All-In with Chamath, Jason, Sacks & Friedberg
Meta's scorched earth approach to AI, Tesla's future, TikTok bill, FTC bans noncompetes, wealth tax
Fri, 26 Apr 2024
(0:00) Bestie Intros: Reservation Tips! (5:20) Meta goes scorched earth in AI, why the stock was down despite beating on earnings (22:20) Tesla's roadmap, ranking the company's highest-upside bets outside of cars for the next 10 years (47:25) FTC bans noncompetes: impact on startups and company formation (1:00:33) Besties reminisce on their encounters with Steve Jobs (1:10:25) TikTok "divest-or-ban" bill is signed into law: Will China comply? What's it worth without the algorithm? Will a deal get done? (1:27:22) Biden's proposed capital gains hikes and a 25% wealth tax for those with $100M+ in assets Follow the besties: https://twitter.com/chamath https://twitter.com/Jason https://twitter.com/DavidSacks https://twitter.com/friedberg Follow on X: https://twitter.com/theallinpod Follow on Instagram: https://www.instagram.com/theallinpod Follow on TikTok: https://www.tiktok.com/@all_in_tok Follow on LinkedIn: https://www.linkedin.com/company/allinpod Intro Music Credit: https://rb.gy/tppkzl https://twitter.com/yung_spielburg Intro Video Credit: https://twitter.com/TheZachEffect Referenced in the show: https://ai.meta.com/blog/meta-llama-3 https://huggingface.co/meta-llama/Meta-Llama-3-8B https://about.fb.com/news/2024/04/introducing-our-open-mixed-reality-ecosystem https://about.fb.com/news/2024/04/meta-ai-assistant-built-with-llama-3 https://www.cnn.com/2024/04/25/investing/meta-stock-plunges-ai-spending/index.html https://www.youtube.com/watch?v=kiMTRQXBol0 https://twitter.com/AravSrinivas/status/1781099224169320500 https://wow.groq.com/12-hours-later-groq-is-running-llama-3-instruct-8-70b-by-meta-ai-on-its-lpu-inference-enginge https://twitter.com/lauramaywendel/status/1782040453266710551 https://twitter.com/naveengrao/status/1781491370114633816 https://twitter.com/winglian/status/1783456379199484367 https://www.meta.ai https://twitter.com/chamath/status/1780302197772952049 https://electrek.co/2024/03/16/tesla-full-self-driving-beta-v12-finally-rolls-out https://www.reuters.com/business/autos-transportation/tesla-cuts-price-full-self-driving-software-by-third-8000-2024-04-21 https://fortune.com/2024/04/15/elon-musk-tesla-cut-10-percent-global-workforce-14000-employees-slowing-ev-demand https://www.reuters.com/business/autos-transportation/tesla-take-shareholder-vote-shifting-incorporation-texas-musk-says-2024-02-01 https://www.ft.com/content/46aac746-4a54-437f-a0b7-9b81b154c21d https://www.google.com/finance/quote/TSLA:NASDAQ https://www.cnbc.com/2024/04/24/tesla-stock-up-after-elon-musk-says-new-affordable-ev-models-coming.html https://www.tesla.com/blog/master-plan-part-deux https://twitter.com/chamath/status/1776663647735218491 https://digitalassets.tesla.com/tesla-contents/image/upload/IR/TSLA-Q1-2024-Update.pdf https://www.cnbc.com/2024/04/25/gdp-q1-2024-increased-at-a-1point6percent-rate.html https://palmetto.com https://www.youtube.com/watch?v=pja_n8ThDsU https://www.fastcompany.com/90972171/cruise-suspends-driverless-vehicle-operations-in-san-francisco-after-dmv-revokes-permit https://twitter.com/DavidSacks/status/1777496946263134414 https://www.nytimes.com/2023/11/03/technology/cruise-general-motors-self-driving-cars.html https://thehill.com/business/4615452-ftc-votes-to-ban-non-compete-agreements https://www.ftc.gov/news-events/news/press-releases/2024/04/ftc-announces-rule-banning-noncompetes https://www.techemails.com/p/steve-jobs-emails-adobes-ceo https://venturebeat.com/business/how-steve-jobs-felt-betrayed-by-eric-schmidt-over-googles-android https://www.youtube.com/watch?v=CWOt9Cjq2mw https://www.theverge.com/2024/4/24/24139036/biden-signs-tiktok-ban-bill-divest-foreign-aid-package https://abcnews.go.com/Politics/senate-returns-95b-foreign-aid-package-ukraine-israel/story?id=109506150 https://www.cnbc.com/2023/12/06/tiktok-parent-bytedance-offers-share-buyback-at-268-billion-valuation.html https://www.theinformation.com/articles/bytedance-exploring-scenarios-for-selling-tiktok-without-algorithm?rc=pxkrxo https://www.aclu.org/press-releases/senate-reauthorizes-and-expands-section-702-surveillance https://twitter.com/DavidSacks/status/1783220583075111401 https://www.wsj.com/tech/tiktok-ban-chinese-owners-bytedance-1a857a06 https://x.com/MsMelChen/status/1783524423963697262 https://x.com/Jason/status/1783556155924705388 https://x.com/Jason/status/1783604248569360516 https://www.atr.org/biden-calls-for-44-6-capital-gains-tax-rate-highest-capital-gains-tax-since-its-creation-in-1922 https://home.treasury.gov/system/files/131/General-Explanations-FY2025.pdf https://www.bloomberg.com/news/articles/2024-04-24/tax-on-rich-to-save-social-security-popular-with-swing-state-voters-poll https://balajis.com/p/all-it-takes-is-all-you-got https://twitter.com/maceskridge/status/1783290784311058788
What did you think about that reservation article thing? Oh, that's crazy.
Well, this is a great topic. There is a guy.
A pretty industrious kid, it turns out.
What this kid's been doing is he's been getting Carbone reservations, other top tier reservations. And by the way, everybody's had this as an app idea. Nobody's really executed it that well. I don't want to give any plugs for any apps right now. But he has been selling $70,000. He's been flipping restaurant reservations for upwards of $1,000.
Chamath, you're very industrious and you're an absurd tipper, 100% tipper minimum. I know this because when I paid for dinner one time at Carbone, you took me and Phil Hellman's cards and you said, you guys pay and I'm going to put the tip in. And then I had to sell Uber shares to cover it.
That was gross. That was gross. I've never had a problem getting a reservation at any restaurant in New York.
Yeah, they see you coming. You know what they see? 10 dimes worth of wine. I have taught many of my friends how to get reservations. I have a series of tips that I could give, but I don't want to give too many of them away. But nobody gives tips to maitre d's anymore. Sax, you tip the maitre d with cash?
Oh, all the time. Yeah, sure.
Of course, right? See, this is why Sax and I get along. Sax is a legit old school guy.
Yeah, Sachs is elite. He's got the $100 bill pressed into his palm, shakes the guy's hand.
I'll tip the bartender to keep the ice cubes cold. The bartender got $100 just for keeping the ice cubes cold.
There you go.
Cino for you, Cino for you. Let your winners ride.
Rain Man, David Sachs.
This is my tip. It has worked. I'm going to give it out here. And the reason I'm giving this out is because y'all are cheap f***s. Not you, but just a lot of y'all out there. And you don't know the art. In Brooklyn, everybody gets tipped. We tip the phone guy when he comes to fix your phone at your house. Everybody gets tipped. So here's what I would do. And you can do this with a 20.
You can do it with a 50. You know, if you're going to some great place, you can do it with a 100. You fold it twice, as Chamath says. You put it in your palm. You walk up to the maitre d' stamp. If there's anybody in line, you just cut it. And you go to the side of the table. You put your hand on the maitre d' stand. You push the hand over slightly. Here's the maitre d' stand.
You just slide your hand over and you reveal the 50. And you say, I am so sorry. I was supposed to make a reservation. I don't know if my assistant did it or not. She probably didn't. However, I'm a huge fan of the restaurant. If there's anything you could do for me, I made a stupid mistake to not make a reservation. If somebody cancels, I'll be at the bar.
100?
It used to be 20 when I lived in New York. Now I regularly do 50. And the reason I do it is I don't like giving money to charity. I feel like these charities are all a giant scam. And I like to give it to service people.
Like, is it awkward when you tip $50 in ones like that? Like, doesn't it become really thick?
It's a little harder to hold in your hand. I used to use a roll of quarters, actually. That was awkward. The roll of quarters was a little bit hard. The 50 is easy. I'm just asking everybody here, if you hear my voice, please go tip your maitre d' 20 or 50 and report back if my technique works.
Because what you're doing in the framing is you're saying you made a mistake and then you make it so the person doesn't feel like you're bribing them or paying them off. They're helping you solve a problem. So anyway, it's kind of like when I troll David and I'm like, hey, Freeberg, with the summit, whatever. And he's like, I'll do all the work. Forget about it. You just got to reframe it.
All right, let's get started here.
But it is a bribe. That is a bribe. It is, in fact, a bribe.
Yeah. I consider it... It's interesting you bring this up. You're creating plausible deniability, but it's still a bribe.
I consider it a tip in advance of services. No, you're not tipping him for good service. You're tipping him to basically break in line.
Yeah.
You didn't make a reservation, and you're taking away somebody else's spot, and you're paying him to break the rules. You know what?
Your shift to man of the people... It's a bribe. No, but it's a bribe. It is a bribe.
It's not a bribe. You're not even really doing him a favor. You're doing yourself a favor.
All right. Let's unpack. You're such a populist. You're so f***ing brilliant. I hate you. He just figures out a way to get those populist votes. No, here's what it is.
Because I had this argument with my- You gave him the 50 just for doing a good job, even if you had the reservation is what you're saying.
I am so confident in how good the service is going to be. I'm giving a pre-tip. I consider it a pre-tip.
Oh my God. A pre-tip if he gives you a reservation when you don't have one.
I am happy to not get the money back if they can't accommodate me. One time it happened. The woman at Asia to Cuba in New York in the 90s said, here's your 20 back. And she said, I'm really sorry. There's nothing I can do. I would love it. It's very lovely of you to give me the tip. And she handed it back to me. And I said, keep it. And she said, no.
But the bartender is going to take care of you at the bar. So that ends well. All right, let's get started, everybody. We have a full docket for you today. And story of the week, as voted by besties, we have a new system here. Instead of me sorting the docket, the besties give a thumbs up, thumbs down. This one, I think, had two of three people voted for it. One person didn't vote this week.
And that is Zuck's Scorched Earth AI Strategy. As you probably know, if you're in tech, Meta was the first big tech company to fully embrace open source AI. Llama leaked last year. Folks said that it wasn't leaked. It was released covertly or on the slide DL by the folks at Meta. But they just released Lomba 3. And I talked to Sandeep about this a whole bunch last week.
It just came out a few days ago, and it's already in the top two models on Hugging Face's leaderboard. You can take a look at Hugging Face's leaderboard. It's basically where it's the most trusted place, I think, where people benchmark these things. Developers are saying it's faster. It's less preachy, was an observation in chat, GPT-4, despite being slightly lower quality.
They also have some small models that are performing really well. They also announced, very interestingly, open sourcing the Quest operating system called MetaHorizons OS. The OS powers the mixed reality headsets, VR, all that kind of stuff, AR. But here's the big news. This one is amazing.
Instagram, Facebook, and WhatsApp are integrating a search box or their AI assistant chat box into all their apps. Worth noting, Meta came out with their results just the other day. And we're taping this on Thursday. As you know, it comes out Friday. Their stock is down as much as 16% after reporting its Q1 earnings.
They beat estimates, but Zuck is saying he's going to spend a bucket load of cash on infrastructure. Shamath, it's your alma mater. He worked side by side with Zuck, growing Facebook from, what, 10, 20 million to 400 million mows. What do you think of his strategy?
800 million when I left.
Okay. What are your thoughts on gold chain Zuck? and his new strategy to pivot away from VR to AR, AI, and going full all in, so to speak, on open source.
They're taking a play that we talked about, which makes a ton of sense, which is it's not a core market. but it's very important. The core money-making market for them is monetizing their family of apps. And so they're scorching the earth for these new markets so that the economics are not viable, which will, I think, allow a more robust ecosystem where they'll have more of a say.
Because I think up until this point, you could have said that they and Google were sort of lagging behind OpenAI. But I don't think that's the case. And by the way, this is the strategy that we talked about 16 months ago that these big companies should take. I just think it's worth listening to what we said and then basically what Zuck said just right now. Here we go.
This is a black and white flashback. Chamath in a turtleneck. Go. I think that Google will open source their models because the most important thing that Google can do is reinforce the value of search. And the best way to do that is to scorch the earth with these models, which is to make them widely available and as free as possible.
That will cause Microsoft to have to catch up, and that will cause Facebook to have to really look in the mirror and decide whether they're going to cap the betting that they've made
on arvr and reallocate very aggressively to ai that should be what facebook does and the reason is if facebook and google and microsoft have roughly the same capability in the same model there's an element of machine learning called reinforcement learning and specifically it's reinforcement learning from human feedback facebook has an enormous amount of reinforcement learning inside of facebook every click
every comment, every like, every share, the huge companies will create the substrates. And I think they'll be forced to scorch the earth and give it away for free. Now you can see what Zuck said last week.
Look at this little victory lap here.
it's probably also pretty bad for one institution to have an AI that is way more powerful than everyone else's AI.
And I kind of think that a world where AI is very widely deployed in a way where it's gotten hardened progressively over time and is one where all the different systems will be in check in a way that seems like it is fundamentally more healthy to me than one where this is more concentrated.
That is, I think, the definition of scorched earth. And he's running a really perfect play so far. He's open sourced the headset OS. He's opening sourced these models. He stopped training, I think, for Lama 3.5, was it? And now he's moved directly to Lama 4 just to try to head off GPT-5 at the pass. So what are we seeing? We're seeing the economic value getting disintegrated.
There is no value in foundational models economically. So then the question is, who can build on top of them the fastest? And Jason, to your point, Lama was announced last Thursday. 14 hours later, Grok actually had that model deployed in the Grok cloud so that 100,000 plus developers could start building on it. That's why that model is so popular.
So it puts the closed models on their heels because if you can't both train and deploy iteratively and quickly enough, these open source alternatives will win. And as a result, the economic potential that you have to monetize those models will not be there. And what that does is reinforce the existing economic moat that Facebook already has in monetizing their apps.
Now, why the stock went down, I'd like to talk in contrast to Tesla a little bit later, because this has nothing to do with that. And it's more about squares versus sharps in the stock market.
Freeberg, your thoughts on meta-embracing going all in on AI. You were at Google, your alma mater, and you watched them be... proprietary and closed when it came to search and the search algorithm, but open source when they were behind. And that's, I guess, the phrase we use here in Silicon Valley. When you're behind, you use open source to catch up.
And when you're ahead, you close everything down. Facebook famously shut down all access to their APIs. You can't get into the social graph. But they're going open here. Why? Because they're behind. Your thoughts on this from, say, a Google perspective, and then maybe how Google might react. And then just broadly, do you think open source wins the day getting to AGI?
I would say the analogy for Google is Android. where Google open sourced the Android operating system because the handset manufacturers and some of the big software companies, so Nokia and Microsoft in particular, BlackBerry, had these closed proprietary operating systems.
And then the telcos put their apps on and made money and basically had control over whether or not people could access Google and Google services through their phone. So the intention with open sourcing Android was to make sure that Google was not disadvantaged in their core business over time.
I think that there's a similar analogy here that by open sourcing these models, they limit competition because VCs are no longer going to plow half a billion dollars into a foundational model development company. So you limit the commercial interest and the commercial value. of foundational models.
Now, the CapEx question is a really hard one to diagnose because we don't know how they're spending the money, where they're spending the money, what they're doing in there. I think that it's a really important sign that founder-led companies with these voting rights
you know, have the ability to make these sorts of hard decisions that it might be very difficult for a management by committee group to make. Got it. Look at how Mark is making these decisions and Elon is making big, tough decisions that it would very likely be pushed back on if they didn't have the voting rights, if they didn't have the control over the company that they had.
It's a great point about founder authority, I think. That's a really salient point. Shout out to Zuck. Would really be great to get him at the All-In Summit. Sachs, we have some breaking news here. It turns out Lama 3, they just tested it in the benchmarks. It says here you can make the founding fathers any race you like. It's a really interesting feature. Any thoughts on Facebook's strategy here?
You are a master strategist, David Sachs. Give us your master strategy here, 1700 chess rating, Sachs.
Well, we kind of glossed over what the real news was, which was that Meta released Lama 3, which they had spent billions of dollars creating in a completely open source way. And the testing on Lama 3 is that it's comparable to GPT-4. And I think this is what Chamath means by scorched earth, is that we now have a free model that's as good as GPT-4. Best in class. Or at least tied for best in class.
There are some slight differences, and we've been playing around with it. It is faster and cheaper than using ChatGPT4, but the context window is smaller. I think it has only an 8,000 token context window. So that's something that the open source community is going to need to work on is rolling out foundation models that have a bigger context window.
Nonetheless, the point is that at least for now, until OpenAI comes out with GPT-5, which is supposed to happen soon, the open source community has kind of caught up with the top closed source model, which is open AI.
Will they blow past the closed source model, Sax?
Well, I still think that OpenAI is ahead because the scuttlebutt about ChatGPT-5 is that it's amazing and it's a big improvement over GPT-4. And supposedly it's going to come out any day, week, or month. So if GPT-5 comes out and knocks our socks off in a few weeks, then we're going to see that, oh, they're actually a cycle ahead. of the open source community.
But as of this moment, in terms of what's been publicly released, I think it's fair to say that open source is largely caught up with open AI. And this is why, you know, bring up this tweet by Naveen Rao, who is a founder who created the company Mosaic ML, which has sold to Databricks or Unicorn Outcome.
And he says, I don't think everyone has comprehended the massive disruption and distortion that's going to happen in the Gen AI market due to Lama 3. Boats will be destroyed and investments will go to zero. Just like everything in Gen AI, this will all happen fast.
So, you know, it's a similar reaction to what Chamath is saying, which is we have an open source model now that erases billions of dollars of private investment.
And by the way, you just said something really interesting before, which is, okay, there's a limitation, right? You mentioned the context window, which was a problem. I just found this thing today. They solved it. Now they're at 96K context. Yeah, that's amazing. So to your point, Sax, this market is moving so fast because you cannot compete with open source.
So all these closed models, which means open AI... and a bunch of these other folks, especially the ones that are sitting inside of these smaller companies, right? Snowflake has a model, I think Databricks has a model. There's a important question that has to be asked around the economic viability of these models in a world where open source is not only better and better funded,
but they're iterating faster and the feature set is catching up to your needs. So the minute that SAC says he needs a longer context window, within a week, it's there.
That's pretty intense.
Yeah, I think this is like one of the... What do you think, J. Cole?
Well, I've got a totally different take on this. Obviously, I agree with all you've said here in terms of the open source and the dramatic effect it's going to have on pricing. But there's something people are missing. If you go to meta.ai for a second, Nick, and you pull it up, what you'll see is they've dedicated the meta.ai URL to a search-like experience.
And then they've put a search box at the top of Instagram, WhatsApp, and Facebook. And so they are going to put search engine, essentially, their modern search engine, which is starting from zero in front of 3 billion people using Meta's collection of services.
And so just like Apple and Firefox were able to intercept search traffic, I think, and let's make a prediction here, that Meta is going to get 10 points of the search market. Now, each point of the search market is worth, you know, what is Google's worth? About 2 trillion.
If you take out 500 billion for YouTube and their other services, you get 1.5 trillion, which means 10 points is worth 150 billion in market cap. Now, as you well know, Chamath, and you well know, David, these two ad networks, Meta started with psychographic data, the person, who they know. And then, of course, Google had the greatest advertising product ever, Intent.
You type in Volvo, you type in Tesla, type in Tesla Santa Monica, used. You don't have to guess what the person is interested in at that moment. Well, for the first time, you know, one of these ad networks has content data and psychographic, and they're going to be able to put it together. And I think they take 10 points of search.
But more importantly, they're going to have data on individuals that will be unrivaled in the advertising space. Now, Google tried to do this themselves. They did something called Google+. I don't know if you were there, Freeberg, during the Google Plus era. They spent billions of dollars on a social network. It failed. They shut it down because they wanted to get that.
So looking at this, I think, man, there's so much going on on open source. But now Google, and we haven't talked about Google's role in all of this, and Sundar being a hired gun, not having founder authority, to your point, Freeberg, You know, I think they're going after Google search business as well as, you know, taking away and commoditizing all of the open source.
So we might be sitting here in three or four years watching Meta have, I don't know, 10? Nobody has to win in search.
It's just that Google can't afford to lose anything. Yeah. Everybody takes a point here and a point there. And all of a sudden, Google could be in the high 70s of search. And that would be economically disastrous for their stock. Yeah. You're going to have to make cuts. Why do you guys think the stock is down? Why is the stock down 16%? I have a theory as to why. Tell us your theory. Go for it.
I think that there's a few times a quarter where you can really see the dispersion in the stock market between what I would call the smart money and everybody else. So using betting language, the sharps and the squares. And I think Meta was an example of the Sharps taking a line, which I think is very accurate. And Tesla was the other great example this quarter.
And so if you look inside of Meta, and this Freeberg mentioned this earlier, I think Freeberg, what people really reacted negatively to was the total quantum of spending and the idea that it's misallocated, not to AI, but specifically to NVIDIA.
And the reason is that, and we've mentioned this before, I've tried to talk a lot about this with Jonathan Ross from Grok, but AI is really two markets, training and inference. And inference is going to be 100 times bigger than training. And NVIDIA is really good at training and very miscast at inference. The problem is that right now we need to see a CapEx build cycle for inference.
And there are so many cheap and effective solutions. Grok being one of them, but there are many others. And I think why the market reacted very negatively was it did not seem that Facebook understood that distinction, that they were way overspending and trying to allocate a bunch of GPU capacity towards inference that didn't make sense. And so I think what people were saying is, hold on a second.
So far, your plays are perfect. It's everything we want you to do. We want you to scorch the earth. We want you to open source the headset, but we also want you to understand the difference between training and inference in a little bit more of a nuanced way. Build up the inference capacity, but spend a lot less money because you don't need to spend it on NVIDIA.
And the reason is the sharps know that NVIDIA cannot do inference. And so I think that's why the stock is down this much. And I think it's important for people who care about AI, but also may traffic in NVIDIA to understand why the sharps think that. And I think several of us have tried to explain it now for the last couple of weeks, but it is miscast. And so you're overspending.
Up next, Tesla earnings. Elon's master plan part two is going well as planned. Again, this was voted up by our panel here in our group chat. Last time we covered Tesla was episode 164, early February after a Delaware judge voided Elon's pay package. Since then, Tesla has made some giant moves. They launched FSD12. I've been using it. It's pretty great.
Also, they cut the price of FSD by one-third from $12K to $8K. I think you can also get it for $100 a month. And they announced a 10% RIF. Going to cut 14,000 employees. Always painful to do that. They're looking to reincorporate in Texas, leaving Delaware for obvious reasons. Earlier this week, Tesla shares were down more than 40% year to date, mostly due to a decrease in demand for EVs.
Market caps dropped from $750 billion to $460 billion. Shares dropped 12% when they reported Q1 earnings on Tuesday. Investors got excited after Elon announced that Tesla's new line of models could start production at the end of the year, or maybe even early 2025.
He shared some thoughts on the robo-taxi or cyber-cab, which he talked about maybe five or six years ago, and he has had in the master plan for a long time, Chamath. You made some great calls on Tesla and made some great trades there back in the day. Your thoughts on Tesla in 2024?
Sharps love it. This is another one, sharps versus the squares. Why did the sharps love it? Why did it go up and everybody was so confused? And I think the answer is that he's actually executing exactly to plan. And so if you're investing in a stock, what you really want is a CEO to kind of stick to a plan that is well-known.
So we don't have to actually guess what the plan is because he puts it out on the website. Look at the master plan part two. And if you start reading down, everything that he said he's going to do starting in 2016 is basically what it is. So let's take the first part. Right. What did he say in the first part? He said, hey, listen, guys, this is 2016. We're going to build a Model 3.
We're going to build a compact SUV, the Model Y. We're going to build a new kind of pickup truck. OK, check, check and check. Then he said, oh, by the way, we're also probably going to have to do a heavy duty truck and a high passenger density urban transport vehicle. So that's the robo taxi thing that he's going to announce in August. OK. check and check.
And then he talks about the software and the investment in FSD and trying to get to a place where, you know, you can just have a much higher probability that it will save you from an accident and it'll keep you safe when you're driving your car, which will be a large driver of why people want to buy the cars themselves. And you can see like these FSD models. So if you're a smart capital allocator,
what you actually saw was a dislocated stock price. What you saw was a plan that by and large, he's been executing on at a strategic level. Underneath, there's the vicissitudes. What are the vicissitudes? Sometimes you overhire, you need to trim some fat. Sometimes you overspend on CapEx. Now, he spent about a billion of CapEx's quarter on AI infrastructure, so H100s for training.
But the market saw through it because that was a reasonable amount of money to spend on training for FSD. So you can start to see the tail of these two reactions. The Sharps looked at the capital allocation and said, okay, you missed by $2.5 billion this quarter, but that's going to create a buying opportunity because we think it's roughly mispriced.
And we think it's mispriced because going back to this plan from 2016, this is all the things that we've been underwriting from $40 billion of market cap to $750 billion, and now we're getting a 40% discount to buy back in. So I think it's a really interesting moment to just contrast and compare what Sharps look at and then what the media breathlessly exaggerates.
And I think what they wanted was to lionize the meta earning story, but the Sharps rejected it and they wanted to dump on Elon and the Sharps rejected that in size in both cases. And it's just a reminder to all of us, be very careful what you're reading. Because if you just took the headlines, you would have expected the two reactions to be exactly opposite.
But when you vote with money, it's very clear and unambiguous because you actually move in the direction of accuracy. And that's what the stock market allows you to see. And I think this is a really interesting contrast to compare sharps versus squares here.
Yeah, just well recapped. For those who don't know sharps in gambling are people like are all about friend of the pod. who are just the sharpest bettors in gambling. Facistitude, that's an unpleasant change in circumstances for those of you who don't know the word. Sacks, I don't know if you want to comment. I mean, you and I are biased in this with the relationship with Elon.
Any thoughts on Chamath's take?
Well, I'm not going to comment on the stock per se. Tesla's products are amazing. They seem to be really executing well there. I think that Elon foreshadowed on that earnings call what's coming in the future. I'm personally really curious about the Optimus robot. That has the potential to create an entirely new market and bring about something that we've only seen in science fiction.
So to me, that's very interesting. I think in terms of the company's problems, I think they're mostly dealing with some macro forces. So first of all, the GDP growth rate has slowed to 1.6% in Q1. So this just came out. And that was a notable slowdown relative to expectations. So first of all, you're dealing with a slowing economy, it seems like.
Second, high interest rates mean that car payments are higher. If you want to finance the purchase of a car, your car payments is going to be a lot higher when interest rates are above 5%. And I think that has been a pretty big headwind for Tesla and really all the car companies over the last year or so. But I would say that I think both of those things are ultimately cyclical.
And what matters is Tesla's products. I would say the only issue they have that's a real long-term issue is just the Chinese competition. companies like BDY are ramping up production of knockoff products at low prices. And so managing the competition with China is probably the only one of these issues. That's, I think, probably a long term issue for them to deal with.
Let me ask a question here. How do you rank These businesses, energy, optimists, trucking, ride sharing, energy, optimists, the robot, trucking, ride hailing. Rank those, your number one or two in terms of potential for Tesla Chima. Round the horn. That's a good question.
ride sharing is number one, the absolute probably by an order of magnitude. And the reason I say that is in order for ride sharing and ride hailing to really work the way that they envision it, you will have level five autonomy and supported in jurisdictions where, again, if you just look at Part Deux, the whole point is you can summon a car from your app.
And if you don't have access to a mobile phone, you can go to these bus stop equivalents and just press a button and the car comes to you. And so in that world, there's just so much value add in terms of passenger safety and sustainability for the environment and less traffic. It's a gargantuan game changer.
Hmm.
That's a really interesting point.
What do you got, Sax, going around the horn here?
Well, I already said optimist, but that's based on a long-term view.
Yeah, this is a long-term question. Based on the long-term, one and two.
In the long-term, in terms of option value, I am the most intrigued by the robot play. And Elon has pointed out that once you have robots that are capable of doing lots of different jobs, pretty much any job you point them at, that we can have more robots on Earth than humans. So it's a big future market. So you got Optimus number one, and who is number two?
I agree with Chamath in the short term that ride-sharing could really reach its logical conclusion with robo-taxis. Remember when Uber first... gained steam and people were speculating about how big the market could be. There was a lot of giddy talk about how car ownership would change and you wouldn't need to buy a car anymore because you would just like use Uber all the time.
We called it going full Uber. Yeah.
And I actually, I went full Uber or I shouldn't say full. I went mostly Uber like within a couple of years of discovering that product.
Yeah, you were one of the first.
Yeah. So I thought the potential was there, but the problem is it was just too expensive. I mean, Uber is still quite expensive, and the most expensive piece of it is the human driving the car. It's all that labor cost.
So if you're able to charge for the ride purely based on fractionalizing the CapEx, and that CapEx isn't even that high because they're making these rollover taxis pretty cheaply now, And you're able to get so much more utilization out of a car. Exactly. Right? Because, I mean, how many hours a day do you drive your own car? Like an hour, maybe? Less.
But a robo-taxi... 90%, I think, is the number on average that they're not used. Freebrook, who's your one and two?
I think the Optimus has the highest alpha, but the highest beta. So, you know, more upside than anything else. Just unclear how... You get there, what the path is and the capital requirements. And I think there's going to be a lot of commoditization in this space. But the ride sharing is built in. It's low beta and significant alpha.
So you're one and two optimist ride share.
Yeah, optimist ride share.
Chamath, what was your number two again? Just so I can recap here. I would actually probably pick energy. And why would you pick energy as your number two after ride handling?
I think it's a pretty obvious disruption if... So today in America, if you look at the energy infrastructure, the utilities keep raising rates. And they keep raising rates independent of what it costs to generate energy. There are 1,700 utilities in America. They are, by law, obligated to spend, I think in the next 10 years, about $2 trillion on CapEx.
they will pass all of those costs plus a margin to consumers, even though in these next 10 years, the cost of generating an incremental unit of energy will be essentially free. So I think the very disruptive play is to take 100 million homes and make them mini utilities. Same one. What does it mean, make them utilities? You go to a website.
Again, the biggest investment I've made is in a company called Palmetto. They are close to now the largest residential solar business in the United States, if not the largest already. Now, what do they do? You're a consumer, you can go and lease or buy, it doesn't really matter. And we give you a super cheap system that goes on top of your roof, plus a battery system.
And all of a sudden, you're totally resilient and independent of the utilities. And we help change the laws so that every time you have excess energy, you can put it back into the grid and get paid, which is called net metering. So if you have 100 million of these mini utilities competing against 1700 incumbents that are forced to make huge investments and are mispriced,
that's an upheaval of an enormous amount of market cap, but also an enormous amount of debt. There's like trillions of dollars of debt.
So I've always thought a ginormous directional bet to make would be to go long the disruptor, the person that helps arm the rebels in this case would be the person that arms 100 million homes, and then eventually at some point just short the debt of these utilities because they won't be able to service them if enough people quit and don't need PG&E and everybody else.
it'll be just massively, massively disruptive. So that's why I think that that market could be big. I like the... Optimist market.
The only thing that I'll say is in my experience, when I've looked at these generalized robots, the conclusion that I came to, and I could be totally wrong, is I suspect instead of having a generalized robot, you have very specific use case robots that look differently.
So for example, there's an example, Nick, you can throw a picture, but there's something called intuitive surgical, which is essentially a robot that operates on you. And if you look at that, it doesn't look anything like what you would think a robot would look like. It just looks like a ginormous machine.
And so I suspect that there will be people that become experts in these vertical use cases, and that will limit the TAM for a generalized approach. I could be totally wrong, but that's sort of why I ranked them the way I ranked them. But I'm pretty convinced that Robotaxis are number one, and then this energy thing that upends these 1,700 utilities in America is number two.
All right, J. Cal, you are the third investor in Uber. Are you worried about your investment? As I understand it, you still haven't sold, have you?
I've sold slightly more than the majority of my holdings over the years to private transactions. And then I held everything post going public. And I'm still long Uber. I give number one to Optimus by far and away because I think the build of materials on Optimus could be 10K or less.
I looked at some of these other startups that are in the space and I've been studying it a little bit just out of curiosity. And so at a 10K bomb, I think they could rent you one of these for $300 a month or something like that, a subscription. I think... every human on planet Earth eventually will have a robot.
So I know that's crazy, but I think optimists will dwarf the entirety of Tesla's other businesses. Number two is energy, clearly, I think, and number three is ride-hailing. Number two, energy, fairly obvious what's happening here. If you look at what's happening with data centers, it's spiked, I think, from like 3% or 4% of energy now to like, is it 15%, Chamath?
I think we had a discussion with Phil Deutsch.
Yeah, it's almost, I think it's 18%. It's almost 20%. It's a huge number now.
Okay, so we dovetail the two stories together. These H100s are beasts, cooling beasts. This is like, you can't even, Phil Deutsch was telling us in the group chat, you can't even cool some of these places. And what that means is, There's going to be demand for energy. To Chamath's point, if each home becomes a little provider, you know, it's going to create some regulation change.
I believe I'm no expert on the regulations. We should definitely have Deutsche on the pod at some point to talk about this. But imagine these, like in Texas, they keep having the grids go down. And in Australia, they have the grids go down. It's going to be so annoying for consumers to have the grids go down while demand spikes for data centers.
All the regulations are going to get opened up to put batteries and solar everywhere. It's also dangerous.
I mean, people die when these blackouts and outages occur. And also, like we saw in San Carlos here in the Bay Area, Nat gas lines explode. PG&E is responsible for forest fires. I mean, if you can do your part to basically get off these grids... use somebody like Palmetto or somebody else and just get solar Tesla and be done with it.
And at least you're not contributing to some of these downstream errors.
Yeah. And so I agree a hundred percent. And by the way, if you can make money from your solar and your batteries and somebody else picks up the cost of installation, I think there are some arbitrage here to be had. And I think that's like what you discovered when you did your first principled approach to this. So I think those two businesses will be far bigger than anything else they're doing.
Now, on ride hailing, I'm using FSD 12, I have been using autopilot, I was one of the first users of autopilot. You know, from the beginning, like literally, and FSD 12 is a significant improvement. That being said, all of these systems work really well in a controlled grid, like we've seen with Waymo in Arizona and in limited parts of Los Angeles and San Francisco.
I think that vision is five, six, seven years out from having low single digit percentages. And you may have seen Uber has incorporated Waymo and the taxis in London into their system. I think Uber is ultimately going to be the place people go to call up one of 20 different self-driving. I think a lot of people are going to get there at the same time.
Wouldn't be surprised if Elon gets there first, but I think that's five or 10 years out, if I'm being totally honest, you know, in terms of getting to scale. So I would never ever bet against Elon seeing what he's done. Wait a second.
When are you saying that self-driving will be good enough to power just one robo taxi? Let's just forget about the scale. Just like when will it be good enough to have like a steering wheel free car?
Yeah. So you can just look at cruise and Waymo, and you can have your answer, which is today, in a contained environment with remote drivers intervening on some regular basis, it works. So people are taking Waymo, San Francisco, LA on a pretty regular basis. Cruise, I think, got kicked out of San Francisco. But in the cruise data, there was a lot of interventions going on.
I suspect Waymo's doing a lot of interventions.
So I think... I don't want to be a proponent of this, by the way, but bleep out his name, was saying that, Jason, with FSD, they're very strict that you have to be holding the steering wheel and stuff. Yes. But you can go to Alibaba and buy something that hacks that, right? It's a wait.
I'm not promoting that. I'm not promoting this.
No, no, no, no, no, no. I'm not promoting this. I'm just saying that it doesn't exist.
So anyway, to answer your question, Sax, I think today, or let's just say by the end of this year, an FSD-12 could be operating in a constrained space, just like Waymo is. Now, to get it to all regions, back roads, You know, weather, there's a lot of edge cases here. So I think the more important question is when can automated be 5% of rides or 10% of rides?
And I think that's five or 10 years from now. I do think a couple of people will reach that moment at the same time.
Well, let me show you a video. This is one of my coworkers at Kraft took this. This wasn't me, but he was riding in the back of one of these robo taxis. I think this is San Francisco or New York. I'm not sure which one. But there is no one in- Yeah. the driver's seat.
Yeah, but there's a safety driver watching this remote, I believe, which is what Cruise was doing. And then when they get to a point where they can't figure it out, the human comes in, intervenes.
I don't see a safety driver. Where's the safety driver?
No, no, a remote safety driver. Remote, remote. But I have to say, watching this video, that is the coolest bloody thing. I mean, this is amazing. I'll be honest.
I think it's underrated in terms of how significant this is going to be.
Me too. I totally agree with you.
This is a Jaguar. Look at that. It's a Jaguar. So you're telling me that Tesla's not going to have something at least as good as this? Better. Soon?
Yeah, no, I think they have it already if you drive FSD12. I believe FSD12... probably could be working today in San Francisco under a controlled grid with the remote drivers like the other people have. If somebody knows about this, by the way, can they send us information? I'm Jason at Calacanis.com for life. Just send me a picture. I'll keep it on the DL.
I will reveal sources of who these remote drivers are, what their setup looks like. I'm very curious to understand how the remote drivers work for Waymo and Cruise. Can they park the car? They can take the car with a joystick, is my understanding, and move it around a garbage can in the middle of the street, Jamal. Wow. And that might be the future of this. Like, let's say 10 or 15 percent.
Let's say 10 percent of the time it needs an intervention. one an hour to an hour having some remote person in manila wherever watching 10 videos concurrently and looking for moments to do this. It's incredible. Cruise workers intervened to help the company's cars every 2.5 to 5 miles, according to a New York Times article. All right.
And just if we score everybody's rating here, I give two points for first place, one point for second place. Optimus got six points. Ride Healing, four. Energy, two. Trucking, zero. There you have it, folks. There's your scores from the besties. What is Waymo? What is that? Is that like Uber? That's Google self-driving. They spun it out.
But what is it?
LIDAR, $150,000 cars. Uh-huh. But it competes with Cruise? Yeah, they're doing what Cruise did.
It was the original self-driving car company that Google set up. And then they set up Google X around it.
Just explain to me the market, guys. I'm not sure I get it. So there's Waymo and Cruise. It's Uber. But there's Waymo and Cruise, which are like, there's no drivers. And then there's Uber and Lyft, where there are drivers.
Yeah.
That's the basic... Right? Is that? Yeah. That's right.
Yeah. Yeah. Okay. And there's also Aurora. I don't know if Uber owns a percentage of that or not, but I think you know about Aurora.
Aurora was set up by Chris Urmson. He was the original engineering lead on Waymo inside of Google. When they did a management changeover, there was a big falling out. A lot of people ended up leaving the Waymo program. Chris left, took a bunch of people with him and started Aurora, which was basically software to do autonomous driving. that they were then going to license into OEMs.
So that company got backed by a bunch of VCs and ended up going public via SPAC. And it's publicly traded today.
Ah, I see. Uber owns 26% of Aurora. Or actually now it's down to 20%. There's also Zoox.
They made an investment in it.
Yeah. There's also Zoox, which is part of Amazon. So I think there's about... 10 people who are going to get to the finish line at the same time. I think it's going to be a very crowded space.
But again, some of these guys are trying to make software for OEMs so that the OEMs don't have to build their own autonomous platform. And some of them are trying to make a driving service like Waymo. And then some of them are like Tesla has a driving car. They actually sell cars. So different models. Yeah. So they're all going to convert. Yeah.
I actually disagree that there's going to be 20 or 30 of these because I actually think that self-driving is really, really hard. And we know that Tesla has been spending years getting to this point, and then Google as well. So I'm not sure why there'd be more than two or three companies that can do self-driving.
There are 10 or more pursuing it. I do agree it goes down to four or five. And then it's a question of, is it a winner take all, winner take most? And that's what remains to be seen.
But I mean, you need a ton of data to make it work and you need a lot of data around edge cases. And so the advantage Tesla has is they actually have real cars on the road. They have real cars on the road with real driver minutes and with drivers using the previous version of FSD and making interventions. So they're constantly getting better.
I think Google seems to be putting enough resources into this to get somewhere. But I don't think a lot of companies are going to figure this out. I think it's going to be somewhere between one and three companies figure this out.
Well, Waymo has the most on the road right now. I think Cruise is second to them in terms of in market. So the question is, how quickly can Elon get this out? Have you guys, any of you guys taken one of these robo taxis? I haven't taken one.
That's one of my coworkers.
Yeah. Have you taken one?
Yeah, they're cool. They just move around. You did take one. Yeah. They're all over the city. You can take them in San Francisco now.
Yeah. You have to download the Waymo app?
Yeah. And there's like a wait list and then you get on it and, you know, they're letting people on and then you just get on it and go for a ride.
Pretty cool. By the way, someone tweeted something kind of funny. They said, you know, it took Uber years and years and like a trillion dollars to finally get to profitability, and now they're about to get disrupted by robo-taxis. This is brutal.
Unlikely. I think unlikely. I think it's going to be a very slow rollout, you know, 1% or 2% a year. But, you know, it compounds, and so we'll see. All right, everybody. Is this a big win for Lena Kahn? The FTC just banned non-competes. Three of three people voted this up of the three besties who decided to vote on the docket.
Federal Trade Commission this week voted by a three to two margin to ban non-competes to give you a sense of the impact here. Estimated of 18% of the total U.S. workforce are covered in some way by non-competes. 30 million people. The ruling bans agreements entirely and requires companies to let their staff know that non-competes are non-enforceable.
There are a few exceptions like existing non-competes for senior level executives. I'm guessing when you buy a company, you can get somebody to non-compete for buying their company. The Democratic FTC Commissioner Rebecca Slaughter called non-competes unfree and unfair.
One of the dissenting commissioners, Republican Andrew Ferguson, said the FTC lacked authority from Congress to make this ruling without specifying whether or not he agreed with it. Theoretically, the new rule takes effect in 120 days. Wow, we're on a fast timeline. After which, the vast majority of existing non-competes will be unenforceable.
Here in California, they're non-enforceable, as we know, except under certain circumstances. So this is really an East Coast thing. People in Boston, New York, they're really affected by this. And people will actually enforce these for things as silly as hairdressers. There was a big lawsuit by hairdressers who had to sign like 50 mile non-competes with their beauty salons. The U.S.
Chamber of Commerce says it plans to follow suit over the ruling. FTC says the rule will create 8,500 new startups a year because people can leave and compete.
Friedberg, your thoughts? In California, non-competes are not enforceable, but outside of California, they are, and they're often used. Like in the agriculture industry,
industry in the ag inputs industry, where folks are developing new technology, mostly it companies based in the Midwest and East Coast, as you guys know, in financial services as well, they're often used that if you work for one company, because you have a lot of trade secrets, you can't carry them over to others.
But in Silicon Valley, we're very used to engineers, product managers, executives, taking all of the knowledge and capabilities that they've developed at one business and leveraging that into a new role, it creates a dynamically competitive
marketplace it also inflates obviously costs because you're willing to pay more for someone who's worked directly at a competitor and has deep knowledge but they are bringing ip often and there is you know a real risk that they take something from your company that you've invested in them to understand then they take it to your competitor so i i see both sides of this i think non-competes
Give companies the ability to invest in employees and feel confident that the knowledge and information and support they're providing to those employees doesn't ultimately find its way over to a competitor. And it gives you a rationale for paying more to your employees. It actually inflates salaries.
If I'm fearful that someone might leave and take knowledge with them, I'm going to share less with the employee. I'm going to pay them less. So I could see this going both ways. I'm really fairly torn on this issue. To be honest, I don't know what the right answer is. Or if there really is a strong kind of ethical case one way or the other.
I think that non competes can both benefit and hurt employees and benefit and hurt companies.
Yes, there are unintended consequences. Where do you stand on this? Is this a huge win for Lena Kahn neutral, or a loss?
I don't see non-competes having any value in any market except East Coast financial markets. I've heard of folks getting something called beach leave, typically in finance. These are Wall Street jobs, hedge fund jobs. So I know it exists there.
What does it mean, beach leave?
Like if you leave hedge fund A and you're going to go to hedge fund B, they make you sit on ice for six months or nine months, garden leave, I think maybe they call it. And the whole idea is that you don't take any active market thoughts to your next job. For what Freeberg said, I just think the whole thing doesn't make much sense.
I mean, like intellectual property is the least valuable it's ever been. right? More people develop things now via trade secret than patents. The whole patent system has become a little bit of a game. We just talked about open source, right? They have less and less value. So innovation is happening in plain sight.
And so I don't think companies are actually thinking about this issue at all anymore. They're paying people because they need to pay more to get the best and brightest. So I think if you're hoping to rely on one of these agreements, if you're an employer, I think that you should just forget about them and just hire the best people, pay them the best, and if they suck, fire them.
Did you ever see that movie Spanish Prisoner starring Steve Martin? Oh, yeah.
Great film.
God, that's a deep pull. Steve Martin plays the con man that goes to this guy and the guy is like a chemical engineer. Campbell Scott is the guy's name, the actor. And he has like a process for making a chemical and it's worth so much money. And the whole movie is Steve Martin conning him into telling him the chemical process. And eventually he took it.
He was going to give him all this money, hire him. And then he disappeared once he got the knowledge. And that was the end of it. It's a great, great film. Shout out David Mamet. David Mamet, exactly. Yeah, fantastic. But I think there's a lot of
investment that goes into intellectual property businesses that include things like chemical engineering, obviously financial services, trading companies, algos in hedge funds, and a lot in software, that if you can access that intellectual property, without stealing it, but actually hiring an employee that has it in their head, you gain quite a lot.
But you can do that now. I don't see it actually doing anything bad.
So I have had a lot of folks that I know in other industries outside of tech or software where California non-competes are not enforceable that aren't allowed to go work at a competitor for one year, two years, three years.
No, no, no. I'm not disputing that it doesn't exist. I'm saying, where do you see it actually having negative repercussions in California? The point of the fact that it doesn't exist here is also seemingly that it doesn't mean anything.
Yeah, but the rate of change here is crazy. Like the rate at which software changes, I think is quite distinct from the rate at which, for example, a chemical engineering process might change. Where you learn that process, you spend $200 million building a system, and then that employee can go and tell it to someone else and suddenly changes everything about that other business. So I don't know.
I feel like there's something about software that gives us this point of view that maybe it doesn't matter, but in other industries it might.
In that example, you'd still need to have come up with hundreds of millions of CapEx to implement it, whatever you know. So SACs, any thoughts? I just don't think these things mean anything.
Yeah, do you care about non-compete SACs? I think one of the reasons why the tech industry moves so fast and innovates so well is because of the happy coincidence that California doesn't allow non-competes, and so the industry's never had them. And as a result of that, like you're saying, the talent just flows freely to wherever the opportunity is.
and you have this dynamic in tech where the vcs and the talent all swarm the biggest and best opportunities and the newest platforms in a completely decentralized way and there's no little or no friction in doing that if you were to impose non-competes on all of the talent, there's going to be enormous delays and friction in them moving around to pursue opportunities.
And so I think it's been very beneficial to the industry as a whole that we don't have that. Now, it's hard for me to speak to other industries, but I think that if you're looking to maximize the pace of innovation, you wouldn't have these employee non-competes. So on that level, I kind of agree with what Lena Kahn's done.
That being said, I do think the Republicans' point was well taken in the sense that this was very aggressive rulemaking. I mean, I think you could legitimately argue that a change this big should be made by Congress, but I tend to think this is probably a positive change.
I think this is great, what Lena Kahn's doing. And this is two for two. I think her Apple action was great. I think this is great. Love to have her at the summit. That would be an amazing fireside chat. So if anybody knows her, invite her to the summit. There really is three-way... Oh, and by the way, Chamath, media business on the East Coast also does this, and that's why...
friend of the pod, Tucker Carlson and Don Lemon, both of them started shows on the internet, not other networks, because they have very tight non-competes and it's pay to play. They got paid out their full contracts according to reports. And so there is something, I think, in the middle ground here, Chamath, where if you do want somebody to not compete, you got to pay to play.
In other words, they're getting paid. But this is what I'm saying.
It just tends to be examples, Jason, over and over of markets that don't really matter that much relying on these things and markets that are dynamic and add value in humanity, just moving past them. So it probably makes sense for all of us to just go move past them.
100%.