DeepSeek, Open Source, Tariffs, DOGE, Market Impact | BG2 w/ Bill Gurley & Brad Gerstner

00:00:00.000 | And I think you make a good point and a good defense for globalism,

00:00:02.420 | but I think the exact response would be if you're the president of the United

00:00:05.420 | States, you know, you're, you're not, you're not looking out for, you know,

00:00:08.820 | the standard of living for all humans or for all people around the globe,

00:00:11.980 | you're looking out for the standard of living of people in the United States,

00:00:14.820 | you know, who, who you have a constitutional oath to.

00:00:17.580 | But my point is that even with that lens,

00:00:22.780 | you have to think about the dynamics around the globe and pulling up the wall

00:00:27.580 | and trying to get people to make a $40 microwave in America is going to

00:00:32.340 | fail.

00:00:33.180 | Hey Bill. It's great to see you. Good to see you, sir.

00:00:48.020 | Are you going to apply to run the sovereign wealth fund or actually better yet?

00:00:51.340 | Can I,

00:00:52.020 | will you give me permission to nominate you to run the sovereign wealth fund?

00:00:55.860 | I, you know, I would think there are people that, uh,

00:00:59.180 | that I've met through my days that are LPs at rather large funds that have way

00:01:03.740 | better experience for something like that.

00:01:05.620 | I think you would be a hell of a choice. Um, certainly on,

00:01:09.020 | maybe on the board of the sovereign wealth fund,

00:01:11.260 | but it's going to be a fascinating experiment.

00:01:13.580 | One observation I had is as we're getting ready for the pod today is last night

00:01:18.100 | I was helping, uh, Lincoln study for his AP history exam. And, you know,

00:01:22.340 | he's studying the gilded age and more importantly,

00:01:24.520 | kind of the McKinley presidency. And,

00:01:27.020 | and it's all about the McKinley tariffs. And I said to him,

00:01:30.780 | it's pretty amazing that you and I are working on the same thing. You know,

00:01:34.980 | I'm reading on chat GPT about the McKinley tariffs. And he said, yeah,

00:01:38.700 | but dad, don't worry about it. You don't have to take the test. And I said,

00:01:42.500 | worse yet, I have to figure out how much money to be exposed to the market.

00:01:46.780 | What are, what are risks going to be in the hedge fund? I said, don't worry.

00:01:49.540 | I get it. I get it. I get a scorecard on that too. Um,

00:01:53.100 | a little weightier than the test.

00:01:54.560 | No doubt. I mean, you and I could go full Lex Friedman today, by the way,

00:01:58.880 | his five hours show pod recently with Dylan, those guys was great,

00:02:02.920 | but we're going to try to keep this, you know,

00:02:04.880 | pretty tight and jam quickly on deep seek on tariffs on doge on, you know,

00:02:09.640 | maybe what the market reaction might be to all of these things.

00:02:12.760 | And as I was thinking about this setup bill, you know,

00:02:15.200 | I thought it would be helpful. You know,

00:02:17.600 | you and I got into this and we said we wanted the pod to be about this

00:02:20.400 | intersection of tech and markets investing in capitalism.

00:02:23.640 | And we specifically kind of wanted to stay away from Washington and, and,

00:02:27.900 | and politics,

00:02:28.820 | but it's really impossible at the moment to talk about capitalism and markets

00:02:32.820 | and what's happening without talking about the big things occurring in

00:02:36.380 | Washington. So we're going to do that,

00:02:38.260 | but we're really going to try to stick to the economic and market lens of the

00:02:41.940 | events that are happening rather than the political lens.

00:02:44.300 | There are incredible pods you can listen to that give you kind of the political

00:02:47.560 | analysis of all of this. So maybe we just dive in.

00:02:50.820 | Yeah, that'd be great. And listen, I mean,

00:02:53.220 | when you expand the lens to include that and you look at the

00:02:57.820 | ridiculous pace of,

00:03:01.940 | of innovation in the AI space,

00:03:04.300 | I don't ever recall a single time in my career where it

00:03:09.100 | feels like you could just have your ear to the ground 24 seven,

00:03:13.140 | and you're picking up something new constantly.

00:03:16.140 | No doubt about it. No doubt about it. Speaking of, you know,

00:03:20.200 | we covered DeepSeek last week, Bill, it seems like it was just yesterday,

00:03:24.920 | but a lot's happened since then. For one,

00:03:27.640 | these usage charts for DeepSeek are really quite amazing.

00:03:31.160 | And Driessen tweeted this, you know,

00:03:33.000 | which shows the percentage of DAUs relative to chat GPT,

00:03:37.600 | you know, their top geos,

00:03:39.320 | I think they have something like 15% in India and like 10% in China,

00:03:43.640 | 8% in Indonesia. It's really a global phenomenon. You know,

00:03:47.060 | one of the things that struck me after all of this is you tweeted,

00:03:51.020 | it's a better world if the most disruptive LLM model

00:03:55.780 | is foreign and open source versus domestic and proprietary, right?

00:04:00.680 | As you said, it's better for safety, for security, for free speech, et cetera.

00:04:04.060 | So talk us through where you think we stand today

00:04:08.420 | based on a little space now reflecting on R1, you know,

00:04:12.620 | and we know that you and Benchmark have been like the staunchest proponents of

00:04:17.540 | open source over the course of the last few decades. Help us,

00:04:22.140 | help us understand how you think open source versus closed source is going to

00:04:25.620 | evolve.

00:04:26.280 | Yeah. And so, so, so let me give you some reflections now,

00:04:31.280 | having, having this been a week in the,

00:04:34.020 | in the rear view mirror and reading and watching as much as I can of other

00:04:38.060 | people talking. So here, here's some things I think we know about DeepSeek.

00:04:43.020 | And R1, the, it was quite innovative, you know, 0.1.

00:04:47.180 | And if you have five hours to listen to Dylan and Nathan Lambert on Lex

00:04:52.300 | Friedman, they get into some of this. They, um,

00:04:55.220 | DeepSeek could put more experts simultaneously against a problem.

00:04:59.740 | They were able to do that because they figured out a way to, um,

00:05:03.500 | separate the parameters and work on things with smaller parameter counts faster,

00:05:08.540 | which, um, no one else had done before.

00:05:12.060 | And so you end up with something that's cheaper and faster. And,

00:05:15.620 | and it's just important to recognize that they did innovate, right?

00:05:19.140 | Versus just copy. Cause I think there's a lot of noise around this whole thing.

00:05:23.780 | Um, three, they did choose to be the most open model we know of today.

00:05:29.100 | And there's a lot of people that like to get into nuanced conversations about

00:05:33.220 | whether it's truly open source.

00:05:34.860 | I guess people are hoping for one day when someone shows all the data and all

00:05:40.180 | the training processes. And so they may have been short of that,

00:05:43.100 | but with the MIT license,

00:05:45.140 | which has no restriction whatsoever on how you do this and open weights.

00:05:49.700 | Um, it gives a lot of freedom to a lot of people to take this thing and run in

00:05:53.820 | opposite directions, which I'll come back to.

00:05:55.780 | The thing you mentioned is also I think worth noting the success was a

00:06:01.340 | breakthrough and maybe not compared to open AI, you know,

00:06:05.500 | but certainly I think Mistral or Anthropic or any of these players would have

00:06:10.500 | loved to have had the, the launch moment,

00:06:15.060 | if you will, the deep sea cat and,

00:06:18.060 | and weren't able to achieve it for whatever reasons.

00:06:20.340 | I don't know if we can fully explain why, you know,

00:06:23.740 | this app is still number one in the app store today. Um, I, I, I,

00:06:28.900 | the thing you mentioned that I think is super interesting is rest of world,

00:06:32.020 | you know, and, and we're going to go into, you know,

00:06:35.060 | how this might play into China, U S you know,

00:06:37.780 | sanctions and restrictions and rest of the world could be up for grabs.

00:06:42.380 | A couple of other points, I think worth mentioning one, it's,

00:06:46.460 | it's validated now that they went around CUDA.

00:06:48.980 | And I just think that's interesting. Like,

00:06:51.380 | it's interesting to think about why it's interesting to think about performance

00:06:55.300 | optimization and how you might get down to the bare metal. You know,

00:06:59.420 | when we went from on-prem software to SAS places like

00:07:03.980 | AWS, they started pulling out as many layers as they could to get to optimization.

00:07:08.940 | And so you ended up with very special versions of Linux and whatnot,

00:07:13.660 | where things are just ripped out and ripped out and ripped out.

00:07:16.060 | So I think that's worth noting. And, and I've,

00:07:18.940 | I know of at least another example where someone that's working on inference

00:07:22.820 | optimization went underneath as well. And then,

00:07:25.740 | and then the last point I would make that I think is just worth understanding

00:07:30.140 | you and I stumbled upon, I think with the help of Sonny,

00:07:33.340 | this interview in Chinese with the founder and we translated and read it.

00:07:38.340 | And I think it's impossible to read that and,

00:07:41.780 | and say this isn't an exceptional founder.

00:07:44.500 | Like he's just intelligent, independently minded.

00:07:49.500 | I think it'd be very hard to,

00:07:52.060 | to make an argument that he's not a remarkable founder.

00:07:55.060 | And of course, of course you're referring to Ling Wen-Fen, you know,

00:08:00.260 | the founder of DeepSea.

00:08:01.100 | And we'll put the link to that interview and you can,

00:08:03.460 | you can copy it and put it into chat GPT or whatever and get a quick

00:08:08.140 | translation.

00:08:08.980 | Well, and Bill, as you know, you know, our,

00:08:11.060 | our team knows him quite well and knows a lot of members of the DeepSea team,

00:08:15.660 | but you know, he's become really a national hero in China. And, you know,

00:08:19.220 | I think it's a, you know, there's a little bit more of a,

00:08:21.500 | of a telescope on him today, but we did,

00:08:23.860 | we did learn some things when we talked to the DeepSea team and a few of those,

00:08:28.420 | I think are pretty salient. And, and I, I would share again,

00:08:31.860 | just setting the table for kind of the facts that I think a lot of them are very

00:08:35.620 | confirmatory of what, what you just had to say. Number one,

00:08:38.620 | he's an incredible founder. Like there is no doubt about that.

00:08:41.300 | He's been working on this problem for upwards of, of a decade,

00:08:45.340 | been thinking about it. He's very successful and,

00:08:48.540 | and has made a lot of money, you know, in the hedge fund business.

00:08:51.540 | And so, you know,

00:08:53.100 | it reminds me a little bit of the Jim Simon story at Renaissance, right?

00:08:56.380 | These are brilliant folks who happened to apply this,

00:08:59.900 | this early AI and deep learning edge to you know, to,

00:09:03.980 | to the hedge fund business and quant trading. But, you know,

00:09:07.660 | a couple of the key things that were debated last week,

00:09:09.860 | one was the total compute CapEx. And when you take power into account,

00:09:14.060 | it really does, it does get you closer to this billion dollars of TCO,

00:09:18.740 | which was out there, you know, discussed, which is similar, I think,

00:09:23.140 | to the TCO of, of the comparative models. Now, of course,

00:09:27.060 | they talked about the $6 million final training run. And this,

00:09:31.100 | it's important to understand this is also correct, right?

00:09:34.900 | And as I said on CNBC,

00:09:36.980 | this compares to about 10 or 15 million for oh one out of open AI.

00:09:41.620 | So apples to apples,

00:09:43.420 | they were 30 to 50% more efficient to your earlier point about real

00:09:48.260 | algorithmic breakthroughs. And now of course they were doing this a few months

00:09:52.180 | later maybe, maybe six to eight months later than what was going on at open AI.

00:09:56.860 | So to expect some of those savings, but take nothing away from them.

00:10:00.180 | A lot of them were, came from algorithmic improvements,

00:10:03.060 | many of which I think are going to be copied, but,

00:10:05.060 | but breakthroughs nonetheless.

00:10:06.540 | I would add one thing to that, which is just worth paying attention to.

00:10:10.860 | And Dylan and Nathan went into this on the LexPod, but, but they believe,

00:10:15.820 | so I can't really defend it.

00:10:17.780 | They believe if you look at models that are apples to apples on the API right

00:10:22.780 | now, that the deep seeks pricing about one 20th of open AI.

00:10:27.260 | And so,

00:10:28.420 | and they argue about whether open AI might just have higher margins or,

00:10:32.780 | or whether deep seek might be subsidizing,

00:10:35.100 | but that differential is bigger than the ones you,

00:10:38.540 | than the one you described for training.

00:10:40.500 | Yeah, no, you're, you're referencing what it,

00:10:43.300 | what they're charging the customer for inference. And remember,

00:10:47.620 | you know, like, and it gets to a couple other points I'm about ready to make,

00:10:51.340 | they can run it well below cost,

00:10:54.300 | or they could choose to be running at well below cost in order to,

00:10:57.220 | to have these outcomes. I don't think that's going to last.

00:11:00.020 | And open AI might be charging well, you know, well above cost for that, but,

00:11:04.980 | but let's, let's keep going.

00:11:06.340 | I think Lyft, I think Lyft endured rides well below cost for a decade.

00:11:11.260 | No, you, you can do it for a very long time. So in that regard,

00:11:16.060 | but let's get back to the compute stack,

00:11:17.660 | because this was something we learned from their team that I think is really

00:11:20.540 | important to understand.

00:11:21.460 | Their compute stack was smaller than the compute stack that open AI used to

00:11:27.500 | train a one, but it wasn't that much smaller. Okay.

00:11:31.900 | And so the problem they now face is that this changes

00:11:37.140 | dramatically. Remember, these are log linear scaling functions.

00:11:41.700 | So in order to get to kind of the Oh three level,

00:11:44.980 | now they got a 10 X the amount of compute, right.

00:11:48.940 | Assuming that they don't come through with some massive

00:11:53.740 | architectural improvements that caused them to be able to do log linear scaling

00:11:57.380 | without more compute. Right. But to get to that next step function,

00:12:01.460 | they acknowledge it's going to be a lot harder. So set another way,

00:12:05.300 | this was the moment in time where their compute comparison to open AI was the

00:12:09.540 | closest it's ever going to be for a few reasons. Number ones with,

00:12:13.580 | with the export controls, they acknowledge they,

00:12:16.580 | they're going to have a very hard time keeping up with Oh three and Stargate's

00:12:20.860 | going to be even much more challenging.

00:12:22.660 | They're not going to have access to Blackwell that is, you know,

00:12:25.700 | two to three X improvement on top of the Hopper series, which already exists.

00:12:30.100 | This differential in GPUs is, is,

00:12:33.980 | as they begin to train Oh three is far greater than it was for Oh one.

00:12:38.340 | On top of that, we learned, you know,

00:12:40.780 | they are massively compute constrained right now.

00:12:43.860 | So you've seen some tweets about this people that are, you know,

00:12:47.460 | getting server delay and all this stuff with,

00:12:49.420 | with these guys because of their massive success.

00:12:51.460 | So they have a limited cluster to begin with and their current deploying,

00:12:55.140 | they're currently taking most of that compute and deploying it against

00:12:58.500 | inference, right.

00:12:59.620 | Just to support the demand that they have coming in the door,

00:13:03.420 | which further constrains their ability to do training.

00:13:06.500 | And we know they got a 10 X the training to get to that next, you know,

00:13:09.620 | to that next function. So this is a really tough situation.

00:13:14.580 | I think, you know, we'll get into this, the export controls,

00:13:18.420 | remember on Oh one,

00:13:20.260 | they already had somewhere in the order of 30 to 50,000 GPUs that they had

00:13:25.260 | previously purchased. And there's a lot of debate is exactly how many they had,

00:13:29.740 | but the differential wasn't that great.

00:13:31.660 | But now when you have to step up to 10 X that it becomes very challenging.

00:13:35.900 | So that we learned, those are things that we think we've confirmed, you know,

00:13:39.700 | directly, directly from the company.

00:13:42.260 | And so I would expect one of the things that would impress me even more, Bill,

00:13:46.540 | you know, I asked the team at OpenAI, like,

00:13:49.740 | what did this surprise them to deep seek surprise them? And they said,

00:13:52.940 | the only thing that surprised them is that it was a Chinese company that was able

00:13:56.260 | to get there before Meta and others. Right.

00:13:59.060 | And I think it's a really important question.

00:14:01.220 | The thing that would impress me even more,

00:14:03.660 | if somehow they figure out an architectural or algorithmic way to catch up with

00:14:08.660 | Oh three and deep research and the stuff that's now, you know,

00:14:12.300 | really truly frontier without having access to GPUs,

00:14:17.340 | that will to me be the definitive, you know,

00:14:20.500 | kind of the statement that they have somehow breaking the broken,

00:14:24.540 | the paradigm on, on cost and, you know, scaling.

00:14:28.060 | Yeah. And we,

00:14:28.980 | we wouldn't be the first person to say it because even Farid Zakaria said it on,

00:14:33.860 | on his show on national television, but you know,

00:14:37.260 | everyone's talking about the fact that constraints can lead to innovation.

00:14:40.900 | And the reason Lama probably didn't do it is they had access to scaling.

00:14:45.900 | And so only if you're limited on that function,

00:14:49.220 | might you making an algorithmic change.

00:14:51.020 | My guess is we could have one or something that Nathan Lambert might disagree

00:14:56.140 | with you about the 10 X requirement.

00:14:58.740 | But I will say this because it's open because they published the paper,

00:15:03.780 | I'm a hundred percent certain that people at Anthropic and at OpenAI are

00:15:09.340 | studying what DeepSea did.

00:15:12.540 | So if there was some innovative breakthrough, it's going to,

00:15:16.460 | the borrowing is going to be bi-directional and it's going to go right back.

00:15:20.300 | And then that gets you into Jevin's paradox,

00:15:22.700 | which everyone else is also talking about,

00:15:24.780 | which is if we make this stuff cheaper, isn't people are just going to buy more

00:15:28.380 | and more,

00:15:28.940 | but let me go back to your original question real quick on open source.

00:15:32.220 | Cause I do think there, I never really got around to answering it.

00:15:35.780 | And I, there are some things that I think are worth mentioning.

00:15:38.580 | So as you said, I tweeted, you know,

00:15:40.860 | I think a lot of people were worried about these models and whether they're

00:15:44.860 | control people, what people say and disinformation and what's embedded in them.

00:15:48.660 | And what do they do? You know, I'm a big believer and many,

00:15:52.700 | many people in academia are as well,

00:15:54.620 | that more transparency leads to more understanding,

00:15:59.620 | more safety, more security, more free speech, this kind of thing. So,

00:16:04.100 | and I'm not the only one that said this.

00:16:05.860 | If you had a singular model proprietary from a singular company,

00:16:10.140 | it would on all those fronts would give you more risk,

00:16:14.180 | more ability for someone to control that kind of thing.

00:16:16.460 | And let's be very clear, the very reason OpenAI exists, right?

00:16:20.580 | I remember when Elon first talked about it on stage at launch was to defend

00:16:25.380 | against singular control by Google of a closed tyrannical

00:16:30.660 | AI.

00:16:31.420 | So, so I put those all in one group and then I, then I said, look,

00:16:35.460 | it's also better for innovation, you know, startups,

00:16:39.020 | cost performance and global prosperity.

00:16:42.940 | And I'll give you a data point.

00:16:45.460 | I was talking with Clem over at Hugging Face.

00:16:48.780 | And so like 48 hours after R1 was posted,

00:16:53.460 | they had 500 variants on Hugging Face.

00:16:56.580 | And today I pinged them this morning before we started, they're up to 1300.

00:17:00.780 | And so these are just forks in different directions, right?

00:17:05.140 | And it allows people to do massive optimization.

00:17:08.940 | It allows people to solve problems.

00:17:10.940 | Any concern you may have about R1 someone can go work on, you know,

00:17:15.940 | it Linus's law from the original Bazaar and Cathedral

00:17:20.860 | paper was given enough eyes, all bugs are shallow.

00:17:24.180 | And this thing like the R&D force becomes the world,

00:17:28.900 | not just an individual player.

00:17:31.260 | And so it allows for so much optimization.

00:17:34.220 | It also makes enterprise companies happy.

00:17:37.660 | So Aaron Levy and Mark Benioff were out there very boldly

00:17:42.580 | supporting this R1 breakthrough, and it just makes sense, right?

00:17:46.660 | If there's a piece of technology that's a commodity

00:17:49.540 | that they need to be successful, they're better off

00:17:52.380 | than if there's some proprietary piece they have to license from someone.

00:17:56.580 | And it was quickly deployed in all the clouds and, you know, in AWS

00:18:00.900 | and in Azure. It's shocking to me, especially,

00:18:04.380 | especially the Microsoft deployment.

00:18:06.580 | So obviously that feels like a piece of the strategic

00:18:11.140 | back and forth between OpenAI and Microsoft around this contract.

00:18:14.740 | But yet everybody went up fast and it just shows you what's possible.

00:18:19.660 | And I think the amount of innovation that you can have and not just up

00:18:23.980 | the stack, I think, you know, people say, oh, that's going to allow people

00:18:27.060 | to build special models and all this stuff, but down the stack as well.

00:18:30.900 | If you, you know, are trying to compete with NVIDIA,

00:18:34.620 | you know, with with a TPU or a non GPU.

00:18:39.140 | We've talked about all these companies before.

00:18:41.260 | Or, you know, we're an investor in a company called Fireworks.

00:18:44.740 | It's trying to be this optimization middle layer,

00:18:47.300 | like high performance inference.

00:18:50.220 | And they have a lot of really amazing

00:18:52.540 | customers doing runtime inference right now.

00:18:56.940 | Production, knowing more about the model allows them to optimize even more.

00:19:02.300 | And so I think the rest of everybody else,

00:19:06.940 | other than the big proprietary models, are probably thrilled

00:19:10.740 | to have this type of of product out there.

00:19:15.020 | And once again, the variants will just will go and everyone will borrow from it.

00:19:20.100 | And hey, Bill, can I just hit pause for a second? Sure.

00:19:23.420 | You know, a year ago, there was a lot of excitement about Llama, OK?

00:19:29.220 | And Zach really took the leadership on kind of open source in the U.S.

00:19:34.780 | And, you know, here we are evangelizing, you're evangelizing about open source.

00:19:40.100 | But it's not about Llama, right?

00:19:42.100 | It's about it's about DeepSeek and R1.

00:19:44.380 | What do you think happened there?

00:19:46.180 | Do you just think it's scarcity and the leapfrog?

00:19:49.140 | You know, obviously, there are lots of reports that that there was a lot of

00:19:53.300 | trauma within within the Metaplex last week.

00:19:58.140 | People very upset about, you know, the fact that they were leapfrog here

00:20:03.100 | and the amount of money they're spending and they didn't get there first.

00:20:05.740 | But any speculation by you as to why?

00:20:10.220 | Why not them?

00:20:11.740 | I watched different open source battles

00:20:15.660 | and a whole bunch of different verticals since my firm was

00:20:18.980 | an original investor in Red Hat, I think, ninety nine to twenty five years ago.

00:20:24.220 | And there's always a continuum of openness.

00:20:29.220 | And there's a whole bunch of licenses.

00:20:31.700 | I tweeted this list of like a hierarchy or continuum of licenses.

00:20:37.260 | And they're from from most open to least open.

00:20:40.260 | And almost every company that tries to play in

00:20:43.460 | in the open source area is playing this weird game

00:20:47.260 | where they want the proliferation of openness.

00:20:51.020 | But they want some kind of hook to be able to to kind of call back

00:20:55.180 | and have proprietary advantage.

00:20:57.220 | And so, you know, we know that Meta had not gone fully open.

00:21:02.940 | The wait, they had never published the weights.

00:21:05.220 | And there was this clause that says, if you use too much,

00:21:09.220 | you got to come see us again.

00:21:10.580 | And that clause got got spread around.

00:21:14.060 | Yeah. So so so folks like Amazon actually have to pay Meta

00:21:18.060 | for the use of of llamas. Right.

00:21:20.620 | And so they were playing that game, that same game that all these companies

00:21:24.140 | have played, Mongo and Elastic, like they've all they've all had to play this game.

00:21:28.180 | And so, yeah, so it looks like these guys just decided to be more open.

00:21:33.740 | And the MIT license is pretty against the rail.

00:21:37.780 | And and as I said, there are people that say they could be even more open.

00:21:42.340 | But this is the most open for sure today.

00:21:47.220 | So keep keep going on.

00:21:50.140 | Actually, let me make let me let me make two last statements.

00:21:53.380 | And then and then we can shift. So one.

00:21:55.300 | China, I think it's really interesting

00:22:00.180 | to know that China is not a newcomer to open source.

00:22:03.820 | If you look at all of the major projects like Linux or MySQL

00:22:09.820 | and most of these open source projects have a website

00:22:14.660 | and you can see who the leading donors are and just go to the Linux one.

00:22:19.660 | I'll put a link in here from the Linux

00:22:22.500 | member group, and you'll see a ton of Chinese companies.

00:22:25.740 | And someone may say, why is China pro open source?

00:22:29.220 | Well, for the past 30 years,

00:22:32.220 | the West has done nothing but accuse them of being IP thieves.

00:22:37.180 | And so if you believe you have the fastest, cheapest,

00:22:42.860 | most capable entrepreneurs or engineers that can run faster

00:22:47.140 | and work harder than everyone else, you'd rather live in a world

00:22:51.380 | where there's no IP protection than one where you're just being held back.

00:22:56.980 | And so I think they jumped into open source full throttle.

00:23:03.700 | And, you know, it's not just Linux.

00:23:06.100 | It's not just this.

00:23:07.740 | If you look at RISC-V, they're one of the biggest supporters of RISC-V.

00:23:11.420 | And every time we put more constraints on what they can get to,

00:23:15.060 | they invest more in RISC-V.

00:23:16.620 | And so and I think this is particularly important

00:23:20.100 | relative to the rest of the world, as we brought up earlier,

00:23:23.380 | because and we'll get into this, we'll get into sanctions and whatnot.

00:23:27.020 | But if you pull the wall up and we don't support open source

00:23:31.420 | and they do and everybody else kind of likes them leading that way.

00:23:36.140 | Ooh, that could be a dangerous situation.

00:23:38.900 | And then the last point I just want to make, I want to go back to Fareed Zakari.

00:23:43.300 | I'm a big fan of his.

00:23:45.420 | He had two takeaways on DeepSeek, and I was impressed

00:23:49.260 | that he landed on both of these.

00:23:53.020 | But one was that that there was a lot of discussion,

00:23:58.420 | especially in Washington, that the U.S.

00:24:00.860 | was two years ahead of China.

00:24:03.860 | And he he said, look, it looks like after the fact

00:24:08.780 | that that's hubris, right?

00:24:11.540 | If if if that's now six months, three months, whatever, it's closing.

00:24:16.860 | And we need to think, I think, with our eyes wide open

00:24:20.420 | as we make policy decisions.

00:24:22.980 | And and I think that's important.

00:24:24.860 | And then the second one, I was just really impressed

00:24:27.060 | with his understanding of open versus closed and how

00:24:31.380 | you can reach a tipping point where things just move in that direction

00:24:37.100 | because so many different entities get behind it.

00:24:40.820 | I like to see water runs downhill.

00:24:42.980 | OK, so Sam Altman did this AMA last week

00:24:46.700 | and he was asked about DeepSeek and about open source.

00:24:49.060 | And I thought his response was really interesting, Bill.

00:24:52.700 | He said DeepSeek is an impressive model.

00:24:55.620 | You know, and to your point about Jevin's paradox, he's like,

00:24:57.900 | we're going to need a lot more compute, you know, because, you know,

00:25:00.940 | as we've said, demand for this is exploding their compute constraint, etc.

00:25:05.260 | And he said, and on the issue of open source, they said,

00:25:08.540 | would you consider releasing model weights and publishing your open source

00:25:12.580 | research? And Sam said, yes, we are discussing.

00:25:16.100 | I personally think we have been on the wrong side of history here

00:25:21.060 | and we need to figure out a different open source strategy.

00:25:24.780 | So I tweeted in response to, you know, something Mark Andreessen has said

00:25:29.540 | that I that I thought all current closed source model companies.

00:25:33.380 | So let's just say OpenAI and Anthropic would open source their models.

00:25:39.060 | And in the case of OpenAI, I could see them open source, you know,

00:25:42.660 | one which competes head to head with DeepSeek.

00:25:45.220 | At the same time, I could see all companies

00:25:49.180 | that are currently open source and closed, right, which includes DeepSeek.

00:25:53.580 | I could see them in the future.

00:25:55.580 | And Mark Zuckerberg has said he reserves the right

00:25:58.060 | not to release all the models in the future. Right.

00:26:01.140 | So I think we may end up with a world where the true frontier,

00:26:05.700 | the actual underlying model is not released at all.

00:26:08.500 | And the only thing that gets released is the agent. Right.

00:26:11.420 | But then one or two generations behind,

00:26:13.740 | you're going to see them all open sourcing these models.

00:26:16.780 | But let's start with Sam's comments.

00:26:19.020 | Are you encouraged that Sam has said, yes, we need to come up?

00:26:23.220 | You know, we're on the wrong side of history here.

00:26:25.420 | Well, I mean, to a certain extent, it validates what R1 did. Right.

00:26:30.700 | That he would feel the need to say that I have found just

00:26:35.300 | and you talk to him more than me, but I found that whenever a threat

00:26:40.820 | or a challenge is made to open AI, Sam tends to go towards it.

00:26:46.140 | Like that's his kind of go to move.

00:26:49.740 | And I think it works for him.

00:26:51.660 | Like and so I'm not surprised that he said that.

00:26:55.780 | I, I was surprised

00:26:58.860 | on a side note that a couple of our friends

00:27:02.700 | who are co-investors with you and OpenAI, when the R1 thing hit,

00:27:06.460 | kind of very quickly took to X to say that,

00:27:10.980 | you know, something that R1 cheated or that the government needed to come in.

00:27:15.140 | And to me, that was a validation point as well.

00:27:19.660 | Like you wouldn't take the trouble if this thing wasn't real.

00:27:23.060 | But yet it could it could tip us more towards open.

00:27:26.380 | I mean, as a as a as someone who who really enjoyed

00:27:32.180 | my business school classes on finance and economics, you know,

00:27:36.100 | one of the reasons I like open so so much is it's the closest thing

00:27:39.780 | to pure competition.

00:27:41.020 | If you look up an economics book, pure competition is like commodity,

00:27:45.860 | like hard to have, you know, prices leads to, you know, innovation,

00:27:50.900 | low price points, Jevin's paradox blowing up.

00:27:53.380 | And so I'm thrilled that it's tilting that direction.

00:27:57.020 | Now, this may be a perfect time to transition

00:28:00.420 | into the other thing that happened as a result of R1,

00:28:04.100 | which is there are a number of increased

00:28:07.540 | efforts to, I think.

00:28:11.620 | Raise the wall of regulation and sanctions.

00:28:16.140 | And it's funny, I watched a number of people go,

00:28:19.980 | oh, my God, look at R1, you know, Washington must act quickly.

00:28:24.420 | But but but the thing that each person intended

00:28:29.060 | are radically different from one another.

00:28:31.540 | You know, some people think this means, oh, we need to embrace open source

00:28:34.820 | and encourage more open innovation in America.

00:28:37.140 | And the other people think, oh, my God, you know, and Dario put out a long piece,

00:28:42.020 | you know, I guess, consistent with his entire tenure here,

00:28:45.220 | just begging for more lockdown and regulation.

00:28:48.340 | So let's let's go into that, because I do think it's industry interesting.

00:28:52.860 | Right. Sam and the industry response seems to be tipping more in the direction

00:28:56.980 | where we're going to go open to.

00:28:58.900 | And my suspicion is you will see that this year,

00:29:01.700 | you know, out of folks like like OpenAI.

00:29:04.740 | But the discussion about DeepSeek clearly touched a national nerve.

00:29:08.900 | Right. Jensen, you know, got got called to the White House

00:29:12.660 | last Friday or we all see me.

00:29:15.060 | I think they said it was pre-scheduled.

00:29:16.980 | Well, not I'm not sure whether or not that was the case, but let's assume it was.

00:29:20.580 | But it's opened up this broader conversation about whether the U.S.

00:29:23.780 | can or even whether it's wise for the U.S.

00:29:27.820 | to try to stop China from advancing along the AI frontier. Right.

00:29:31.820 | And some of the arguments are human talent in China

00:29:34.660 | will always find a way to innovate.

00:29:36.420 | You know, keeping keeping China six months behind is not worth the cost.

00:29:41.260 | Scarcity fuels innovation.

00:29:43.380 | It turns AI into a global arms racer or the one that I've been

00:29:47.540 | are, you know, been been advocating is I think we've we've focused

00:29:51.580 | so much on slowing China down.

00:29:54.020 | We haven't focused enough on speeding America up. Right.

00:29:57.820 | Removing rate, removing the regulations around power generation,

00:30:01.260 | all the things we need to get America running full speed.

00:30:04.060 | But I would say this is probably the biggest divide in Silicon Valley

00:30:08.140 | among technologists regarding, you know, kind of this president's policies. Right.

00:30:12.860 | So there's a camp, as you know, of China hawks led by, you know, folks like,

00:30:18.620 | you know, Alex Karp was on the Palantir call last night who viewed this

00:30:22.380 | as an existential holy war and that we must battle on every front.

00:30:26.740 | And, you know, in order to slow China down.

00:30:29.580 | And then I would say there are people who are more what I would consider

00:30:32.340 | China constructivist.

00:30:33.740 | And I put you in that camp.

00:30:35.220 | I would I would put myself in that camp.

00:30:37.020 | I would put, frankly, you know, Elon in that in that camp and others

00:30:41.340 | who seem to think that it's a losing battle just to focus on slowing China down.

00:30:46.100 | And what we really need to focus on is more engagement and speeding the US up.

00:30:50.340 | So can you lay out a little bit, you know, your views on on

00:30:53.980 | on those two computing sides and where we may end up coming down on?

00:30:57.500 | Brad, I think you framed it perfectly.

00:30:59.820 | The problem is that the by the way,

00:31:03.820 | I think on the on the anti China side in Silicon Valley, you have

00:31:08.460 | you have like three groups, you have the people like Dario

00:31:13.340 | who are, you know, maybe worried about competition,

00:31:16.340 | maybe worried about more, but certainly question that.

00:31:19.260 | You have the new kind of VC backed defense companies

00:31:24.660 | who all and maybe I kind of put Palantir in that group, but they all,

00:31:29.100 | I think, have an incentive to kind of have tension with China, if you will.

00:31:34.660 | It actually increases revenue.

00:31:36.380 | I call them the new neocons.

00:31:38.380 | And then and then I think you just have a large group of people

00:31:41.940 | who were raised to be anti China.

00:31:45.180 | And they're just it's it's it's what they were taught growing up.

00:31:49.420 | It's the anti communist thing.

00:31:50.940 | It's what got us into the Vietnam War.

00:31:52.540 | It's been around forever.

00:31:53.660 | But but your parents might have taught you that.

00:31:55.900 | Like, it's just in the ethos.

00:31:57.900 | One thing I would add to you on the risk side, you listed them.

00:32:00.420 | You listed a lot of great points.

00:32:02.140 | You know, one thing I would add is that protection of U.S.

00:32:05.260 | companies causes harm, like like

00:32:08.100 | Detroit is not globally competitive anymore.

00:32:12.540 | And putting tariffs on these cars

00:32:16.300 | is not going to make Detroit more competitive.

00:32:18.900 | It's going to make them less competitive and they're going to fall further behind.

00:32:22.420 | And I think this idea of raising the wall

00:32:27.620 | and increased decoupling is a super dangerous idea.

00:32:32.180 | We may find that the rest of the world is perfectly fine

00:32:36.980 | buying ten thousand dollar BYD cars and using DeepSeek.

00:32:41.260 | And we may we may just be shutting ourselves off.

00:32:44.660 | I found this interesting data point I wanted to share with you.

00:32:47.380 | To highlight my fascination with China, I've kind of studied it

00:32:51.980 | over a very long time frame.

00:32:53.620 | It turns out most people have no reason to know this.

00:32:56.220 | But in 1820, China's economy was wide open

00:33:00.700 | and they actually had 33 percent of the global economy.

00:33:05.940 | 33 percent of global GDP was China.

00:33:08.860 | Most people probably wouldn't know that.

00:33:10.620 | And the reason they might not know it is because 150 years later,

00:33:14.580 | at the end of Mao's reign and Mao had raised the wall

00:33:18.700 | and and kind of turned China inwards,

00:33:21.620 | they had fallen to five percent of global GDP.

00:33:25.220 | And they've been working their way back from that.

00:33:26.940 | So we know this emergent China from that place.

00:33:29.500 | But, you know, that's the real risk.

00:33:32.500 | You know, if you're not a globalist, if you don't believe in,

00:33:35.700 | you know, all the great economic work that shows how,

00:33:39.740 | you know, specialization can work to the benefit of everyone

00:33:43.340 | and you close that wall, you may be surprised at what happens.

00:33:46.980 | To wrap this section, and that's the perfect segue to talk about tariffs.

00:33:50.380 | But, you know, I would say this.

00:33:52.860 | I think there is a middle ground, right?

00:33:55.780 | I you know, I took a little when I first read your tweet,

00:33:59.140 | you would rather, you know, it's better for an open source

00:34:01.340 | Chinese model to win versus a foreign.

00:34:05.300 | Oh, foreign. OK, I knew I knew what the hell you meant.

00:34:08.100 | Then a then a closed U.S. model.

00:34:11.460 | I want Team America to win on this.

00:34:13.740 | I want a you know, I would love to see the U.S.

00:34:16.700 | frontier labs open source more stuff.

00:34:19.460 | I agree with you fundamentally on the principles of open source.

00:34:22.500 | I believe they will.

00:34:23.900 | And unquestionably, I want to see the U.S.

00:34:26.940 | win when it comes to the race in AI.

00:34:29.420 | And I know you do, too.

00:34:31.300 | Secondly, I would say on this is that I think a lot of the things

00:34:36.780 | that we've done in the name of being tough on China, right,

00:34:40.580 | are actually counterproductive.

00:34:42.660 | I agree. It takes the eye off the prize.

00:34:44.580 | It slows us down. Right.

00:34:46.660 | And it doesn't focus on speeding us up.

00:34:49.420 | And frankly, it's not very effective or it backfires entirely

00:34:52.900 | in terms of slowing China down.

00:34:54.780 | And and one of the places where I think that this bill,

00:34:58.620 | you know, takes us to is is is really Trump's tariffs

00:35:02.300 | that that that brought, you know, brought brought us to the fore this week.

00:35:06.700 | But you have something to say.

00:35:07.700 | Yeah, I just want to say two things to what you said.

00:35:11.260 | And then we'll go to the tariffs.

00:35:12.340 | So, one, you know, I was being provocative when I said foreign.

00:35:16.340 | And you could read China.

00:35:17.780 | But if you think about it like Linux doesn't have a geography. Right.

00:35:22.580 | And so one possible reality

00:35:26.340 | that I don't think meets your goal of America wins

00:35:29.940 | is you get to a place that's Linux like where

00:35:33.700 | the model doesn't have a sovereignty and and

00:35:37.740 | because of the 1300 variants of our one that are on hugging face already,

00:35:44.100 | like the and because that MIT license, though, it could be that model that wins.

00:35:48.940 | You know, it doesn't have to be the one that they're doing.

00:35:51.780 | It could be a fork of it.

00:35:52.940 | So anyway, I wanted to make that point.

00:35:54.940 | And then the other thing I wanted to say, I'm just on a risk side,

00:35:58.580 | you know, and by the way, there there are proposals in Congress right now.

00:36:03.300 | I just that would kill open source.

00:36:06.020 | They would say we couldn't use variants of our one.

00:36:08.820 | There's there's like a lot.

00:36:10.700 | There's a huge breadth of perspectives, as you already said.

00:36:15.020 | But I think if you I think I can imagine that if you just poke China enough,

00:36:21.580 | if you keep poking and you keep poking and you keep raising the constraint,

00:36:25.980 | you increase the odds that they make a run at Taiwan.

00:36:29.580 | And I just think it's important to always think from their perspective,

00:36:34.500 | you know, and and I just think we need to be careful about how

00:36:39.820 | how hard we push. We may end up with the exact worst outcome.

00:36:42.940 | Unintended consequences. Right.

00:36:44.980 | So this week we woke up, you know, really ended Friday

00:36:49.380 | at a late press conference the president had, and then it went into effect

00:36:52.820 | over the weekend. Right.

00:36:54.220 | Twenty five percent tariffs on on on Mexico and Canada.

00:36:57.420 | Fifteen percent tariff or ten to fifteen percent tariff on China.

00:37:01.220 | Before we dive into the economic debates for and against tariffs,

00:37:05.500 | let's just kind of lay this out.

00:37:07.420 | You know, so Monday morning, the markets overnight Sunday,

00:37:10.060 | the markets are falling a lot.

00:37:11.580 | Monday morning, Kevin Hassett, chairman of the National Economic Council,

00:37:15.300 | comes out on the White House lawn.

00:37:16.660 | He said, oh, these are all being misinterpreted.

00:37:18.700 | This is not about a trade war.

00:37:21.060 | This is a drug war. This is about fentanyl.

00:37:23.300 | You know, he did happen to say we may revisit in the future

00:37:26.380 | as part of a tax reform strategy.

00:37:28.500 | So he left the opening to to, you know, tariffs for other reasons.

00:37:32.380 | Then the president on Monday talks to both sides.

00:37:34.940 | They both commit 10000 troops to the border to fight fentanyl.

00:37:38.060 | And you delay the tariffs for 30 days.

00:37:40.980 | And I think the market reaction now is that they're not going to hit at all.

00:37:45.460 | So let's first start on what we think happens here.

00:37:48.460 | So help me predict what you think happens.

00:37:50.940 | And then I would love to get into kind of the merits and demerits

00:37:54.340 | from an economic and from maybe the tech industry perspective

00:37:58.020 | on on on this tariff strategy.

00:38:01.580 | So, Brad, I'm going to be brief, because, look,

00:38:04.060 | this is more your world than mine.

00:38:05.580 | You're looking at a lot of large public companies

00:38:08.060 | and and all all the things that impact them.

00:38:10.740 | You know, I guess even in the medium sized public companies I work with,

00:38:14.780 | especially if you have physical goods, you've got supply chains

00:38:18.460 | all around the globe. Right.

00:38:19.940 | And I imagine one thing you've had to do in your shop,

00:38:23.780 | you know, is when a new tariff pops up is just immediately ask,

00:38:28.060 | well, who's impacted, who sources there, who like who

00:38:31.300 | who and, you know, probably just creates a lot of chaos. Right.

00:38:35.180 | In the short term, as we try and figure those things out,

00:38:37.980 | you know, there's different Foxconn plants all around the globe. Right.

00:38:42.220 | And different people source different, you know,

00:38:45.420 | whether it's fashion products or anything, you know, from Vietnam

00:38:49.980 | or from Indonesia or from China during covid.

00:38:53.500 | I think one of the things we realized is there is some flexibility.

00:38:57.700 | They can move, move a lot faster than people thought.

00:39:01.300 | But it's still chaotic.

00:39:03.180 | And based on like, as I said, you know, my default is a globalist.

00:39:08.460 | Now, I don't know enough to know if we have unfair deals

00:39:13.180 | that need to be honed and that this is just a means to an end.

00:39:16.700 | And if so, maybe it's not that big a deal.

00:39:19.260 | I don't believe that creating a lot of, you know, bringing the wall up,

00:39:24.500 | as I said earlier, around the eye.

00:39:26.540 | I don't believe that that'll be in the U.S.'s long term best interest.

00:39:29.900 | You make a good point about, you know, clearly

00:39:33.900 | Trump extracted a concession that was a pretty damn good

00:39:36.940 | concession from Canada and Mexico when it when it comes to defending the border.

00:39:41.260 | Right. So as a tactical negotiating tool,

00:39:45.020 | you know, his batting average is exceptionally high.

00:39:48.420 | You know, whether it was getting Columbia to take the detained deportees

00:39:52.380 | where he threatened a tariff.

00:39:53.820 | And so I think that that's the market's reflexive belief.

00:39:56.700 | In fact, Scott Besant, the Treasury secretary, in a letter

00:40:00.260 | he wrote about a year ago to his investors, he used this concept

00:40:04.220 | that Trump's tariffs, you shouldn't be afraid of them

00:40:06.820 | because he said his strategy was to have a fully loaded gun,

00:40:10.620 | but rarely discharge.

00:40:12.780 | Yeah, right. Fully loaded gun, but rarely discharge.

00:40:15.140 | And so the interpretation is he's just using this as a big stick

00:40:20.100 | to achieve very tactical goals.

00:40:21.940 | Now, I call this bill the Besant consensus.

00:40:25.460 | I believe this is the market consensus view, the Besant consensus.

00:40:30.020 | But I want to throw out an alternative view, right?

00:40:33.140 | I think that Trump may, in fact, have a much, much deeper

00:40:37.220 | and more principled belief in tariffs.

00:40:39.540 | If you listen to his speeches, if you read them, and this goes back

00:40:42.660 | over a decade, OK, he believes that McKinley was one of the best presidents.

00:40:48.460 | He thinks that the country, he believes that the country

00:40:51.180 | and there are arguments for this was at its peak or at its best

00:40:54.580 | during peak tariffs in 1880 and that you could, in fact, replace

00:40:59.620 | when we move to replace tariffs.

00:41:02.060 | So prior to 1910, which is when we got the income tax.

00:41:06.300 | Right. The vast majority, we'll put this chart in the pod.

00:41:09.820 | The vast majority of revenue to the U.S.

00:41:12.180 | government came from tariffs.

00:41:14.660 | And then you see, starting in 1910,

00:41:17.260 | basically revenue from tariffs plummeted and the amount of revenue

00:41:21.500 | that came from income tax and Social Security tax,

00:41:24.260 | you know, thereafter skyrocketed.

00:41:26.980 | And so I think there's a belief that replacing tariffs

00:41:30.740 | with high income taxes and corporate taxes has gutted the middle class.

00:41:34.940 | Not only did it destroy jobs in America, but in fact, caused these people

00:41:40.180 | to be burdened with taxes, you know, to pay for social services

00:41:44.820 | that could have otherwise been paid for by tariffs.

00:41:47.580 | So that to me is a variant view.

00:41:51.780 | Right. If you believe that to be true, there is a much more principled

00:41:56.140 | architecture that he wants to move to, that this is not the best consensus,

00:42:00.700 | which is have a fully loaded gun that's rarely discharged.

00:42:04.060 | But it's a fully loaded gun that you fully intended to discharge.

00:42:07.860 | So let's assume for the moment that he does have that view,

00:42:11.420 | maybe more of a fortress America. Right.

00:42:15.500 | I think you've outlined where where you stand on this.

00:42:18.660 | Do you believe that that hurts us technologically and it would hurt us

00:42:22.260 | in in terms of our global economic standing? Is that right?

00:42:25.660 | Well, look, I'm not a tariff expert,

00:42:29.940 | but but when you told me you wanted to talk about this, I did.

00:42:33.380 | I did some research and maybe maybe we could have your son

00:42:36.460 | or your son's professor on to talk more about this.

00:42:40.500 | But all of the even the success stories around tariff,

00:42:45.180 | if you ask your favorite, I didn't tell you.

00:42:47.580 | They seem very short windowed.

00:42:49.500 | Even the McKinley ones were like four years.

00:42:52.500 | So I don't know of a long term successful high tariff program.

00:42:59.420 | And I think it gets back to what I was saying about China, like pulling up

00:43:03.460 | the wall is I don't think I don't think there's any economic argument

00:43:08.700 | that that works to help a country in the long run.

00:43:12.460 | And going back to the to the I do want to take a brief

00:43:17.060 | second to talk about what you just said about the like

00:43:22.220 | the American middle class may have been been affected by this.

00:43:25.900 | You know, there's a time and place when a country is in a great place

00:43:31.580 | to be competitive globally in scaling out production.

00:43:35.980 | And it relates to having a educated workforce

00:43:42.340 | that has a very low standard of living, that's willing to work for a wage

00:43:46.340 | that's highly competitive globally and may be willing to work

00:43:50.660 | nine, nine, six, you know, right, like like way more hours

00:43:55.660 | than nine in the morning to nine at night, six days a week. Right.

00:43:59.020 | And so if you look at when America,

00:44:03.020 | you know, was mostly successful scaling out post World War

00:44:06.580 | two where you're been decimated and and Japan's been decimated.

00:44:11.500 | Yeah. And we had a lot of people moving up the social ladder

00:44:16.220 | and prosperity ladder as a result of being willing to do that.

00:44:20.700 | You know, if you fast forward to where we are today, you know,

00:44:24.780 | I don't think there's any way to say this other than to be blunt.

00:44:27.780 | Like there are people in China, Vietnam, Indonesia, Mexico

00:44:33.980 | that are willing to work harder and longer for a wage

00:44:38.500 | that is radically lower than what people in the U.S.

00:44:41.860 | are willing to work.

00:44:42.900 | And they're going to move from a place on the prosperity

00:44:47.060 | and social ladder that's low to a place that's still beneath the average American.

00:44:52.340 | And I don't know how as a humanist you can say they don't deserve that. Right.

00:44:58.220 | And I think we misinterpret that this is somehow a result of tariffs or trade.

00:45:04.460 | It's just global fairness. Right.

00:45:07.580 | It's just my point of view.

00:45:09.940 | No, totally.

00:45:10.780 | And I think you make a good point, a good defense for globalism.

00:45:13.260 | But I think the exact response would be if you're the president United States,

00:45:16.580 | you know, you're not you're not looking out for, you know,

00:45:19.660 | the standard of living for all humans or for all people around the globe.

00:45:22.780 | You're looking out for the standard of living of people in the United States,

00:45:25.620 | you know, who you have a constitutional oath to.

00:45:28.420 | And no doubt. But my but my point is that even with that lens,

00:45:33.860 | you have to think about the dynamics around the globe and pulling up the wall

00:45:38.100 | and trying to get people to make a 40 dollar microwave.

00:45:41.500 | And America is going to fail.

00:45:43.700 | But you're just going to end up with more expensive products.

00:45:46.860 | We watched this happen in Europe.

00:45:48.780 | Like we watched it play out in Europe.

00:45:50.780 | You're just going to make yourself Europe.

00:45:53.260 | I do think there are some great economic arguments on this.

00:45:57.260 | Like I said, Kevin Hass is chairman of the National Economic Council

00:46:00.500 | is at the Hoover Institute.

00:46:01.940 | You know, he's going to be on the front lines of carrying the tariff policy,

00:46:05.780 | defending the tariff policy.

00:46:07.420 | But if you look at the McKinley tariffs, they certainly caused a lot of strife.

00:46:10.580 | But they're very good arguments

00:46:12.420 | that they helped us industrialize in a way we never would have.

00:46:15.500 | And more importantly, they helped us build critical strength

00:46:18.860 | heading into World War One.

00:46:20.460 | So had we not done the industrialization in 1880, 1890,

00:46:24.180 | would we have been even prepared?

00:46:26.140 | But the world does look very different today, Bill.

00:46:28.660 | Global supply chains, the cost of shipping is radically lower, etc.

00:46:32.780 | But I thought what was interesting is, you know, the Fed actually did a study

00:46:37.020 | on the tariffs in 2018.

00:46:39.780 | They did it, I think it was on washing machines, you know,

00:46:42.620 | and they basically said it led to a lot higher pricing for washing machines.

00:46:46.620 | Right. Where there were, you know, tariffs put on them.

00:46:50.460 | And even for dryers that had no tariff put on them.

00:46:53.580 | But because they're usually sold together, the prices went on those as well.

00:46:57.140 | And interestingly enough, even the domestic producers raised prices

00:47:00.420 | because now the competitive market had raised prices.

00:47:03.100 | So they now had a pricing umbrella that they could raise prices into.

00:47:06.260 | And I think they concluded that very few jobs were actually created.

00:47:09.700 | Now, they were looking at this only two years in arrears, Bill.

00:47:13.260 | So, you know, again, they weren't looking at the long run effects of,

00:47:17.100 | you know, did this play out?

00:47:19.460 | So I think, you know, one area I'm really focused on for Silicon Valley.

00:47:24.860 | So imagine a tariff on, you know, on GPUs or on chips.

00:47:29.260 | Right. Which has been threatened this week.

00:47:31.900 | Like they're talking about just a tariff on Taiwan,

00:47:35.420 | which is a tariff on chips.

00:47:37.380 | Right. And today we actually can't make those chips in America.

00:47:40.700 | Right. We don't have two nanometer, three nanometer fabs

00:47:44.020 | where we could build them even if we wanted to.

00:47:46.260 | So when I look at that, that's really just a tax on chip manufacturers

00:47:50.740 | and on the end buyers. Right.

00:47:52.140 | So that would be a tax on on the Nvidia's and AMD's and and others.

00:47:56.740 | And it would be a tax on Metta and Amazon and all the folks

00:47:59.500 | that would have to have to pay that tax, which which likely means

00:48:03.620 | that you get less chips purchased and less AI research. Right.

00:48:07.940 | So if our number one goal is to win the race in AI, this is a classic case

00:48:11.660 | where in the short run, I think it's self-defeating.

00:48:14.500 | But I do think there are ways to do this where you can bring more fabs to the US.

00:48:19.580 | But if you wanted to have, you know, if you want to have a permanent

00:48:22.780 | 50 percent tariff on all chips in the US starting now,

00:48:26.500 | I think like that's just that's just going to have negative repercussions.

00:48:29.380 | But if you said of a 50 percent tax on chips,

00:48:32.500 | but I'm going to delay it for two and a half years

00:48:34.900 | and you have to meet these hurdles for building fabs in the US, et cetera.

00:48:38.420 | So more of a negotiating tactic than it is kind of a permanent and higher tax.

00:48:42.900 | I think there are some, you know, some some really good outcomes of that.

00:48:46.860 | We're less dependent upon Taiwan, which is, you know, always threatened by a

00:48:51.020 | you know, what what what most people believe is a foreign adversary,

00:48:54.540 | you know, less dependency in the case of of some situation evolving there.

00:48:59.060 | So I think it's going to have to be something that we watch.

00:49:01.540 | But the main thing I wanted to point I wanted to make today on this

00:49:04.580 | is don't be lazy in believing the best consensus.

00:49:09.020 | Don't think that this is just about,

00:49:12.220 | you know, a negotiating tactic.

00:49:14.980 | Go back and read the speeches.

00:49:16.980 | I think Trump and Trump's administration has a much more principled view here

00:49:20.860 | that may be a value added tax equivalent.

00:49:23.780 | A tariff that is somehow, you know, a proxy for these value added

00:49:27.620 | taxes is a more efficient and better mechanism for helping

00:49:30.980 | the middle class in the United States than an income tax.

00:49:33.660 | And, you know, and we saw that that back door was kind of left open.

00:49:37.860 | You know, we said terrorists, maybe we may come back to terrorists

00:49:40.700 | as part of tax reform.

00:49:42.060 | So keep your eyes out for that.

00:49:43.660 | Well, then, look, any analysis of this situation is made more difficult

00:49:48.620 | by the blatant reality that even if Trump

00:49:53.980 | is just using terrorists as a negotiating chip,

00:49:56.860 | he can never say that out loud or it would take away

00:50:01.420 | the their ability to be used in that way.

00:50:04.700 | So he has to be obscure about it either direction.

00:50:08.420 | And so it makes it harder to know exactly which which ways up.

00:50:12.540 | So let's move on to Doge.

00:50:13.860 | So you were in Washington.

00:50:15.380 | What took you to Washington and and what did you see?

00:50:18.500 | What's your perspective of Doge?

00:50:21.340 | And like I love the update on your specific visit.

00:50:25.540 | But then could you reflect on why or why not Doge matters to a tech investor?

00:50:31.980 | Yeah, no, I think this is so I think it's so important

00:50:35.020 | because I think, again, like with tariffs, we're in this fog of war, Bill.

00:50:39.100 | Right. Where, you know, like their change brings a lot of contentiousness.

00:50:45.140 | And I think sometimes we lose, you know, the basic facts

00:50:48.420 | of what we're trying to achieve here.

00:50:50.100 | It's super important to understand that there is nobody like Elon.

00:50:54.220 | I mean, he's truly an end of one in working on issues like this.

00:50:57.900 | You know, so he's working 20 hours a day.

00:51:00.500 | He's working through the weekend.

00:51:02.060 | His team is, in fact, sleeping across from the White House

00:51:04.740 | and, you know, in the executive office building.

00:51:07.580 | But most importantly, right, Elon is a systems thinker.

00:51:10.620 | I mean, he literally showed up in Washington

00:51:13.660 | and he didn't do what normal people do when they show up in Washington,

00:51:17.540 | which is, you know, fed, you know, fed all the politicians and understand,

00:51:22.500 | you know, what he needs to do in order to play by all the rules

00:51:25.620 | that everybody else says.

00:51:26.540 | He just starts asking questions.

00:51:28.180 | And not surprisingly, the first question he asks is who sends out the wires?

00:51:32.500 | Like who controls the wires?

00:51:34.420 | Can I just get a list of all the wires that are scheduled to go out?

00:51:37.860 | Like, what are we what are we spending the money on?

00:51:39.900 | You know, over the next month, have they been audited?

00:51:42.620 | You know, and as I think he started doing that, of course,

00:51:46.100 | the Leviathan of Washington just convulses. Right.

00:51:49.540 | Because they're like, whoa, nobody questions.

00:51:51.620 | Nobody. Nobody looks at the wires.

00:51:53.580 | He's like, well, that's kind of what, you know, the president has asked me to do.

00:51:57.180 | So I need to do that.

00:51:59.380 | And so the antibodies, you know, really started attacking.

00:52:03.660 | When the only thing he really asked to do in the first instance is,

00:52:07.340 | you know, his first principles led him to thinking that, you know,

00:52:10.660 | like where where to look.

00:52:12.540 | So, you know, I want to bring it back to this idea

00:52:16.060 | that change is hard, but change is necessary. Right.

00:52:19.260 | We're simply talking about his whole purpose here

00:52:22.100 | is to balance the budget that both parties

00:52:25.460 | have proven the inability to do in the normal process.

00:52:29.220 | And most people agree it's it's bankrupting the country.

00:52:32.580 | And so it's not that hard.

00:52:34.940 | You know, we we showed this on the pod where we went through.

00:52:38.260 | If you just return to the baseline of 2019, Bill,

00:52:42.420 | if we just go back to the baseline, grow it by two and a half percent from 2019.

00:52:47.140 | You balance the budget in this president's term. Right.

00:52:50.500 | But that requires us getting a trillion dollars cut off the covid high.

00:52:55.660 | We lost our minds.

00:52:57.380 | Remember the letter to met a time to get fit.

00:52:59.980 | It was like we lost our mind.

00:53:01.260 | Well, Silicon Valley's gotten fit.

00:53:02.620 | We we've made some reductions, but government hasn't gotten fit at all.

00:53:06.260 | Hasn't done anything except stay at that covid high.

00:53:09.660 | And all Elon's saying is, listen, let's just go back.

00:53:12.380 | Let's start with just getting a trillion of this out, which is,

00:53:15.100 | you know, the excess that we put in. And so.

00:53:18.140 | The other thing that he's doing is he's literally live blogging this on Twitter.

00:53:23.260 | You know, he hosted this Doge spaces on Sunday night.

00:53:25.700 | Anybody, you know, can join this thing.

00:53:27.820 | It's not like they're hiding anything.

00:53:29.420 | It's Sunday night's a euphemism.

00:53:31.020 | It was hit midnight Easter.

00:53:33.300 | Which, you know, to me, I mean, I it's not that I mean,

00:53:39.340 | I think most people thought that was like him being mischievous.

00:53:42.140 | What what I do, I mean, the guy works around the clock.

00:53:45.300 | Right. That was when he had a free moment. Right.

00:53:47.420 | And they're like two two congressmen.

00:53:50.980 | Yeah. And so one of the things I tweeted on Sunday night

00:53:55.100 | after that was if we do this and if we tell

00:53:58.620 | if we if we tell the American people, you know, that we're going to do this

00:54:02.500 | and we put together a believable plan where you're going to cut a trillion

00:54:05.340 | dollars and balance the budget in the next few years,

00:54:08.180 | I'll tell you what's going to happen. Right.

00:54:10.340 | Interest rates are going to come down. Right.

00:54:12.740 | Because the whole reason bond vigilantes moved into the bond market

00:54:16.260 | and started shorting it is they thought, OK, here we go.

00:54:19.140 | Trump's going to stimulate the hell out of the economy

00:54:21.380 | with a continuation of taxes, et cetera.

00:54:24.020 | And nothing's going to really change on costs.

00:54:27.300 | And I think the big thing that I came back thinking is

00:54:30.980 | people are wrong.

00:54:33.300 | Right. Like there are there is a fundamental difference

00:54:36.580 | in how these folks are attacking,

00:54:39.060 | removing inefficient spending from the federal government

00:54:42.260 | and getting us back to what is a very sensible 2019 baseline.

00:54:46.100 | Remember, nobody thought in 2019, Bill,

00:54:48.980 | that we were like starving babies in the streets because our spending was so low.

00:54:53.300 | Nobody thought that. OK.

00:54:55.180 | Like everybody thought we were spending plenty of money in 2019.

00:54:59.100 | And that's all they're talking about.

00:55:01.220 | And yet, if you watch the convulsion coming out of Washington,

00:55:04.700 | you would think that, you know, something very draconian was going on.

00:55:09.460 | Well, I mean, but you would expect that, right?

00:55:12.140 | Like we don't have term limits.

00:55:14.580 | We have lifelong

00:55:15.740 | politicians in Washington.

00:55:19.660 | They we because of Citizens United,

00:55:22.580 | you basically can raise money, unlimited amounts of money

00:55:25.700 | from corporate interests.

00:55:27.180 | I you know, I gave the speech a year and a half ago on on regulatory capture.

00:55:31.860 | I I'm not surprised that the the entity

00:55:36.900 | that is Washington pushes back on someone that wants to take away

00:55:41.660 | the tools that give them power.

00:55:43.860 | I'm just not not surprised.

00:55:45.620 | Well, I think I think what people expected, frankly, is, you know,

00:55:49.900 | immediately after people started seeing the relationship with Trump,

00:55:52.820 | what's the first thing people did?

00:55:54.180 | They all started speculating how long until the relationship blows up.

00:55:58.060 | And, you know, Trump always fires everybody.

00:56:00.220 | And and just the opposite is happening.

00:56:02.340 | And then I think they all expected Elon just to come to Washington,

00:56:06.060 | not do anything like just to maybe make some recommendations

00:56:09.020 | to Congress on things that could be cut.

00:56:11.100 | But you and I know, Eli, like there's no chance he's going to Washington

00:56:14.700 | to just like, you know, run some research and make some recommendations.

00:56:19.260 | So I think that was misplaced.

00:56:21.340 | So let me tell you how I think Doge fits in with the normal budget process,

00:56:26.460 | because I also think this is very misunderstood. Right.

00:56:29.780 | So remember, I think the way to think about this in your head

00:56:33.740 | is we have two tracks going on here.

00:56:36.660 | Track one is the normal budget process.

00:56:39.660 | And in this case, they're using a parliamentary tool

00:56:43.580 | called reconciliation, OK?

00:56:46.060 | And basically what that means, I'll spare you the details.

00:56:48.980 | But this is out of the White House, led by Kevin Hassett in the House,

00:56:53.180 | obviously led by the speaker in the House Budget Committee.

00:56:55.660 | But basically, reconciliation is a special budget process

00:56:59.220 | that allows you to get an omnibus budget bill past Congress

00:57:03.060 | without having to get to the 60 votes in the Senate

00:57:06.460 | that is filibuster proof, OK?

00:57:08.900 | And they're working hard on this.

00:57:10.620 | I expect some meaningful improvements that will come out of this in spending.

00:57:14.420 | I suspect that Doge will be offering their ideas how to save some money in this.

00:57:18.500 | But this is kind of the normal process that occurs in Washington.

00:57:21.980 | And the president, I think, has said he wants something to sign out

00:57:25.180 | of the reconciliation process in April or May. Right.

00:57:28.900 | And so it has to go through this normal.

00:57:31.140 | All the committees are going to have their hearings.

00:57:33.340 | They're going to put together the budget that they think complies

00:57:36.060 | with reconciliation.

00:57:37.060 | There's going to be a grand negotiation,

00:57:39.140 | you know, that occurs with 10 people around the table.

00:57:41.540 | And, you know, all the horse trading that usually occurs in Washington.

00:57:44.780 | So that's track one bill.

00:57:46.820 | Track two is Doge and cuts in spending by executive authority.

00:57:53.100 | And this is the part that I think has Washington up in arms.

00:57:58.020 | So that's what you see that's causing the fury.

00:58:01.500 | Elon is advising the president and then the president is deciding in real time

00:58:07.340 | whether certain people need and need to be cut

00:58:10.940 | and whether certain spending should be stopped.

00:58:14.660 | And when the answer is no, this amount of money and these people

00:58:18.620 | are not required to faithfully execute the laws that I've been given.

00:58:22.260 | They say they're just going to downsize the downsize,

00:58:25.020 | the executive agency tasked with executing the law,

00:58:28.020 | and they're going to stop spending the money that they believe is wasteful

00:58:31.180 | and not needed to fulfill the law.

00:58:32.980 | So they're saying, especially in the face of a national fiscal crisis

00:58:37.220 | where we're falling further and further into a debt spiral, we need to do this.

00:58:41.820 | So they in the town hall on Monday night or on Sunday night.

00:58:44.860 | You know, for example, Elon called USAID.

00:58:48.780 | So this is an organization that's quite controversial.

00:58:52.180 | You research it.

00:58:54.220 | That spends 50 billion dollars a year on foreign aid. OK.

00:58:57.940 | And it has thousands of people in the agency.

00:59:01.780 | And he said, well, between I think among employees,

00:59:06.020 | it's probably closer to a thousand or two and then a lot of of of contractors.

00:59:10.780 | And basically what Elon said on Sunday night is I called the president.

00:59:14.460 | I told him, unfortunately, there's no apple to be saved.

00:59:17.340 | It's a total ball of worms.

00:59:18.740 | If there was just one worm in the apple, we'd pull the worm out.

00:59:21.420 | But the whole thing is a ball of worms.

00:59:23.140 | So the whole thing needs to be shut down.

00:59:25.860 | And we're we're going to let thousands of people go

00:59:29.100 | and we're going to save 50 thousand dollars on the budget

00:59:31.300 | or 50 billion dollars on the budget.

00:59:34.220 | It subsequently looks like on Monday that, you know, they made a deal

00:59:38.340 | where Marco Rubio, right, who's the secretary of state,

00:59:41.780 | is going to become the acting director of the agency.

00:59:44.220 | And now it looks like they're going to eliminate

00:59:46.020 | whatever they think is wasteful, and then they'll consolidate

00:59:48.580 | perhaps other parts of that spending into the State Department.

00:59:52.380 | But basically, this is what Bill caused Schumer and folks

00:59:56.780 | to come out on Monday morning, declare all of this activity unconstitutional

01:00:01.860 | to say that, you know, nobody elected Elon.

01:00:05.980 | He can't do this. It's unconstitutional.

01:00:08.100 | And this is where I think you're going.

01:00:09.700 | The whole challenge is now going to move.

01:00:11.660 | But remember, this has nothing to do with track one.

01:00:14.180 | Right. Except you're angering a lot of people on the Democratic side.

01:00:18.140 | But this is really about track two.

01:00:20.060 | Does the president have executive authority

01:00:22.420 | not to spend money that they deem is wasteful?

01:00:25.540 | So you asked me a question and maybe we'll touch on it

01:00:30.580 | for a second earlier, which is, is it constitutional? Right.

01:00:35.900 | And so I think that's a pretty fascinating constitutional question.

01:00:39.700 | I've consulted with a lot of people I think are experts in the area area.

01:00:43.460 | And I do expect that Schumer or a group of members as early as this week

01:00:47.900 | is going to file, you know, a claim, a lawsuit in federal court

01:00:53.700 | where they say that this is a violation of the Constitution

01:00:56.660 | under Article one, Section nine, Clause seven, where Congress

01:00:59.700 | has the power of the purse strings and the Supreme Court,

01:01:02.460 | you know, has long upheld this.

01:01:05.140 | You know, basically, the Supreme Court has said separation of powers

01:01:07.740 | generally support the idea that it's Congress who appropriates funds

01:01:11.700 | and anybody else who doesn't spend those monies that would be unconstitutional.

01:01:15.100 | So that that's likely the argument they're going to make, Bill,

01:01:18.860 | and they're going to say immediately they got to cease and desist

01:01:22.060 | from, you know, Elon shutting off wires

01:01:25.540 | or not spending money or shutting down USAID.

01:01:28.140 | Now, I happen to think that's on pretty weak footing.

01:01:30.500 | OK, but it is.

01:01:32.260 | I think it's going to happen on weak footing. Why?

01:01:35.420 | So just think about this for a second, right?

01:01:37.260 | The president has the authority to execute the laws.

01:01:41.020 | And there's this doctrine that's known as impoundment,

01:01:44.820 | which the courts largely recognize.

01:01:47.180 | And it's basically the president saying, OK, I see the law

01:01:50.300 | that we're supposed to uphold and I don't need all this money.

01:01:53.740 | And in fact, I have a further and maybe supreme duty,

01:01:57.860 | an overriding duty to the Constitution that supersedes

01:02:02.580 | the constitutional control of the purse, to execute faithfully the laws,

01:02:06.340 | to protect the general welfare of the American people, which he might argue

01:02:10.220 | includes protecting the country from bankruptcy.

01:02:13.060 | Right. So he's just saying, listen, I'm doing my duty.

01:02:16.700 | Yes, I'm executing all the laws they told me to execute.

01:02:20.180 | However, I'm doing it for less money.

01:02:22.500 | And and given that we're in a national debt crisis,

01:02:25.420 | I need to do that in order to protect the American people.

01:02:27.900 | So I think that this is going to eventually come to head.

01:02:32.180 | Imagine it goes to the to the Supreme Court to decide.

01:02:35.180 | And I think there's a decent chance along the way that at a minimum,

01:02:39.740 | think about what what Chuck Schumer is going to have to defend.

01:02:42.980 | He's going to have to defend some of this really crazy spending

01:02:46.700 | that we all know exists.

01:02:48.620 | I mean, I don't think there's anybody in either side of this argument

01:02:51.460 | who doesn't think there's a bunch of inefficient

01:02:53.460 | and silly spending by the government.

01:02:55.460 | So effectively, that's the that's what you're going to have to defend

01:02:58.980 | if you want to defend this lawsuit.

01:03:00.740 | So I think the political pressure is going to be massive.

01:03:03.460 | That's brought to bear, particularly because Doge is being so transparent

01:03:07.660 | on this, right?

01:03:08.460 | Like you do not want to be defending every single line item

01:03:12.460 | to the American people,

01:03:13.700 | which is exactly what Doge is going to put you on the spot to do.

01:03:16.900 | And so I think two potential outcomes.

01:03:19.820 | Number one, the political pressure causes, you know, them to cut a lot more

01:03:24.660 | as part of track one, right, this reconciliation process.

01:03:28.820 | Or number two, that the Supreme Court actually does, in fact, recognize

01:03:33.540 | some more expansive, you know, executive power around impoundment.

01:03:37.500 | But, you know, I think either way, you know, I imagine before

01:03:41.860 | this is all said and done, Bill, that we're going to see headlines

01:03:44.900 | that say Elon causing a constitutional crisis, right?

01:03:48.780 | That, you know, we have the we have the courts involved

01:03:51.540 | and you have the solicitor general that's that would be defending

01:03:56.340 | the executive branch in the White House on on this matter.

01:03:59.580 | Now, bring it home, you know, and as I said in the question, bring it back.

01:04:05.860 | Like, why does this matter for tech investors?

01:04:09.140 | What? Let's presume it goes either way.

01:04:12.380 | What's it going to mean to how a tech investor should be thinking

01:04:15.580 | about the markets and and tech stocks?

01:04:18.660 | Yeah, I mean, just think about what were the three topics we talked about today.

01:04:23.460 | The first one was like just massive technological uncertainty, right?

01:04:27.140 | Like where you said it, it's a pace of change you've never seen in your career.

01:04:31.500 | Highly disruptive, multi, you know, companies that are valued

01:04:35.100 | at one hundred and fifty billion dollars that are being challenged by,

01:04:39.060 | you know, a Chinese startup on a shoestring.

01:04:42.100 | So you and I would both say our ability to forecast the future

01:04:45.580 | as to where this is going, right, is is challenged because it's moving so fast.

01:04:50.340 | Then we talk about tariffs.

01:04:51.580 | Well, massive economic uncertainty, Bill.

01:04:54.420 | I mean, you know, free trade has been generally established

01:04:58.140 | as a principle in the economy for the better part of

01:05:00.940 | certainly for you and you and my entire investment career.

01:05:04.580 | The markets could count on that.

01:05:06.780 | And now I'm suggesting that at least there's some probability

01:05:10.020 | that this president is going to move in a very different direction, right?

01:05:14.300 | That maybe it's not the best consensus, that maybe it's something else.

01:05:17.900 | It's it's a we might call it the Trump new normal, right?

01:05:21.900 | Where tariffs become standard practice and maybe as a replacement to income tax.

01:05:26.460 | So, OK, there's a lot of uncertainty around that.

01:05:28.540 | And now this third one is political, right?

01:05:31.420 | Like it's been a while since we had a looming political

01:05:34.100 | constitutional crisis where, you know, where an issue between

01:05:38.020 | the congressional branch and the executive branch went to the Supreme Court.

01:05:42.220 | That also yields a lot of uncertainty when you add these uncertainties up.

01:05:46.340 | What does it do for the value of assets that you and I look at?

01:05:49.060 | Right. You and I are valuing those future cash flows.

01:05:51.820 | We have to apply a discount rate.

01:05:53.820 | Discount rate measures the risk associated with those future cash flows.

01:05:58.580 | So you and I have to take the discount rate up.

01:06:00.820 | Why? Because we're a lot less certain about technology, politics and economics.

01:06:05.740 | So to me, that means multiples come down and asset prices have to come down.

01:06:09.820 | Why? The world sifts through all this.

01:06:11.580 | The surprising thing to me really, Bill, is how well

01:06:14.620 | the public markets have held up in the face of all of this. Right.

01:06:18.020 | And I think part of that has to do with they believe Trump

01:06:20.860 | is going to be a super pro-growth president.

01:06:22.980 | You're going to have lower taxes, et cetera.

01:06:25.500 | But I think that's the risk on the table as a risk manager.

01:06:28.980 | What I have to say to my team is, OK, we've got a downsized risk, right?

01:06:34.340 | We you know, you don't do the 10th best idea, right?

01:06:37.980 | Or the 11th best idea.

01:06:39.220 | You really got to make sure that you that you better understand this stuff.

01:06:44.060 | And so I think for the long term investor, you know, perhaps

01:06:46.780 | they can just ignore the noise and they could say, you know, I'm fully involved.

01:06:50.100 | I'm fully invested.

01:06:51.260 | I believe in this super cycle.

01:06:52.860 | AI is going to be great for everything.

01:06:54.940 | But what I would say to our friends in Silicon Valley is expect way more volatility.

01:06:59.420 | I think the next six months, all of this uncertainty

01:07:02.940 | means that you're going to have a lot of volatility.

01:07:05.620 | It's exactly what we felt all weekend long.

01:07:07.860 | It's exactly why the markets were gapping down overnight on Sunday.

01:07:11.340 | And then they did this U-turn because we got a change in policy

01:07:14.220 | or what appeared to be a change in policy out of the White House.

01:07:17.060 | And so, you know, welcome back to 2017, Bill.

01:07:20.180 | All of this change may be absolutely necessary and totally good for Team America.

01:07:26.180 | But it's going to it's it's going to mean that we have more sleepless nights.

01:07:30.900 | Yeah. And look, I think I think

01:07:33.940 | because of so much chaos and because of such

01:07:38.300 | massive uncertainty in regulatory action

01:07:42.300 | and the fact that it can often backfire.

01:07:44.260 | I mean, I give you the example, like when Deep Seek popped up.

01:07:47.660 | First of all, I think this whole economic thing kind of got blown out of proportion.

01:07:51.860 | The paper originally said six million was just the post training.

01:07:56.420 | And somebody interpreted that as the whole thing.

01:07:59.220 | And then it led to it.

01:08:00.300 | But but a lot of people then I would say a lot of Nvidia bulls

01:08:04.220 | ran out and said, no, no, no, they had way more Nvidia.

01:08:07.460 | They had way more GPUs than they thought.

01:08:09.460 | Well, guess what?

01:08:11.860 | Pounding the table on that may cause some in Washington to say, oh,

01:08:15.380 | we should have had higher restriction.

01:08:17.780 | So you're an Nvidia bull.

01:08:19.340 | Think you're protecting Nvidia by exposing Deep Seek.

01:08:23.620 | And you may end up with sanctions that end up hurting Nvidia's revenue.

01:08:27.380 | So it's a it's a it's a dangerous place to play.

01:08:30.740 | Well, I mean, you just you just showed I mean, you have all of these forces at play.

01:08:35.780 | Listen, what do I do as an investor?

01:08:37.900 | You know, what did I do in the fall of 22 that led me into Nvidia in the first place?

01:08:41.700 | I just studied what was happening in technology in the company.

01:08:44.820 | I didn't have to think about free trade or tariffs.

01:08:47.420 | I didn't have to think about export restrictions.

01:08:49.860 | I didn't have to think about, you know, constitutional crisis.

01:08:53.380 | All I had to figure out is, is the forecast for Nvidia too low

01:08:57.060 | because of the explosion?

01:08:58.140 | We're about ready to have an AI.

01:08:59.820 | And that's the bet we made and we won big on.

01:09:01.980 | But now, as I sit here today, the valuation for Nvidia is much higher.

01:09:06.060 | And now I have to take into consideration all these other risks.

01:09:09.700 | And all I'm saying is all else being created equal.

01:09:12.980 | I think that the super cycle is, is, you know, if I'm just doing

01:09:16.580 | my fundamental analysis, I think it's on fire.

01:09:19.020 | I think we're going to need way more compute than we have.

01:09:21.220 | I think DeepSeek unleashes the amount of inference we're going to need.

01:09:24.500 | I think deep research out of OpenAI unleashes that.

01:09:27.140 | So I think the fundamental is bigger than ever.

01:09:29.540 | But at the same time, I also am humble in the face of what's known and knowable.

01:09:36.060 | About the next 12 months, you know, around tariffs, around export controls,

01:09:41.420 | around all these other risks in the economy.

01:09:43.580 | And I just think we have to, you know, you have to look in the mirror

01:09:46.820 | and acknowledge that a lot of this is unpredictable.

01:09:49.780 | And that impacts what folks are willing to pay,

01:09:52.580 | what multiple folks are willing to pay.

01:09:54.380 | And for our friends in the VC markets, right, particularly

01:09:57.820 | some of these high valued companies, you know, in mid and late stage VC,

01:10:01.420 | it's going to impact, right?

01:10:03.140 | There's always that lag effect.

01:10:04.580 | But the public markets and the risk appetite and the multiples they pay,

01:10:07.740 | that rolls downhill.

01:10:09.060 | And so I would just say, I think that we may get to the back

01:10:12.260 | half of this year or into next year.

01:10:13.940 | And it may, in fact, be the golden age and off to the races.

01:10:16.740 | But I think at the moment it's the golden age of uncertainty.

01:10:19.740 | Hey, Brad, let's let's I think that's that's well said.

01:10:23.980 | Let's close with where we started with this sovereign wealth fund thing.

01:10:28.300 | So pick either the pro or the con, make the argument,

01:10:33.820 | and then I'll take the other side.

01:10:36.740 | Well, I mean, like I think you saw me post in our thread

01:10:39.900 | as a general matter, right?

01:10:42.700 | I'm you know, I love the fact that we have people with business

01:10:46.940 | sensibilities and and incentives looking out for America

01:10:50.460 | who want to negotiate on behalf of America. Right.

01:10:53.500 | And who, you know, we sell we sell wireless spectrum and licenses. Right.

01:10:57.660 | I would love to see that go to the benefit of all the citizens in the country.

01:11:01.700 | You know, we we, you know, have, you know, drilling licenses on national lands.

01:11:07.220 | That's that that money belongs to the citizens.

01:11:09.780 | So I love that idea.

01:11:11.260 | You know, in the fund, here's my challenge with it, Bill.

01:11:13.460 | We have 40 trillion in debt

01:11:16.460 | and we probably have another 50 trillion of unfunded liabilities.

01:11:19.740 | So we're a debtor nation and we're paying five percent on all that debt.

01:11:25.140 | So the hurdle rate to our return that is needed on the sovereign wealth fund,

01:11:30.060 | right, is five percent.

01:11:31.140 | Otherwise, you would just take all those monies and you would pay down the debt.

01:11:34.260 | Right. If this was our personal balance sheet.

01:11:36.300 | So what you and I would describe this as is we're levering up

01:11:40.100 | the balance sheet of the United States to earn the spread

01:11:43.300 | between the sovereign wealth fund returns and the five percent

01:11:47.060 | that we're paying to borrow all the money.

01:11:48.860 | So I think net net, I probably have it in place

01:11:52.300 | because I think it's a good tactical lever.

01:11:54.420 | You know, I do worry about what administration administration

01:11:59.180 | it could lead to some crony capitalism and deal making that benefits

01:12:02.940 | certain people, et cetera.

01:12:04.500 | So I don't know.

01:12:05.740 | It's a close call for me, but I think it's going to happen either way.

01:12:08.500 | I would you took both sides.

01:12:10.820 | So I'll do a quick both sides.

01:12:13.100 | Yeah. I mean, in addition to the scenarios you talked about

01:12:15.980 | and you and I have debated this in the past, but if I look at Goldman Sachs, GM,

01:12:21.940 | even United, like if the government's going to be the lender of last resort,

01:12:26.660 | I would argue they should take all of the equity.

01:12:29.020 | And this could be a vehicle for that.

01:12:31.100 | Although there's nothing that's kept the government from doing that.

01:12:34.220 | I think in the GM case, they did take equity.

01:12:36.260 | So it's like it has done it in the past without the vehicle.

01:12:40.100 | I'm way more skeptical than you on the crony capitalism.

01:12:44.220 | It would be a ninety nine percent certainty that this asset would be rated

01:12:49.660 | from from transition to transition in the government.

01:12:53.740 | And I would highlight that probably the most successful

01:12:58.060 | sovereign wealth funds in the world are all in autocracies.

01:13:02.340 | They're not in democracies.

01:13:04.740 | I don't agree. I don't agree with that.

01:13:07.740 | I don't agree with that.

01:13:08.900 | I mean, from from from Norway with Norges to Korea to Canada,

01:13:14.380 | we have a lot of great sovereign wealth funds.

01:13:16.660 | But I do think, you know, it's fair.

01:13:18.420 | I think one of the things I would say in all those countries,

01:13:21.380 | what you do have, Bill, is consistency and independence

01:13:25.460 | in the management of the sovereign wealth fund, independent

01:13:29.100 | from like some unilateral control by the by the executive branch.

01:13:32.820 | And then if you want to see, I can go really wrong.

01:13:36.340 | Please go read Billion Dollar Whale.

01:13:39.060 | It's one of the most exciting books you could possibly read

01:13:43.020 | about what happened to the Malaysian sovereign wealth fund.

01:13:45.980 | So, well, as always, it's fun to get together.

01:13:50.380 | Thanks for making the time. We'll talk soon. All right.

01:13:53.260 | As a reminder to everybody, just our opinions, not investment advice.

DeepSeek, Open Source, Tariffs, DOGE, Market Impact | BG2 w/ Bill Gurley & Brad Gerstner

Chapters