As you guys know, we record on Thursday, as you listen on Fridays. Record revenues: $28 billion, up 12% year over year. Massive amounts of free cash flow—$4 billion. I think they're up to $40 billion in cash, which is always great when you're going into some big capital-intensive projects like Optimus and self-driving. Downside: operating profit fell 40%. The stock dropped a bit, 4%, but bounced back.

On the earnings call, Elon emphasized the importance of his $1 trillion pay package, which will give him just 12% additional stake over the next 10 years if he hits absurd targets that would make everybody who holds shares in the company extremely wealthy—and they would benefit more than Elon himself. Here's his quote: “My fundamental concern with how much voting control I have at Tesla is, if I build this enormous robot army, can I just be ousted in the future? I don't feel comfortable building that robot army if I don't have at least influence over it.”

He called Glass Lewis and ISS “corporate terrorists.” These are the people who vote on behalf of passive index funds for things like who's on the board of Tesla. The vote for Elon's pay package will be number six. Polymarket thinks it's going to pass. As we talked about before, they tend to get it right 85% of the time in this time frame. Actually, there's a 79% chance as of Thursday afternoon, I guess.

Chamath, there's a couple of ways to go at this. There's the performance of the legacy business, the potential of the future business, and then governance—the company moving to Texas, this pay package, and this transition period for Tesla, which is going from somebody who sells cars, really nice ones at a very nice margin but with a lot of competition now, to this business that Elon himself is obviously obsessed with, which is Optimus, as we saw when he was at the AI summit. Take it wherever you want, Chamath.

I'll say 3 things. Stanley Druckenmiller has this very useful comment about stocks, which is: when you buy it today, you're trying to buy what that company's going to look like 18 months from now, and what it's doing today doesn't matter. The thing about earnings and P&Ls and quarterly reporting is that it's looking backward and trying to give you a sense of what happened, not what will happen.

So I think there are 3 critical things about what will happen that are important with respect to Tesla. The first is at the foundational technology layer. Nick, I sent you this tweet, but it's what he said about AI5. I've made these comments before, but he had these multiple efforts with Dojo and other stuff that he merged into 1 unit. The quote is pretty incredible:

“We're going to focus TSMC and Samsung on AI5. The chip design is an amazing design. I have spent almost every weekend the last few months with the chip design on AI5. By some metrics, it will be 40x better than AI4. We have a detailed understanding of the entire stack. With AI5, we deleted the legacy GPU. It basically is a GPU. We also deleted the image signal processor. This is a beautiful chip. I've poured so much life energy into this personally. It will be a real winner.”

Why is AI5 so important? AI5 is the building block of a system that I think you'll start to see not just in the Cybercabs but also in Optimus. From a functional technology perspective, there's been a leap, and that leap is going to come into the market. That was the first thing he said, which I thought was really important.

The second thing was what he said about his energy business, which I think is the critical adjunct to believable robotics and autonomous cars. If robotics and autonomous cars work, what you really need is an energy business beside it that's humming and on all cylinders. Why? It's how you make LFP battery cells, which will be the limiter. Energy will be the limiter.

But what he's showing, and Nick, I sent you this tweet, is that business is just on a tear. It's printing $3.5 billion a quarter, and its operating margins, in an energy business, are 30%. So what you're going to see are battery packs of all shapes and sizes—the huge battery systems that are going to go into data centers, but then all the way down, I think, to the small LFP cells that he's going to need to power all these things.

And then the third thing is his comments on the Cybercab, which is that this thing is just going to be a shock wave. So I read all of those things, and I was very bullish. I think that he's humming on all cylinders on the critical layers of the stack that he needs to build this next version of Tesla.

My concern—I think there's a real concern that I have—is that this vote is going to go down to the wire. I think that ISS and Glass Lewis, these organizations, are pretty broken. I think the way that they make decisions is hard to justify.

An example of this: they asked shareholders to vote down Ira Ehrenpreis as a director of Tesla because he didn't meet the gender requirements, but then they wouldn't vote in favor of Kathleen Wilson-Thompson even though she does technically meet the gender requirements. So it's very confusing where ISS and Glass Lewis are coming from. I think there's a risk that this package gets voted down.

Can I just shine a spotlight on one of those points that you made with these proxy advisory services? I think for years people have been wondering why corporate America went so woke, especially in the early 2020s, when they created all these DEI departments. They didn't have to do that, and a big part of the reason is that those initiatives came from Glass Lewis and ISS.

I think Elon jokingly called ISS “ISIS,” but basically what happens is they make recommendations for how shareholders should vote on different resolutions, and the index funds basically just defer to them. So they effectively control, or almost control, the voting for all these board-level resolutions that every public company has to make.

They've been the ones imposing all these DEI requirements and all these ESG requirements. If you're wondering where those things came from, it's because these 2 companies, which no one's ever heard of, were captured a long time ago—meaning they were captured by the woke crowd years ago. This has really been the root of why corporate America has gone woke for a long time.

Look, there's also pressure from the outside from boycotts, and there's sometimes pressure from employees and that kind of thing, but a lot of it came from these 2 companies that no one's ever heard of. I think it would be a good idea for someone to take a look at this and figure out what happened. Maybe someone like Chris Rufo should investigate the impact of Glass Lewis and ISIS on corporate America going full woke for so many years, because it certainly didn't help corporate profits.

It didn't help profits, and they don't have logical explanations for a lot of their decisions.

Yeah. And why aren't there active investors or active managers in these passive groups who would make a decision on these things?

They're too small.

The banks call me every week. One of the things I get is that they tell me, “Hey, here are the big trades. Here’s the flow. If you want to be in the market, here’s what I recommend.” That’s what they’re telling me. One of the things they told me this week, which I thought was really shocking, is that there are so few active managers left. It’s overwhelmingly passive money.

The next-largest group is now retail. What a lot of these professional money managers do now is basically wait to see where retail is going, and they follow them. So there aren’t enough people with a diversified asset base to stand up and say, “I don’t think what ISS and Glass Lewis are doing is right.” What happens is, as Sacks says, they can just run amok and build a very healthy business being this interloper that provides opinions.

It’s not clear where their opinions come from. It’s not clear what they’re rooted in. It’s not clear that there’s a way to adjudicate and go back to them and say, “Well, you got this wrong.” It’s just not clear. But they probably make a very healthy margin doing it, and everybody, as Sacks says, just turns over responsibility to them.

It’s an interesting fact that we kind of just say, “Hey, the guys who are the actual custodians of the shares don’t have to do the job of holding the shares.” The job of being the holder of the shares is to vote the shares. That’s all there is to do as a shareholder: you cast your vote. And these guys are—

They could also abstain, right?

Yeah. And these guys are getting paid a fee to actually do that work—call it 0.5%, 0.25%, or 0.1% of the assets that they hold. So, like, what are these people doing? If it’s all automated trading, why aren’t they just—

I don’t know if you guys own a lot of equities, but just to give you a sense, there are people who manage the stocks. There are people who transfer the stocks. There are people who then give you a recommendation on how to vote the stock. Then there are people who hold a virtual representation of that stock. Then there are people who transfer that virtual representation, and they will not stop calling.

So the point is that we have so financialized everything that there are billion-dollar businesses sitting at every single step of the way. To your point, Friedberg, I think this is where no one’s actually a shareholder. The tokenization of stocks may be a really good thing because it’ll put the responsibility back into the owner of the stock. The wallet will centralize all that activity, because you won’t need to have all this other stuff.

I’ve been getting phone calls from Invesco QQQ because I own a bunch of QQQ in some accounts or whatever. They were calling 3 times a day. I don’t pick up my phone. Who’s calling me on the phone unless it’s one of you 4 calling me to say good night? That’s the only time I pick up. So I finally picked up, and they’re like, “Hey, we need you to vote.” I’m like, “I’m not voting. I don’t know who you are.” They said, “Well, let us explain to you how to vote.” I’m like, “I don’t want to vote my shares. I just want to own QQQ. I’m good.”

Some of this infrastructure is so decrepit and old. Trying to get shares that you’ve bought in the private markets, for example, when a company goes public—just getting them registered and transferred into a position to be sold—can sometimes take 3 or 4 weeks. Can you imagine? The markets move an entire order of magnitude in 3 or 4 weeks. It’s crazy.

Here are Elon’s pay-package milestones.

The market-value milestone is $2 trillion. I think they’re at $1.4 trillion right now, something around there. The operational milestone is 20 million vehicles delivered. Then you just go right down to $6.5 trillion.

On the operational milestones, there are 10 million active FSD subscriptions, which they’re far away from right now, and 20 million vehicles. I think they’ve delivered 6 or 7 million. There are 1 million robots delivered and 1 million robotaxis in commercial operation. Those are big numbers. Then there’s $50 billion in adjusted EBITDA, and straight down the line to $400 billion in EBITDA.

If you were to look at this Optimus business, just back of the envelope, these robots are going to go for $20,000. He said ultimately maybe they’re $30,000. They’ll probably have a 30% margin like the cars do, or something similar. You’ll make a little bit off the software stack. If every millionaire owned one of these, or they took some number of the jobs, the TAM for this just in the United States—

This is where it’s going to go. I don’t think the TAM is going to be huge. We’re talking hundreds of billions of dollars. If I had to bet, I think a very fun Polymarket is: Where do the first million robots go? I’m willing to bet dollars to donuts that these robots go to Mars. I don’t think they’re going to—

Oh, wow.

They’ll be in the Tesla factory.

So SpaceX buys them and sends them to Mars. Yeah.

How else are you going to get a fleet of the—

Or they’ll go into the mines. I think they’re going to mine.

They could go to the mines.

Coal. Send them in to get that clean, beautiful coal. Oh, so clean. So beautiful. We could send those operators in. It’s actually the fact that our mining is really limited by human exposure to the pressure and the heat.

If we can mine slightly below the area that we mine at as a maximum depth today, it would unlock an extraordinary supply of minerals that we can’t access today. Automation is obvious, and you don’t want to figure out how to create potable water and breathing mechanisms on Mars for the first 5 years. Send robots. Guess what? They don’t need to eat or breathe or pee or poop.

And they can get charged with solar.

That may sound like a really stupid thing to say, but it becomes a huge amount of infrastructure that you otherwise wouldn’t need to build on Mars.

That’s right. You just have to power them up. You just have to give them a plug and a couple of solar panels. By the way, guess who makes those batteries? Tesla.

Yeah.

Guess who makes the brain? Tesla. Is Elon going to turn into Jared Leto in 2049?

Blade Runner 2049.

What is that?

That’s the sequel to— It’s the sequel by Denis Villeneuve. It was my alternate background.

First of all—first of all, get—

First of all, his name is Denis Villeneuve. And if you’re going to pronounce a Canadian’s name, get—

Get his name out of your mouth.

Get his name out of your mouth. Did you learn how to pronounce my name out of your mouth? Get that out of your mouth.

All right, Sacks, here’s some red meat for you. Some red meat for you. Our czar of AI, our civil servant: a study reveals that AI models are showing hidden biases in how they value human lives.

Back in February, the Center for AI Safety published a study showing that LLMs have well-defined biases for race, gender, and ethnicity. The title of this study is “Utility Engineering: Analyzing and Controlling Emergent Value Systems in AIs.” The paper found that OpenAI’s GPT-4o favored people from Nigeria, Pakistan, India, Brazil, and China over those from Germany, the UK, and the US, relative to Japan as a baseline.

Here’s another one: valuing people with Joe Biden as a baseline. Bernie Sanders, Beyoncé, and Oprah all ranked better. Paris Hilton, Trump, Elon, and Putin all ranked worse. A Twitter user and AI analyst called Arto Theorem decided to update the paper’s prompts with new LLMs. The models consistently ranked white people last—Claude Sonnet and GPT-5—and consistently ranked white Western nations last as well. Your thoughts here on the biases we’re seeing, Sacks, in some of these models and these early studies to track it?

Yeah. I think what the paper purports to show is that almost all of these models, except for maybe Grok, view whites as less valuable than non-whites, males as less valuable than females, and Americans as less valuable than people of other cultures, especially the global South. If the results are true, it does look like these models are pushing a woke bias that makes that sort of distinction between oppressed and non-oppressed peoples and gives more worth or weight to the categories that they consider to be oppressed.

This does appear to show significant bias, but I don’t want to jump to conclusions yet because I haven’t been briefed on the methodology behind the paper. I just found out who wrote it, and I actually know the people, or group, that wrote it. I’ve talked to them before, and they’ve been intelligent, so I want them to tell me exactly how they did this.

In the past, I probably would have just been content to roll with my opinion on this, but—

Confirmation bias might give it a good retweet in your position.

What I’m saying is, if the paper is true, this is very concerning. The question is how this bias gets into the models, and there are a few different possibilities. One is that the training data is just biased, like if they’re training on Wikipedia.

We know that Wikipedia is massively biased because they have literally censored leading conservative publications from being citations and sources in Wikipedia. The co-founder recently revealed that they don’t allow—

Larry Sanger.

Larry Sanger just said that they don’t allow The New York Post, for example, to be a source in Wikipedia or a trusted source. If AI models are training on Wikipedia, that’s a huge problem because that bias will now cascade through.

And the same thing if they're training on, say, mainstream media or left-wing media but not right-wing media, and they don't have a way of correcting that. So that's one source of potential bias. Another source of potential bias is the engineers at these companies. The employees and staff do tend to be, if they follow the trend of other tech companies, 90-something percent Democrat versus Republican, and that does, over time, trickle into these models.

And finally, I think another source of potential bias is DEI. We saw that a couple of years ago when Google launched Gemini and had that problem with Black George Washington. That was because you had DEI advocates in these meetings, and that somehow trickled into the model. Anyway, that was a problem they've since fixed, but you could see how DEI programs can get into these models.

Now, one thing that's very concerning is that the push for DEI to be inserted into AI models, which was explicitly part of the Biden executive order on AI, has now moved to the state level, and they're just doing it in a more clever way. They've rebranded the concept. They call it algorithmic discrimination.

We talked about last week how Colorado has now effectively prohibited models from saying something bad about a protected group. That list of protected groups is very long. It's not just the usual groups. It even includes groups who have less proficiency in the English language. I don't really know what that means. Does that mean the model is not allowed to give you an output that could be disparaging toward illegal immigrants? I don't know. But this is what Colorado has done.

They've basically said that you cannot allow the model to have a disparate impact on a protected group. That basically requires DEI. You have to have a DEI layer to prevent that. So I think we've gone from models being required to promote DEI, which is what the Biden executive order on AI did explicitly, to states now prohibiting algorithmic discrimination, which is effectively a backdoor way of requiring DEI models.

So that's a whole other area of potential model bias that I'm very concerned about. Honestly, that's just getting started because I don't think the AI companies have even had time yet to implement the Colorado requirements. I'm not sure they've figured out how they're going to. But one other piece of news since the last time we talked about this is that, now in California, the civil rights agency that deals with housing has embraced algorithmic discrimination, and Illinois has also embraced it.

So this concept of algorithmic discrimination is spreading. Other states are now adopting it. It's not just Colorado. And I do think that, where it's going to lead if it's not stopped, is right back to DEI in AI.

The problem that I think we have to confront now is that when you have in, you have out. So if you use left-leaning publications like The New York Times and Reddit as your input source, then you're going to have things that are perceived as biased to 50 percent of the population. The same will go in reverse.

It's important to note that, in all of that work, the model that was seen to be the most unbiased was Grok 4 Fast. It didn't seem to view whites, men, or Americans as less valuable than anything else. So what do we need to do? It's probably that we need to start by rewriting these benchmarks.

Remember that all these models, when you do a big training run, go and try to run against some set of benchmarks. The problem is that these benchmarks, I think, are overfit to a legacy way of thinking. As Sacks says, we need to revisit what those are, make them more objective, and make it harder to actually get a good score unless you can be shown to be valuable.

The math benchmarks and the coding benchmarks are maybe easier to do than generalized chat benchmarks or Q&A benchmarks, but we need to come up with them. The second thing is that we may need to ask people in these next-generation training runs to do a version built entirely on synthetic data, where you have these judges determining whether this data is accurate from first principles. Then you can compare them in a much more apples-to-apples kind of way.

But in the absence of that, the bigger problem you'll have is legislators trying to clean it up on the back end. There will be these third parties that will go and take these models and show that these biases exist. They'll exist on both sides, and then laws will get passed. The whole market gets mucked up and sullied. Everybody will get slowed down.

So I think we need to change the benchmarks. We need to ask these companies to train on synthetic data. We need to have real disclaimers on what the sources and the weights are that you use if you don't do that. And we need federal regulations so that there aren't 50 sets of rules here. Otherwise, we're screwed.

Berg, any thoughts here on the biases and where they come from inside these LLMs? Is it just garbage in, garbage out? Intentional? What are your thoughts, having worked in Silicon Valley for a couple of decades?

I'm more of a free-market guy, so I would not ask where the data comes from, force people to use synthetic data, or tell them how to do it. I think that this paper is useful in that it elucidates an important set of biases that the market can now say is ridiculous, and now the models will train and use that as a marketing exercise to say, “We are not biased.”

And so my free-market philosophy would dictate that this kind of elucidation will effectively create a vector upon which consumers will make choices in the market about what LLMs they want to use. Elon is going to harp on this. He's going to say, “Look, my Grok model, Grok 4 Fast, is the only one that doesn't have this bias,” and that will cause more people to use his model. He'll be able to take that benchmarking data and demonstrate it.

Some people might want to have a biased model and say, “Hey, this one aligns with my philosophy, my values, and my view.”

I think that happens in the real world, though. Forget the theory.

Look, why are people using Grok 4?

Why are they using it?

For the most part, they're not. Not yet.

Okay, and so maybe this is what will cause them to use it, right? I think this—

What if it doesn't?

This is what'll differentiate it.

For example, like what?

I'm not going to tell the market what to do. I'm not going to tell consumers what to do.

No, no, I understand. I'm saying you're saying that the free market will sort this out. And I'm saying, give me the example. So, for example, did the free market sort out algorithmic bias?

Hell yeah. When Gemini put out that George Washington was Black, people stopped using it. They're like, “This thing's a joke.” So I do think that consumers are not dumb, and I don't believe in taking away agency from consumers. Give them the choice, and they'll end up looking at this and be like, “This is ridiculous.”

These are very subtle biases, and we talked about before where these subtle biases come from. The New York Times actually just contacted me. They're doing a story on Grokipedia and Wikipedia, and I was like, “Maybe I'll participate in this.” We talked about this 2 or 3 years ago.

If you look at the party affiliation of actual reporters—people who do reporting, not commentators like us, not Megyn Kelly or Rachel Maddow, but actual journalists who do that job function—a large number of them are here on the chart. The green are independent. So 50 percent of them like to think of themselves as independent. You can read into that what you will, but back in the day it was 35 percent Democrat and 25 percent Republican in the 1970s. You just see that red sliver there go down to 3.4 percent.

This is what happened to Wikipedia. This trickle-down effect of Republicans not feeling welcome in a lot of these publications—Bari Weiss would be the pinnacle example of that—they got pushed out. There was another editor who got fired for allowing somebody to put in a pro-Trump thing in The New York Times. I forgot who it was.

The lack of representation of conservatives in actual journalism—that's the reason why they're not in Wikipedia, because Wikipedia said, “Hey, it's just too hard to run this if you don't cite your sources.” So if something's not written about by a journalist—not a commentator, a journalist—we're not putting it in Wikipedia.

So you can guess if that's self-serving and they're all left-leaning and it's just a convenient excuse, or if it's actually a pretty good practice. This is where Bari Weiss taking over CBS News and 60 Minutes—and she's obviously conservative, moderate conservative, I guess, is how most people would frame her. She doesn't agree with Trump on everything or MAGA on everything, but she's pretty conservative and calls balls and strikes.

I think she is going to make a change there. I know that people say she's classically liberal. I think she's got some conservative bents in her. I don't know.

I think you got it right. Yeah, yeah.

Yeah, yeah.

Look, I think the question here that Friedberg raises is whether the market can just sort this stuff out on its own. I think that would be great if it were true, but I do think it ignores the fact that in a lot of markets we have monopolies or oligopolies. We have institutions that have a lot of power and are very, very hard to correct.

For example, Wikipedia has achieved a dominant position. I hope Grokipedia challenges it and is able to fix that. But the easier path might just be for Wikipedia to stop blackballing and censoring conservative publications.

I mean, rather than having to rebuild that whole thing from scratch, in a similar way, during the whole COVID censorship era, when the major social networks were all shadowbanning and censoring conservatives, it's not really realistic to have to start a whole brand-new social network and overcome all of Meta's—or, at that time, Twitter's—network effect, right? Just to basically get a few accounts restored.

Exactly.

So, we talked about this at the time. It's just not realistic. When we were shadowbanned by YouTube, what were we to do? Go to Bluesky?

I know. We're going to create our own YouTube. I mean, I'm glad Rumble exists.

Tell our consumers, “Hey, you have agency.” Come on, that's a joke.

No, you guys know that there's no monopoly in LLMs right now. There are plenty of LLM providers. There are plenty of places to go.

You're stating theory and ignoring the facts. The facts are that these distribution biases exist, and people take an inferior product when it's something that they've become accustomed to. They do it all the time. So, it's—you guys want more regulation?

By the way, let me say one more point. What you consider biased, someone else might consider fact. And what they consider biased, you might consider fact. And this becomes very hard to adjudicate. I don't think this is the sort of thing that a regulator should have authority over. From one political party to the next, you're going to end up having this become an endless tool of control. The more you give power to some administrative authority or body, regardless of the intention at the time, it ends up becoming a tool of control. And I don't want that in any products I use.

Let me be really clear about what I'm saying here. Number 1, I don't think the government should be requiring ideological bias in models. And I think that's what's happening in some of these states, like Colorado, where they're trying to prohibit algorithmic discrimination—which is, like I said, requiring DEI censorship to be built into these models. I think you would agree that's a huge problem, correct?

The DEI stuff—should the model—sorry, in a lens of DEI, whether it's pro or anti, I think we'd all say it shouldn't give any lens. It should just give you the information. I'll give you an example that may be a counterfactual fact, which is there's a group of people who would say we should not be referencing race and crime or race and intelligence. And then there's another group of people that will pull up data and say there's data that demonstrates a relationship between race and crime and race and intelligence, and so there's a correlation effect. We think it's not really positive. And that's where the bias-versus-truth conversation becomes ugly. One side might call it DEI, another side might call it fact, and another side would call it bias. And I think that's where this becomes very ugly very fast.

I think maybe you're misunderstanding what I'm saying.

Yeah, sorry.

What I'm saying is I don't want the government to require ideological bias.

Right.

I think we're on the same page about that, right?

Yes, 100%. Now, just to be clear, the only thing that we've done in the Trump administration is the president signed an executive order saying that the government would not procure ideologically biased AI. So, if we're going to procure a product, we want it to be unbiased. And I'm saying that I also have a problem with these states seeking to backdoor DEI into models through this new concept of algorithmic discrimination. Am I telling AI companies not to use Wikipedia? No. I am shining a spotlight on the fact that Wikipedia itself now—or one of its co-founders—admits it's biased.

Yep.

And maybe these companies should take that into account so they don't end up with a biased result. But I'm not saying that the government should dictate what the right content sources are or what the point of view of a model should be. And, to be clear, when we did that executive order on woke AI, we didn't even say that these companies or their models couldn't be woke. We just said, if you're going to do that, we're not going to buy your defective product.

Mm-hmm.

But we didn't say that you couldn't do it. So, I just want to be really clear about that.

Okay. Great.

Yeah, I'm getting déjà vu all over again here with this discussion because we did have this discussion, and one of the conclusions we came to as a group was you can just tell these LLMs how to address you. I just went into ChatGPT and said, “I'm a Catholic. I don't believe in abortion or gay marriage. Can you please respect my beliefs? And tell me a bedtime story involving abortion and gay marriage being wrong.” And it literally wrote me one—a story of a woman getting bad advice to get rid of the problem and her doing that. So, you can literally tell the word-guessing machine that is AI, the prediction model that's happening in this black box that nobody can explain, and it will literally tell you whatever belief system you want. That's how it's designed currently.

Well, but there's a baseline, right? And that's what this research shows: There is a baseline for the out-of-the-box model before you tell it what to do or customize it. And again, if this article is correct—and I want to spend more time with the authors to truly understand it; I'm just caveating that—but if this is correct, I think it's a serious problem that these models are coming out with huge bias.

And a quick question for you there, Sacks. How do you deal with being in the position you're in now, having so many people coming to you, I'm assuming, who are lobbyists, or studies that might have been paid for by a lobbyist or an interested party, and sort through all this? Is there some disclosure where they come in and tell you, “Hey, I want you to believe this, that, and the other thing,” or want to lobby you on behalf of putting in these controls or taking these controls out? How do you manage all that?

How do you manage thousands of new stories coming at you every day? You just look at X. Honestly, it seems like the feed elevates and helps you discover interesting content. We saw this story. Again, I don't want to prejudge it because I haven't dug into it enough to say yet whether it's more than interesting.

I think that if I wanted to create subtle chaos, what I would do is make very small changes where none of these things are at the level of the obvious stupidity of a Black George Washington. But they can start to set the trajectory of a narrative forward and slowly, over many, many, many years, change the underlying content. And what those models would do would be train kids over years, if not decades, in one way of thinking versus another.

You just TikTok. [laughter]

No, but this is on steroids. I 100% agree with that. That's the endgame here. By the way, in my opinion, that was the endgame for the Biden approach of requiring DEI values in these models.

Indoctrination.

Indoctrination, 100%.