The Transformer Shortage, Uber's Autonomous Future, More Size Gongs, FAL Series B

29:55Just due to that sort of efficiency increase. So we're just seeing more and more of these stories. I've never experienced a time where we had this many new companies ripping from zero to double digit, you know, millions of dollars in ARR. And so it's, you know, just fantastic to see. Yeah. Yeah. And the products are clearly real. Obviously there was, you know, a similar boom during the crypto boom in 2020, 2021. But there was always a question of like, is this really a durable problem? Or are we just speculating with this? Sure, the stocks might be a little bit frothy, who knows where the valuations land.

30:34But you know, when I opened one of these apps, and I use generative AI to edit a photo, I'm getting real value out of that. And I love that. And I think that's going to stick around for sure. Anyway, let's move on to our next fundraising announcement. Harvey has raised$300 million from Sequoia in a Series D. Winston Weinberg says on X, excited to announce our Series D led by Sequoia with participation from Conviction, Kleiner Perkins, OpenAI, GV Team, Conviction, Allad, Gil, and LexisNexis. LexisNexis is an interesting one because they have all the data that you'd want to train on. Let's go through the announcement here.

31:14Today, we are thrilled to announce our Series D. In 2024, we saw 4X annual recurring revenue growth and expanded from 40 customers to 235 customers in 42 countries, including the majority of the top 10 US law firms. So that means six at least of the top 10. That's pretty huge. And I'm sure they're paying a pretty penny for this. We've also seen the legal and professional services industry shift faster than ever before. Lawyers are adopting technology at an unprecedented rate. Centuries old firms are experimenting with new business models and enterprises are driving significant savings with AI enabled workflows.

31:56The pace of change will only accelerate in 2025. Very cool. Have you seen a lawyer use Harvey or has it been revealed to you? What's your take on Harvey? I haven't yet. I, one of my neighbors and our, and our, our, our counsel for TBPN, uh, is at a fairly large law firm, Wilkie. And I started asking him as soon as Harvey was like on the map, you know, already, they quickly came out the gates and raised, you know, a few rounds back to back. They're already at their series D I think it's a two ish year old company, maybe a little bit more. Um, they, so I started asking him, have you heard about this have you seen it and so like nine months ago hadn't hadn't heard about it used it anything like that um i actually should i'm not going to text him this moment because he's on paternity leave but um i'm sure they they're already having conversations with the firm even like six months ago there was people flooding harvey on the timeline oh i don't know a single lawyer that even knows about it or like who's using this blah blah blah hard to fake uh this Hard to have the majority of the 10 largest law firms in the US using your product.

33:07So they're clearly crushing it. I think that some of the best applications of generative AI are the most obvious. You look at the legal industry where every small thing costs hundreds of dollars, even if it's just using boilerplate contracts. And so this was just such an obvious application, but it took a ridiculously talented team and, you know, extreme focus to actually get it to this point. And so I'm sure there will be, you know, what what hasn't been obvious yet is, you know, in the generative sort of engineering coding space, there's so many heavily funded big players, everything from, you know, Cursor to Devin, which are taking different approaches to Windsurf and Poolside and all these different companies.

33:56We haven't seen the same thing play out in legal AI. There certainly are, you know, cool. You know, there are a number of, you know, legal AI companies, but none that seem to have the momentum and the sort of like, you know, potential power law effect of Harvey now has hundreds of millions of dollars to, you know, deploy against this problem space. And again, I don't think this is the kind of thing that law firms are going to want, like, you know, five or six tools. They're also not the kind of thing that engineers can sort of more organically adopt software. I don't think this is going to be the same type of situation just because it's such a privacy, you know, oriented industry and so heavy around regulation.

34:41So I have a theory that I want to run by you and kind of spitball and work through. there's always this question of, okay, yeah, the company's doing well. They got more than half of the top 10 US law firms, but what is so capital intensive about this? And there was a big question about, should they be raising a ton of money and training a new foundation model? Because at a certain point, OpenAI is spending$500 billion on Stargate to train GPT-6 or GPT-5 is coming soon. And can you, if you're in a narrow niche, like a vertical software niche in AI, afford to train a new model? And I don't know that they are.

35:25I don't think that they need to necessarily. I think a lot of what they're doing is post-training, not pre-training, not the really expensive training runs, more of the fine tuning. But then the question becomes, well, these companies are making money. They don't have that many employees. Why are they raising so much money. And I think for these very specific niche AI companies, it could be driven by a need to do deals and essentially pay for access to data that's not widely available on the open internet. And there's actually a case that we can go through, but what's your take on that theory that, hey, maybe you need a lot of money because you need to go pay the license and get the data illegally and bring it into the LLM.

36:10And that's the value add. I think that that's, I'm sure a, you know, percentage of this fundraise will be used to do that. That said, I would guess that it's this, that this was more oriented around Sequoia wanted to own another 10 % of Harvey and, or, you know, there's a mix of investors in the round, but it's more around, Hey, we're in the position where we can do another financing. We're only willing to sell 10%, you know, and we just figure out what the number is. Okay. 300 on 3 billion. Right. And so I don't think at when a company raises this much money so fast, sure. They have to have a plan for it, but, uh, of how to deploy that capital.

36:50But oftentimes it just comes down to, you know, we're willing to sell 10 % of the company right now. And then the market sort of sets the price. Right. Interesting. Interesting. Uh, well, I, well, we should go through this, uh, Thompson Reuters versus Ross intelligence case. Ben Thompson broke this down on Stratechery yesterday, and I thought it was very interesting to understand how data and fair use are being used in these AI trainings. So Reuters, which is funny because they're the ones in the lawsuit, but they're also reporting on it. Another one hand washes the other situation. Reports a federal judge in Delaware on Tuesday said that a former competitor of Thomson Reuters was not permitted by U.S.

37:33copyright law to copy the information and technology company's content to build a competing artificial intelligence-based legal platform. U.S. Circuit Judge Stephanos Bibas' decision against defunct legal research firm Ross Intelligence marks the first U.S. ruling on the closely watched question of fair use in AI copyrighted litigation. And so Ross Intelligence was using Thomson Reuters headnotes as AI data to create a legal research tool to compete with Westlaw. It is undisputed that Ross's AI is not generative, wrote the judge. And so this is a pretty important distinction. Ben Thompson argues that Ross Intelligence was building a legal research tool based on natural language processing, to enable plain language search queries that would deliver relevant court decisions.

38:30This tool was a direct competitor to Westlaw, which is owned by Thomson Reuters. Here's the key aspect of the case, which does have a bearing on gender of AI. The product itself did not violate Thomson Reuters copyright. Rather, the judge ruled that training material that Ross Intelligence used to develop its natural language search functionality did. Specifically, while Thomson Reuters refused to give Ross Intelligence a license to its headnotes, which summarized a case, and key number system, the numerical taxonomy it used to organize cases, Ross Intelligence did use a company called Legal Ease to generate its training data.

39:06And Legal Ease is alleged to have basically copied Westlaw's headnotes. And so the reason why this is pertinent is that assuming this case holds up through a jury trial and any relevant appeals, it establishes the precedent that using copyrighted data for creating a product, even if that data is not in the final product itself, does lead to potential liability. This is also precisely why this case might not hold up. There's two other cases, and eventually it's going to go to the Supreme Court. And so it squares the circle in generative AI cases, but not if the problem was the training data. It seems to me that the entity that violated the law is legalese, not Ross Intelligence, which, by the way, has long since gone out of business because of this lawsuit, which is funded by insurance.

39:51Regardless, I wouldn't read too much into this or any other case just yet. All of this is going to need to be resolved by the Supreme Court at some point or Congress. And so this is interesting because there's been this debate over, oh, OpenAI is scraping YouTube or they're scraping, you know, there's all these allegations, scraping the New York Times. And then there's the question about, is it transformative? And what is fair use? And so, you know, if you're just, if you're just a, you know, a BuzzFeed reporter, even a podcaster like us, we read, I'm reading Ben Thompson's article right now. I'm like stealing his content.

40:24You're generating content from it. Exactly. So I'm reacting to it. And so in theory, this is fair use. Now, the question is, is the output of these LLMs is usually transformative because it is new sentences. And as long as they're not dumping direct quotes out, it should be fine. But there's a question about, can you scrape someone's and can you use their content as training data? And this is something that was just clearly there was no legal basis for it. Because you look at the historical precedent of anybody just physically walking into a library, reading books, pulling out quotes, using that material to make other things.

41:12James Cameron. Yeah, yeah, yeah. James Cameron. And then suddenly, you know, when a computer does it, oh, whoa, whoa, whoa, whoa, whoa. Can't do that. And so we we've taken the other side of this. We've said, you know, I'm sure OpenAI is listening to live stream right now. I'm sure. I'm sure they're just actively feeding this into into their models, but we encourage it, right? We're trying to make the models have our extremist beliefs. And so it's actually Yeah, yeah, yeah. It's part of our strategy. But, you know, it is interesting. You know, the other side of this is that for a long time, companies were able to monetize their data, you know, very effectively on like a per user basis, you know, paywalls, enterprise contracts to access that data.

41:56uh i have a buddy who has like this fantastic media roll-up company and he charges millions of dollars for big hedge funds to get access to that data and other you know policy you know makers and things like that and so it is this really interesting challenge of like what happens the data usa id or what happened i actually pressed i i pressed him on that monday i said you know how you doing bud uh our hands are clean um but uh the new york times can't can't say the same thing but uh politico is the one yeah so anyway it's just this interesting challenge of you know even startups will tell you yeah startups have long included and not usually that compelling aspect of their pitch which is that we're going to have so much data and that is going to be valuable inherently and what's happening now i think is that you need to take the data that you have.

42:50And of course you want to protect it because if it's part of your monetization strategy, but it's coming down to, it does seem like on a long enough time horizon, the walls on all, you know, data that's shared online are going to drop. Right. Because, um, and clearly it depends how easily it can be scraped because some, some data, I mean, if you're talking about like healthcare financial records, like these are truly locked down. Yeah, sure. There's hackers that break stuff open every once in a while. They get a list of passwords or Home Depot receipts or Walmart stuff. But like in general, there have not been massive leaks of really super proprietary data.

43:29And certainly once it leaks, it's on the dark web. Like as a company, you are not just going to be like, yeah, let's ingest that. Whereas, yeah, sure. You might write a bot that scrapes the internet like Google does. And then there's a legal question of, is that okay or not? And how can that go into pre-training? So an interesting scenario that I'm sure Harvey is in is that law firms themselves generate so much data, right? And some of that ends up becoming public through cases and stuff like that. But then a lot of it is highly confidential, highly private. it and if you're a client of one of the companies that harvey works with you don't necessarily want your internal private legal documentation to be used i guess you could argue oh it doesn't affect you if they train on your stuff and allow other people to reproduce documents you know to solve you know similar legal issues but at the same time there's like a bunch of and i'm sure but but that said i'm sure harvey does some deals where it's like you know this is going to cost you five million a year, but like, if you allow us to train on the full set of your law firms, you know, 50 years of, you know, internal communications and, you know, things like that, then we can sort of discount it or, you know, who knows, right?

44:48It could just be built into their practice. I don't have any inside knowledge there. Yeah. Yeah. It's, yeah, it's fascinating to follow. We'll have to dig into Harvey more, understand how they built the business. I mean, the OpenAI startup fund is in this deal. So obviously they're working closely with OpenAI. At the same time, Sequoia is, I think, writing big checks into XAI. And Grok is now going to be able to leave relatively light legal opinions, I think, that that was announced. And so there's always a question about legal aids for consumers versus an enterprise product that integrates with truly proprietary data.

45:28Harvey's clearly taking the latter strategy, but it looks like it's working out. And congrats to the team over at Harvey. Seems like you guys are crushing it. Well, let's move on to more AI news, but this time on the energy side, there's a great article in China Talk all about the transformer shortage. And we're not talking about the transformer architecture. We're talking about literal physical transformers, and the transformer shortage is choking the U.S. supply chain. If Trump wants to build, he needs an industrial policy for transformers. It doesn't matter if you want to build housing, AI data centers, renewable energy installations, EV charging stations, semiconductor fabs, or drone factories.

46:13You need transformers for all of the above. Not even the fossil fuel industry is exempt. Oil and gas drilling both require special transformers to supply power to rig machinery, compressors, refineries, and more. I was talking to, I mean, there are nuclear power companies that want to deliver nuclear power to oil and gas drillers so that they can get more power to run their tools. People don't realize that you need a lot of energy to get the oil out of the ground often. And so these transformers are really key. And I haven't heard of any companies building these, so I thought it would be an interesting deep dive.

46:51So technologically speaking, transformers are relatively simple. They were first invented in the 19th century. And yet the inability to build transformers is causing U.S. industrial policy to short circuit. In the words of the utility resource planner, A.J. Pandy, the transformer shortage is bad. It's like hair on fire, biting my nails, losing sleep levels of bad. If you build new multifamily housing, you could be looking at a two and a half year wait for a transformer to supply the building. And that's if the manufacturers are even accepting orders. I had no idea. The Trump administration has ambitious plans for AI infrastructure, permitting reform, and offshore drilling projects.

47:35Without transformers, however, the U.S. is destined to remain a build-nothing country. no matter how many billions of dollars the federal government dishes out. So how did it get so bad? And that is an interesting question. COVID-19 impacted transformer supply chains, but a confluence of several demand side factors extended the shortage. First, the U.S. grid is aging. We know this. Transformers are typically rated for 40 years of service, and much of the U.S. grid is reaching the end of its allocated life. And so just replacing and upgrading existing transformers is driving a ton of demand. Second, overall electrification is increasing nationally.

48:13Americans are using electricity instead of natural gas to cook their food, heat their homes, take hot showers. And this trend is effectively irreversible. And third, and perhaps most obviously, scores of new renewable energy projects and EV charging stations require transformers to come online. Newly leased offshore drilling projects will have a similar effect. And so we want to do more with transformers and yet uh the supply side is not keeping up and so the n-i-a-c yeah yeah the children the children we need to get children to stop playing minecraft and start mining coal they clearly yearn for the mines let's make them let's make them productive yeah the answer is clearly a steampunk future with uh me on an airship twirling my mustache powered by coal and steam steam revolution that's the end that's the end goal for sure back to blimps yeah blimps for sure yeah i saw a fantastic photo of the hindenburg docked at the chrysler building incredible maybe it was the empire state like it's so ridiculous it just looks so cool and futuristic uh but at the same time i think it's more like i think blimps are more like unicycles than anything else where it's like novel but then there's so many drawbacks that it just doesn't make sense.

49:28So I think I'm anti-blimp, anti-train, and I just want supersonic jets and rockets, really. Anyway, the NIAC identified four structural supply side challenges in a report released in June of last year. Labor shortages. The U.S. does not have an effective pipeline to train and retain the manufacturing talent. We've seen this in so many different industries. There's a historic industry cyclicality. There's been a strong correlation between transformer demand and the housing market. And so when the housing market took off in the early 2000s, transformer production followed. But then when the market crashed, it caused many manufacturers to exit the market.

50:05There's also a lack of standardization. It's difficult to efficiently scale up production with a custom product. And so utilities are very specific about their transformers. There's no standardization really. Every utility has its own specification. If you're a small utility, you're asking for a very, very small batch of this very bespoke thing. And there's also a material shortage. Transformers are made with grain-oriented electrical steel, G-O-E-S. Never heard that term before. Glad to know it now. But there's only one domestic manufacturer who is unable to meet demand. Contributing factors include the destruction of the Azovostal steel plant in Mariupol and sanctions against Russian steel producers, rising demand for non-oriented electrical steel, a key ingredient EVs that comes from the same manufacturing facilities, will also increase supply tension.

50:59The net result is that domestic supply can only meet 20 % of transformer demand and transformer prices have risen 60 to 80 % since 2020. Not good stuff. Should we go into physical attack? This is a great one. so crazy crazy crazy a 2014 analysis by the federal energy regulatory commission the ferc identified 30 critical high voltage substations in the national grid and predicted that losing just nine of these substations as the result of a coordinated attack could cause a nationwide blackout lasting for weeks or even months and so this is your call to action become a prepper get a generator, a gas, a diesel generator, get a battery pack, get a Starlink.

51:46You know, if you have a family, you should be investing in this stuff because the power grid is sadly very, very fragile. That report is 10 years old. But according to an NPR interview with Richard Moroz, former president of the New Jersey Board of Public Utilities, the situation today is not substantially different. Knocking out those high-impact facilities is probably harder than everyone would think, but he did not say that the underlying grid infrastructure had become less centralized. What kind of security measures are in place to protect these high-impact facilities? According to Politico, protections include armed security staff, bullet-resistant fencing, or video monitoring, but these measures may not be enough to protect against a major blackout.

52:28In 2022 alone, there were a total of 1 ,600 security incidents involving the U.S. power grid, including 60 incidents that led to outages. Notable cases include the 2013 Metcalf sniper attack and the 2022 attack on substations in Moore County, North Carolina. In both cases, the perpetrators were never caught. And so basically people just go and buy a sniper rifle or a long range assault rifle. And then they just take pot shots at the substation. And if they hit any critical piece of the, in the chain, one wire, one bolt, anything, it can knock out the entire grid. It's like an extremely like high leverage terrorist activity for a crazy person.

53:16And they could have any sort of motives. They could be left wing, right wing. They can be international. it's like one person a 300 rifle and potentially millions of dollars you know or you know god forbid billions of damage my so you think about the arson that happened in la and the potential for that like there are a lot of people that want to have like a high leverage terroristic impact and um and these electrical substations are sadly uh one of the most high leverage little little anecdote little anecdote my mom was buying a condo and was in escrow and they were waiting on a transformer to be able to power the building and uh it dragged on every single day every single week they'd say like hey we're expecting this like in the next week just dragged on and dragged on and dragged on actually for months and she eventually just failed and bought another place because she was like, I'm not just going to keep waiting around for this, you know, this thing that who knows when it's going to come basically.

54:23So actively, you know, slowing down various aspects of the economy. Yeah. I wonder if I can find this Will Manitis tweet. I know he's talked about it. Uh, transformer.

54:40Yeah, this is it. He, he tells the story of Metcalf here. Uh, we should just read through this because it's so interesting 26k likes it's a banger folks on april 16th 2013 a team of highly skilled gunmen opened fire on the metcalf's power substation in san jose california in just under 10 minutes they disabled 17 transformers and caused 15 million dollars in damages this is the most important terror terror attack you've never heard of quick thread the pg &e metcalf station provides most of Santa Clara Valley with power. Facebook, Stanford, et cetera, are all on this grid. The attackers are still unknown.

55:19They were never caught and the motive is still unknown. This is so crazy. It sounds just like complete conspiracy theory because of course, when you don't have an answer, you're just going to generate theories and who knows, they're going to be hard to verify. A timeline of the attack. But all this is like, this is factual like facts. No one knows how to piece all these facts together. At 12.58 a.m. in the middle of the night, fiber optic lines were cut not far from the U.S. Route 101 just outside of South San Jose. The substation loses internet and phone service. At 1.07 a.m., some customers lost service.

55:58Cables in its vault near the Metcalfe substation were also cut. 1.31 a.m., a surveillance camera pointed along a chain link fenced around the substation recorded a streak of light that investigators from the Santa Clara County Sheriff's Office think was a signal from a waved flashlight. Like, I'm ready to go. It was followed by the muzzle flash of rifles. At 1.37 a.m., PG &E received an alarm from motion sensors at the substation, possibly from bullets grazing the fence. At 1.41 a.m., Santa Clara County's Sheriff Department received a 911 call about gunfire sent by an engineer at a nearby power plant that still had phone service.

56:39At 1.45 a.m., the first bank of transformers riddled with bullet holes and having leaked 52 ,000 U.S. gallons of oil overheated, whereupon PG &E's control center about 90 miles north received an equipment failure alarm. At 1.50, another flashlight signal caught on film marked the end. More than 100 expended 7.62x39mm cases were later found at the site. At 1.51, officers arrived and found everything quiet, unable to get past the locked fence, and seeing nothing suspicious, they left. In the subsequent investigation, it became incredibly clear how professional of an operation this was. Of the 100 shell casings found, all had been wiped clean of fingerprints.

57:24there were also stacks of rocks found all over the site commonly used to gauge firing distance so they went they measured they put down rocks this rock means we're 200 meters away 200 yards away and so you need to adjust your uh your your dope accordingly essentially yeah uh yeah so yeah Just to kind of get into who might potentially be the culprit. If you've ever anybody that listens to the show that that follows Josh Diamond, he's notorious for saying good morning, we are going to win almost every day. And then also, anytime something like this happens, he'll quote, we didn't say how many US or how many foreign sabotage teams are operating on US soil.

58:12And so he was within the 2016 Trump White House, working on national security, has a military background. And so he had actually a lot of exposure. A lot of times these attacks happen and the government figures out what happens, but they don't necessarily deliver that to the press and the media because the government doesn't want the citizenry to feel like they're constantly under attack. or there's sort of these missions being carried out. But something like this, to me, screams foreign sabotage more so than such a professional job. Who else would really have the incentive to do this other than a foreign adversary that just wants to cause chaos?

58:59Yeah. You know, US and test things, right? Yeah, yeah, totally. Yeah, I'm staring at the tinfoil hat. It's just out of reach. But I mean, just thinking about like if me and a bunch of guys wanted to go do something in a foreign country, could we get there on a tourist visa, overstay that visa, start acquiring weapons? There's a million weapons all over America. Go to some gun shows, buy some stuff, figure out some back routes, buy all the gear, plan all this out. Like this does not seem like, you know, like a moon landing level effort. This feels like a couple guys from a pretty sharp team putting some work in, some planning, and then going and executing it with a couple months of prep, and they get a pretty crazy outcome.

59:48There's been a lot of conspiracy theories about whether this was a test of the system or an attack from an adversary or maybe just political extremists who are Americans, and they want to cause chaos for one reason or another. It's hard to say, but the important thing is that. It's a relevant repost from Bill Ackman, who reposted within the last 24 hours from his account that has 1.6 million followers. Wow. If you're not a conspiracy theorist by now, you are basically the R word. And so a little aggressive. Clean it up. You know, remix it to be, say, if you're not a conspiracy theorist. what would your mother say silly if you're not a conspiracy theorist by now you're just silly that hits pretty hard too um so mix it up i have long said that uh conspiracy theories like if it's a conspiracy theory and it comes true it's just history.

1:00:50like 9-11 the the default narrative on 9-11 which is that bin laden determined to attack that was a conspiracy theory let's define it was it a conspiracy absolutely it was a bunch of guys in yeah yeah literally conspiring to attack america and then the second question is was there a theory absolutely because the cia wrote a memo theorizing that bin laden was determined to attack. And so the CIA basically said, hey, we are in the business of crafting theories about conspiracies. We think we are onto a conspiracy. And then it came true. And so now it's just, and now it's just true. It's history.

1:01:30Ignore the theory. Ignore the theory aspect. The definition of a conspiracy is a secret plan by a group to do something unlawful or harmful or the action of plotting. So anytime, anything, whenever something happens, it's usually because a small one or a small group of people plotted to do something. Now you don't know. The opposite is when you're, when you're building in public or when you're, uh, you know, a politician and you're on the campaign trail and you say, if elected, I'm going to cut taxes or I'm going to raise taxes. And then you do it. That's not a conspiracy. You did it out in the open.

1:02:04You said you were going to do something. Like the Tesla secret plan. That is not a conspiracy. There are conspiracies that happen all over the place. But not everything is conspiracy. But sometimes there are conspiracies that happen. And if someone is onto it ahead of time and has a theory about it, it is technically a conspiracy theory. And so we almost need a different word to denote a conspiracy theory that you think is wrong versus a conspiracy theory that you think is correct. Because obviously, there are two paths that these things can take. We obviously, we keep a tinfoil hat on the set.

1:02:43It's usually within me and you're a good counterbalance to me because, you know, I'll have some crazy theory and be like, oh, like, you know, seems complicated. You know, it could be a lot more simple and, you know, less nefarious than that. Yeah. I usually just like to go back to like, you know, is there some sort of market pressure? Does this conspiracy theory require me to suspend the idea that people are rational economic actors. Okay, that's going to be a lot harder. But when you look at this, it's like, if you're a country that's competing with America and you can send a group of five dudes over to America and just wreak some havoc for, what, 100K in flights and gear, that's going to be a great ROI.

1:03:29And so that doesn't break my world model to consider that this might be foreign interference. A lot of the moms and wellness girlies in L.A. had heard that Iran had posted like a day before the fires, which they always post. They always post. They always post. They're posters over there. And they had posted, you know, in the coming days, the United States will experience a, you know, catastrophe. the world will experience a catastrophe the likes of which it has never seen. And then as it became obvious that there was arson happening and these fires didn't all start organically, then the conspiracy theory windmill started turning.

1:04:15But then since then, we've established that, you know, it was arson, but we still don't really know what happened, who did it, what the motive was. Was it mental illness? Was it homeless people? Was it, you know, they just got away with it? I don't know if we'll ever know. Like this is 12 years ago at this point and there's no evidence. There's nothing to go off of. There's no threads to pull. It's like, you know, they've already, they got the shell casings. There were no fingerprints. What are you going to do? Anyway, there are other problems with the transformer shortage. Mainly one is that a lot of these transformers come from China.

1:04:50$54 billion exactly in 2022. In 2023, China was the USA's third largest supplier of transformer components by value, selling American customers 375 million worth of parts of electrical transformer static converters. Only Mexico and Canada outranked China by value. Chinese-made components represent 15 % of America's imports in this category. This might not sound like much until you remember that this calculation is by value. Since the exchange rate is kept artificially low, China likely provides a disproportionate volume of these components compared to other trade partners. The picture gets worse when you consider imports of whole completed transformers.

1:05:36The OEC reports that China was the USA's number one supplier of transformers and transformer components in 2022, exporting 5.47 billion worth of product to the USA. And so, however, for the large power transformers that are most critical, China is the sixth largest exporter. So it's not all bad news, but it is very rough. And you could see that this could become one of the poker chips on the bargaining table in the tariffs. We're restricting chips, but chips aren't the only thing we need to build these AI data centers. If we want dominance and artificial intelligence, we need transformers too.

1:06:17And if China cuts us off, that's going to be a significant setback. So lots of, you know, we never talk about politics, but geopolitics always finds a way in with these stories. So let's wrap up the transformer discussion and see how they close this out. By supplementing the 2020 Transformer Executive Order with Real Industrial Policy, the Trump administration has a huge opportunity to jumpstart the construction of housing, factories, and energy infrastructure in the United States. I love the sound of that. I think that should be bipartisan. We all want to build more stuff. To close, we'll leave you with a full list of policy recommendations from the National Infrastructure Advisory Council, the NIAC.

1:07:02They recommend that the federal government craft policies and designate funding targeting at increasing domestic capacity. They want to use the CHIPS Act as a model. They recommend convening all parties who drive demand to achieve greater accuracy in transformer demand forecasting. So everyone who's buying transformers, let's get them all in a room and let's say, how many transformers do you actually want? How many do you need? Stargate, how many transformers do you actually need? Crusoe, let's get you in here and see so that we have an accurate plan of what we need to build domestically, how important are the tariffs, et cetera.

1:07:38They recommend encouraging long-term contracts and customer commitments. They recommend establishing strategic reserve of transformers with the U.S. government as the buyer of last resort. and they recommend that the government promote collaboration between design engineers from utilities and domestic manufacturers with the goal of standardizing transformer design and reducing complexity. And so there's a couple others here, but basically you need to grow the pipeline of qualified workers in the transformer manufacturing industry and also ensure sufficient supply of electrical steel by coordinating incentives.

1:08:17So a lot of work to do on industrial policy there, but hopefully green shoots ahead. What you got? I got a good transition to some actually breaking news from the re-industrialized summit. Aaron Slodov and Austin Bishop actually just sent us this while we were live. They just announced re-industrialized 2.0 is returning to Detroit. So if you don't know, re-industrialized is a basically conference to get all the different policymakers, investors, and founders together in a single room to figure out how we're going to re-industrialize the United States, right? This is one of the top sort of national security issues of our time.

1:08:56And it's going to take my hotels calling me to kick me out of the room. Hopefully it's not too loud. But anyways, really exciting event. They did it last year. It was a huge success. They have people like Dan Gilbert, who started Rocket Mortgage and owns the cabs and is part of stock tax and he's going to be hosting it so they're going to you know pull out all the stops it's going to be an amazing event there was a lot of good stories that came out of it last year i was having uh my second child literally the same week that it was happening last year i think you were so we both couldn't go but we'll have to go there uh this year one of the guys who's working on it mike slay i were i went to his previous conference uh the defense tech Venture Summit in DC.

1:09:42That was fantastic. Really well executed. I gave a talk on stage. The mics work great. There were tons of cameras. All my slides were up. They look great. It was a lot of fun and a really great community and really top founders went. And so highly recommend checking out Industrialize, signing up, getting on the list, massive demand. And I'm sure it'll be a huge, huge event. Hopefully we can get out there, but if not, I'm sure we'll be live streaming it and reacting to the show as it goes on. I think that'd be really cool to do. Anything else we should cover on the breaking news or should we move over to Dara Khosrowsharhi over at Uber?

1:10:18Do it. Let's do it. And so Dara did an interview with Ben Thompson over at Stratechery and I thought it was interesting because I wanted to talk about self-driving cars, but he has a very interesting background. uh dara's family left iran after the 1978 iranian revolution and to be clear so did dara from delphi's family did the same thing yeah so a lot of dara's just america and just dominate and so you were mentioning how iran was kind of posturing that they may have started the fires in la and that they're you know kind of a near peer adversary at this point to the united states It's certainly not an ally.

1:11:00And it's interesting because the people, like if a family left, what does that say? If a family left Iran in 1978, like what does that say about their politics? Well, it actually means that they're pro-business, capitalist, pro-democracy. And those are the people that got kicked out in favor of an authoritarian regime. And so I'm a huge fan of all the Iranians who have made it over here. And, you know, if we ever do need a new leader in Iran, Dara Khosrowshahi, I think, would be the one to take it over. Or our friend after Dara, you know, you know, LBOs Google and merges it with Delphi, like a generational run, then go back and run Iran.

1:11:43That's one of the things is same thing with Russia and China and Iran, our adversaries, the actual people in these countries we have so much in common with. We all generally the same things. We want to have families and be a great education and great economies. We all want the same thing. And then it's our regimes that are just like, yeah, we are mortal enemies. And so, yeah, you know, Dara, our Dara has told us, told me stories about his family coming over here with nothing, literally nothing, like basically a plane ticket. And then work going, you know, going from being, you know, successful business people in Iran, coming to America with nothing, running it back up, you know, becoming wildly successful here.

1:12:33And just goes to show the ingenuity and creativity and drive of that people. Yeah, it's such an interesting kind of like A-B test on, you know, is there some sort of raw CEO skill or raw business leader skill? And clearly that's the case with Dara. His family ran a very big company in Iran, and they thought that the children were going to join the family company when the time was right. and then they had the slight interruption of the revolution that caused them to flee Iran. They were lucky to have an uncle that lived in the United States with his wife and they took us into their homes and they rebuilt their lives.

1:13:13And so we used to go to France in the summer. So at the time, no one really knew the changes afoot in Iran. So we thought we would just leave, go to France for a while while things cooled down, calmed down, so to speak, and they never did calm down. And so he goes to the United States and completely starts from scratch with basically nothing. They were lucky to come to the States and rebuild their lives. And so he had an average suburban childhood, studied engineering in college and quickly through that all the way to the start of a career in New York in finance. So he was like an investment banker at Allen & Company.

1:13:47Then he meets Barry Diller, who was a client at the investment bank. And he worked for Barry Diller on a couple of investment banking deals. And he swears to himself, Dara Koswashari says, if I ever have the chance to work with that guy, I will. And he gave me the chance to work for him and I went to become his deal guy, so to speak. And so he's a deal guy. We love deal guys on the show. And so he moves up in his career with Barry. He goes over to IAC Travel, which then goes into its own entity, Expedia. He becomes a public company CEO. It was a trial by fire. And the early years were pretty rough.

1:14:24But after a while, I got the rhythm of it. Barry was the chairman and controlling shareholder. And we built that company in partnership for many, many years. And it was a really good run for me. And so I thought that was interesting. He also talks about how he became CEO of Uber a little bit more. We have some details on this just from the reporting at the time, but I thought it was good to revisit. Ben Thompson says, one big story is that as I understand it, there was a tense boardroom fight to even select you. And DK says, yes, Travis Kalanick supported you, even though you're like, you're not going to be part of this company?

1:14:57Was he just trying to get back at Benchmark? Or did you sit down and talk to him and sell him on your vision? And so there's this question. Obviously, Travis was in a rough scenario. He didn't really want to leave. Why is he backing the new CEO? You would think that there'd be crazy tension there. But DK Dara says, I still to this day don't exactly know what went down at the board. I believe that there were two other candidates. I think it was Meg Whitman and Jeffrey Immelt, at least based on the news. And I think I was the third candidate that was the least objectionable to both sides. And since one side couldn't win over the other, they kind of went with me.

1:15:31Go figure as far as decision making. But it certainly worked out for me. And I think hopefully it'll work out for the company as well. And so he has experience. He actually took over Expedia from a founder. So he was a Wall Street guy. Then he was a CEO guy. And now he's sort of a manager. and I think that was interesting. He says Expedia taught him to be an operator. I went through a period where I was CEO of the holding company, but also running Expedia.com and that taught me operations and my background on Wall Street. That taught me a lot about capital allocation. So I think that training really prepared me to come in onto Uber and be the on the ground technical operator that the company needed, but also understand governance, understand capital allocation, et cetera.

1:16:14And so I have a huge amount of respect for Travis and the founding team because it wasn't just him who built the company. And they had to take a very aggressive stance as it related to regulations, the taxi unions, et cetera. They had to bust through an enormous amount of resistance. I do think at some point the company got so big and powerful, but it was still undertaking, call it upstart tactics. And they weren't able to make that adjustment, which is, hey, we're no longer an upstart. We now have millions of drivers. It comes with a responsibility. We have to have different kinds of discussions with regulators.

1:16:44And I think I brought that maturity in hindsight. So I thought that was interesting. And I want to get to the. Yeah. And the big, the big question always with Uber is what, how big would Uber be with Travis still running the company? And there were some dark years where Dara had this extreme pressure on him. A lot of people didn't believe that he could sort of fill travis's shoes and there will probably forever be a debate still of how big would uber be with with with tk running it uh i joked before that that tk would have loved the uh post-dei era uh he he certainly um he's starting to rear his head he did the all-in podcast he's done little stuff here and there he was spotting it spotted in dc during the inauguration uh yeah maybe he'll do a people maybe he'll do a people magazine interview we can hope but the reason i wanted to talk about this was because i have been kicking around in my head is uber a winner or loser in the ai era and i and i haven't fully come to a conclusion i wanted to hear what dara had to say and hear him kind of go back and forth on it so we'll go to uh their decision to exit self-driving cars and where they go from there.

1:18:02So Ben Thompson says, you're doing partnerships with different autonomous driving companies in different cities. Why don't you give me your high-level pitch on how you see this market? Then there's a bunch of specifics that I want to drill down into. So DK says, yeah, definitely. Listen, I don't think it's an if, it's a when. And the AV is a huge opportunity for the entire marketplace. These robot drivers are going to be safer. I think they have the opportunity at scale. And it'll take a while to get to scale to reduce the price per ride and increase the TAM of the marketplace, trillion plus dollars.

1:18:34We think it's an enormous, enormous long-term opportunity. Now, a couple of things have to happen before that opportunity actually turns into real financial opportunity. And so he talks about developing the advanced technology groups with ATG, which was Uber self-driving group. Is it the 2025 era in San Francisco? Well, it could be. It could be, right? But I do think the market has changed and I certainly hope it has. First, you need a consistently superhuman safety record. It's not good enough for a robot driver to be better than a human driver. It's got to be multiple times better than the human driver.

1:19:14I think the capacity for society to accept human failure is much more than machine failure. I completely agree with that. There was a fatality in Phoenix. Did that change Uber's mind as far as getting out of the autonomous vehicle market, or were you sort of already on that way and that sealed the deal? I think we were leading that way. It certainly was a very, very important factor in our determination, but there were a couple other issues that were more paramount. One was they were in the middle of COVID. So all of a sudden capital was much more dear. We went from losing$2 billion a year to losing$4 billion a year, which is pretty scary.

1:19:47My job is to make sure the company survives. And this was a significant amount of money burning part of the operation. The other issue that was a big one was that even though we were developing ATG to some extent separately, we treated them as arm's length. We did have the marketplace separate from ATG and the other players with whom we wanted to partner, like Waymo didn't believe us. So they couldn't do a deal with Waymo as long as ATG was there because they were like, hey, at some point, you're just going to swap Waymo out with ATG. And so they want to be asset light. Hardware was not Uber's superpower.

1:20:20At some point, companies have to know their superpower. There are very few companies in the world who are great at hardware and software. Tesla, Apple are probably the two of the leads. It's very difficult to pull off the accident, capital, how the business was shaping. You either had to make a bet on a vertical or platform. We wanted to make a bet on platform. And then just superpower were the four reasons why we decided to get out of it. And Ben compared it to Netflix. So on autonomous cars, there's a question here. Where does he say? How does that differ around the world, US federalism? So there's a question about, you're obviously the ones that have variable supply.

1:21:01Speaking of the network possibilities before I'm on board with that, I should have been a devil's better devil's advocate here, but let's grant you this. You are the aggregator sitting at the top of the stack. I'm very curious to think through what do those layers, some of which you just listed look like. So let's take a look at the actual cars. Is the software stack and the hardware made by the same company or a different company? How do you see that playing out? And Dara says, accepting Tesla, it will probably be different. So if I look at the stack, it actually kind of looks like the hotel business.

1:21:31I'll expand on that in a second, but let's look at hotels for a second. Right. You've got the brands and the owners. Well, you've got the demand layer, which can be Expedia. It can be booking.com. It could be Marriott. Then you have the brands themselves, whether it's Marriott or an independent. And then you have the operator. The operator, these are the teams, management teams. Sometimes Marriott is an operator of a hotel. Sometimes there's actually local management companies as well. that's an operator. Then there's the asset owners. These are REITs, real estate investment trusts. Marriott doesn't own almost any hotels whatsoever.

1:22:04These financial partners own the hotels. And then you obviously have the financier. And so he thinks that's going to happen in autonomous vehicles. You have the network layer. That's Uber. That's Lyft. It could be Bolt. It could be Waymo too. They have the wherewithal of also going direct. Then you have the driver. To me, that's kind of like the brand, which is the Marriott driver or the Waymo driver. Then you have the operator, the management company. This could be us or Move.io, which is a partner of ours. Actually, 15 % of our inventory today come from fleet operators that are on the ground in these cities.

1:22:37So we think we can move them over to managing AVs very easily. Ultimately, we think of OEM manufacturers. Most OEM manufacturers are going to have different software providers providing AV, sometimes on an exclusive basis, sometimes a non-exclusive basis, a GM, Tesla are going to develop in-house, but many of the other players are going to license it. And so he's talking about this fragmentation and he's applying the Expedia lens. And it's just interesting to me because we've seen this massive AI boom. And my question has always been, we keep hearing, oh, GPT-5 is going to be superhuman or ASI is right around the corner.

1:23:15And shouldn't those models be able to drive if they're superhuman? Like we keep hearing like they're going to be like better than any PhD and it's like, well, most PhDs can drive a car. And so that should be part of the eval, I would think would be, yeah, it can also, it can write papers, but the same model should be multimodal enough to drive a car. And so there's a question of like, who wins that? And then how much staying power is there in this aggregation layer? Certainly Uber has a lot of people that have installed the app. They rely on it. There's a whole system, payments, ratings, all this different stuff.

1:23:51But is Uber vulnerable or are they actually stronger than ever? What do you think? We talked about this before. It seemed like hearing this interview has completely changed my opinion on, so it's smart for Dara to go and talk with somebody like Ben Thompson because it'll actually be a very intellectual conversation more so than a journalist that's just trying to get a headline or potentially dunk. So I think that his point of view of comparing it to the hotel travel industry, which he obviously has a ton of experience at from his time at Expedia, that starts to make sense. I always had this lens on Uber as, okay, Waymo's is dominant force.

1:24:36They're experiencing, you know, they have a technological lead. They have the actual traction lead. People love the product. They have their own app for bookings. And so, you know, we had talked about this on the show before. What happens to Uber? Are they a cable company that just has a lot of existing, you know, consumer demand and, you know, comfort and usage? And they just sort of over time, like people just still prefer human drivers for certain things. And, you know, maybe they just sort of end up taking a huge discount on their on their multiple because, hey, this business is going to be dead in, you know, 30 years completely.

1:25:11And it's not really going to be growing. Maybe you grow earnings, but not revenue. You know, who knows? But now seeing this, I can imagine a scenario where there's, you know, different fleets and operators and technology providers. And Uber maintains that sort of, you know, routing demand from consumers to to these fleets and sort of being that marketplace layer. So I can totally see a world where where they thrive around this new trend. Right. They're stripping out potentially. There's a bunch of costs being stripped out when I ride. I'm about to get an Uber over to the PMF or die cage and it's going to cost me like sixty dollars.

1:25:50the majority of that is going to go towards the driver who just pulling up. I open the door, I get in, you know, say a few words and he drops me off in 10 minutes. Right. I can see Uber really benefiting overall. And I think this is the right narrative for them to take when they don't have this sort of internal technological lead or, or even on the ground traction that a way mode does. Yeah, I do. I do wonder if maybe Uber is particularly well positioned for this like slower transition towards AV fleet rollouts, because even if the next training run of the AV AI system is just flawless and superhuman in every way, it's still going to take a long time to just build that many Waymos like Jaguars going out of business.

1:26:38like where are they getting these cars from they need to make so many of them and then even once they do that it's hard to maintain that fleet that can flex up or down for hey it's new year's midnight everyone wants to be in an uber or it's monday morning and everyone's commuting are people commuting in in waymos now or ubers because we're in this like post car ownership society that's possible. And Uber has this ability to, with the demand-based pricing, the surge-based pricing, they can say, hey, it's New Year's Eve. Surge pricing is going to be crazy. $200 to just hop in your car and go pick up someone.

1:27:19Even if there's a lot of AVs driving around, 100 % of the AVs could be in use. And then you're going to want human flex up to actually meet the full demand. uh and these are going to be people that just commute and they're just going to say hey i i want to make some extra money and the prices are so good i'll jump into the market right now and have a little bit more flexibility there so i i could imagine that um that there won't just be oh yeah one month like all the cars are self-driving immediately just because of how long it takes the counterpoint is that waymo is now venture backed they did around where they raised from you know it's google is majority owner but they did a venture round and so that tells me one i was surprised that google didn't just try to own the entire company but yeah that tells me that their ambitions are to be the next uber so they may play ball with uber at some point maybe there's partnerships that they can do but they're still competing for the same pie john i'm gonna get kicked out of my hotel room um okay and uh uh so we gotta wrap up the show then well i was going to say, if you want to run through a bit of timeline to give the people what they want, you're welcome to, or we could just pick it up.

1:28:32Fortunately, I'm going to the airport this afternoon. I'm flying home. We're going to be back in the studio. Back in the studio. Yeah, I think that's a good place to stop it. We're done with the Dara story. Tomorrow, we'll take you through Elizabeth Holmes' prison interview. We'll take you through a ton of timeline. We're going to have, uh, you know, a million posts built up. We got a whole stack here, folks. So, uh, set your alarm clocks for some time between 8am and 10, because we're not really on a full schedule yet, but we appreciate you dealing with all the kinks of live streaming. We're still figuring this out.

1:29:09I think we've had a fantastic couple of weeks with it and we're really enjoying it. We appreciate you listening to the show wherever you listen. So don't forget to leave us five stars on Apple Podcasts and Spotify. Leave an ad in your review. Send us DMs. Tag us in posts you want us to react to. We will get it on the show. And next week, we'll have some updates on PMF or Die. And I got a quick line before they put me in cuffs and walk me out. But sign up for Ramp. Buy, move your portfolio over to public. Get an eight sleep. Run an out-of-home campaign with adquick and get a bezel watch and then take it to your wander.

1:29:51I think that's great. That's just a simple recipe for a good Thursday. There you go. All in a day's work. It's barely noon on the East coast. It's still 10 a.m. on the West coast. You got plenty of time to get all that done today. So go do it. Thank you, brothers. We will see you tomorrow. We'll see you tomorrow. Talk soon. Talk soon. Bye.