Economists Uncut

What Happens To Markets Once DeepSeek’s True Risks Are Priced In (Uncut) 02-03-2025

What Happens To Markets Once DeepSeek’s True Risks Are Priced In | Bradley Tusk

NVIDIA has fallen almost 19% this week as the Chinese firm DeepSeek has disrupted American tech markets. DeepSeek’s AI model R1 was developed at a significantly lower cost than OpenAI’s models and doesn’t rely on high-end expensive hardware and expensive GPUs, which is why NVIDIA has been falling. But could DeepSeek actually be bullish for chip makers like NVIDIA? And then we’ll talk about some other developments as well, including Alibaba’s Quen, which is supposedly even better than everything else out there.

 

What is going on with the tech landscape? Should we still be looking at American tech companies? Let’s find out with our next guest, Bradley Tuskey, is a CEO of Tesk Ventures and author of the book Vote With Your Phone, Why Mobile Voting is Our Final Shot at Saving Democracy. Welcome back to the show, Bradley. Good to see you.

 

Yeah, thanks for having me. Happy 2025. Are stock markets correctly pricing in the risk to AI from DeepSeek, as you know, earlier this week? No.

 

Yeah. No, absolutely not, in the sense of the stock markets have not been completely rational about any of this for a while now, right? The endless bull run on NVIDIA, in my view, wasn’t totally logical because, yeah, if what we were seeing at that moment in time was going to be the case forever, sure. But as we know, things always change, right? The only constant you can count on is change.

 

So just like it wasn’t going to be that NVIDIA’s GPU was going to be the only ones that anyone was ever going to use, which shows why I think the driving up the NVIDIA stock didn’t really make sense, it also doesn’t mean NVIDIA is now a useless company simply because DeepSeek has said and released a report saying, here’s what we’ve achieved. The report that, by the way, we haven’t even validated yet. I saw that OpenAI came out today and said, well, they used our models to build on.

 

So if that’s true, that sort of invalidates a lot of the DeepSeek argument. So no, I think that we have not seen rational behavior around AI investing, both in the public market side, or quite frankly, in my side of the world, on the private venture side either. The question is whether or not this deeply affects the American chip market.

 

Take a look at my screen here. Let me give you another development. This came in this week as well.

 

Alibaba’s Quen. So Chinese tech company Alibaba on Wednesday released a new version of its Quen 2.5 AI model that it claims surpasses the highly acclaimed DeepSeek v3. Quen 2.5 max outperforms almost across the board chat GPT-4 and DeepSeek v3 and Lama 3.1. Alibaba’s cloud unit said in an announcement posted on its official WeChat account.

 

So this is a company announcement. So engineers still have to verify this. But Chinese companies have been able to do this despite heavy restrictions on exports of American advanced chips.

 

The question is how and how this is going to impact. Come back to us, the investors. Yeah.

 

I mean, a couple of things. The first is ironically, it was our restriction in the first place that led to the Chinese innovations, if they are true and accurate, that found ways to be able to make do without the most advanced chips. So I think as President Trump thinks about all of his different tariffs and trade policies and restrictions and all the ways he wants to have conflict and fight with people, you know, be careful what you wish for, because sometimes what sort of might feel good emotionally or make sense immediately turns out to not be the right solution.

 

So that’s number one. Number two, though, is, you know, on Alibaba specifically, what I didn’t hear was, you know, what the impact was in terms of needs for compute, needs for energy. You know, in terms of just one system outperforming the other, you know, there may be, you know, if you’re trying to detect cancer, let’s say, right, where every extra tiny percentage point could make a meaningful difference.

 

When you’re using something effectively as a search engine, like, does it really matter to me if, you know, perplexity is better than Alibaba or Alibaba is better than open AI or better than Anthropic? No, not really, right? You know, I’m just asking pretty simple questions, by and large. So, you know, there, and I think what Lisa’s a larger point, which is different use cases are going to exist. Some people are going to need the absolute best technology, even if that requires the most compute and the most energy costs.

 

Some people are going to say 90 percent functionality is more than good enough for what I need, and I will take the lower costs and the cheaper chips. And then beyond that, for example, I met with a startup this week and, you know, we haven’t done any technical diligence on it, so I can’t verify it, but they talked about having a full software stack that effectively acts as a universal adapter that could take anyone’s AI code and run it on any chip, right? If that’s true, that changes everything again. So, point just being, there are so many different things being worked on that will reshape this market time and time again, that I think before anyone plunges all in in either direction, they should be careful.

 

This came up this week, the idea that more efficient AI is going to be a boon for AI stocks. Okay, so Jevin’s paradox. What is it? Why do we care? This surfaced this week as Microsoft CEO started talking about Jevin’s paradox strikes again.

 

He tweeted, as AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can’t get enough of. At the heart of the Jevin’s paradox is the idea that as a technology becomes more energy efficient, the cost of using the technology declines, the lower cost spurs consumption, the increased consumption in turn rises to wipe out any decrease in energy use, the efficiency gains might have initially represented. So, is this just actually good news for NVIDIA, meaning we’ll actually end up using more NVIDIA chips? No, not necessarily.

 

I think it’s good news overall for the AI sector and the tech sector, because ultimately for tech to work, it has to become commoditized and the cost of operating, it has to be low enough that it could be done at scale. And so, you want lower computing costs and less energy usage. So overall, yes, I agree with that.

 

I mean, Moore’s law is the same thing, which is if the power of computing on any given thing doubles roughly every two years, of course, it’s going to keep getting more and more efficient and therefore more and more widespread adopted. But one, if you’re Microsoft, that’s all true, except if people decide that OpenAI’s model is just not better enough to justify all of the higher costs. Microsoft’s got a pretty substantial investment in OpenAI, so other Microsoft products might benefit a lot from this overall trend, but their investment in OpenAI might suffer.

 

And two, if you’re NVIDIA, no, I mean, if you’re totally focused on the high-end market and everything gets completely commoditized and you can’t keep up with that, you might be completely out of the game running altogether. So, yeah, if NVIDIA says we have a wide range of chips able to serve every price point and every need, and they can really meet the demand for all that, sure, then I think overall, the increase in adoption is good for them. But if it’s an increase in adoption at sort of lower price point, lower efficiency chips or whatever else that they don’t make, that’s bad for them.

 

Okay. To your point earlier about why, or perhaps the people at DeepSeek took OpenAI’s code. So this is what you’re referencing.

 

OpenAI says Chinese rivals use OpenAI’s work. The maker of ChatGPT, OpenAI, has complained that rivals, including those in China, are using its work to make rapid advances in developing their own AI tools. What do you think the government is going to do, the U.S. government is going to do in response to this particular development? You know, the answer is really finding that right balance, right? So Biden did an executive order on AI around a year ago.

 

Trump rescinded it. I didn’t mind the rescission because honestly, I didn’t think the Biden executive order did all that much in the first place. And the real things that you’re trying to balance if you were a policymaker, and let’s say Trump, you were someone that was just more objectively concerned with trying to reach the best possible outcome for the American people, then you would say, one, I want U.S. companies to be able to compete as aggressively as possible with Chinese AI companies and everyone else.

 

And that might mean, for example, allowing more open source technology in the space. Two, we want to have security so that our AI companies aren’t easily hacked or their technology is stolen in different ways. But three, we also want to recognize that AI might offer tremendous potential to the public, but also harm, right? So social media and Internet 2.0 in many ways has created an entirely new economy that’s done a lot of good, but we also are all very familiar with the absolute toxicity of meta and TikTok and Twitter and all these other platforms and teenage suicide and self-harm and eating disorders and just all the terrible things that we’ve never regulated, right? The U.S. government has completely failed to provide any sort of guidelines or regulation to Internet 2.0, and it has led to absolutely horrible outcomes for millions and millions and millions of people.

 

And so you have to take that into account as well. So look, regulation, you know, here’s where I think I differ from some of my fellow VCs. You know, someone like Marc Andreessen would just say regulation is inherently bad.

 

You know, as you know, David, the first half of my career was working directly in government and politics at the state level, city level and the federal level. And so I’ve seen this from both angles. And the answer is it’s not inherently good and it’s not inherently bad.

 

It’s solely based on whether or not the regulation solves a specific problem, right? Leaving markets entirely unregulated tends to lead to widespread abuse and over-regulating them like they do in Europe leads to a lack of innovation. And what we need is the right regulation. Bradley, you’re a VC.

 

The fact that DeepSeek made its code open source, does that make you more skeptical or less willing to invest in some other AI companies in the space in this particular domain? Yeah. Well, first, yes. So I don’t invest in generative AI because it’s just too expensive, right? We are early stage investors.

 

Our average check size is $3 to $5 million. So, you know, for me to own 0.001% of a generative AI company doesn’t make any sense. But look, take a step forward, which is, you know, companies using AI building on top of models, you know, like OpenAI or DeepSeek.

 

I would now, you know, ask all of those companies that I’m looking at investing in or ones that I’m already in, what are you using and why? And if the answer is the more expensive option, I want a really good reason for that, you know, because if it turns out that there’s an open source solution that is exponentially cheaper, I think you have to seriously explore that. So now look, again, maybe DeepSeek isn’t really cheaper, and they did use OpenAI’s sort of tech to build, train all the models, and therefore actually it’s not true. But let’s just say, take for, you know, a moment, let’s assume that everything they said is true.

 

Then I would be worried if even if I were a generative AI investor in investing in closed systems, because open source ultimately is going to have a lot more adoption. Well, this begs a question as to what the future of the NASDAQ is, broadly speaking. Take a look at my screen here.

 

This is the NASDAQ composite, and it’s just been trading range bound ever since basically the middle of December. And of course, with the recent news with DeepSeek, investors are kind of worried about whether or not semiconductor chips even have a future in the US. What’s your outlook for the next quarter? What does the NASDAQ need to do to turn this consolidated pattern around? Yeah, I mean, I think there’s two approaches, right? One would be what you see open AI doing right now, and you’ll see others as well, which is try to discredit DeepSeek as much as possible to restore that faith in the American chip market and in all the companies that are sort of pushing AI and try to just sort of continue the past.

 

The other would be to say, just accept there’s going to be a lot of evolution and change here, and it’s going to be things we don’t know about until a company announces it, and then we see if it’s verified or not. And therefore, an absolute bull run at all times doesn’t make sense, and nor does selling everything off just because someone releases a white paper on the internet. And therefore, maybe the NASDAQ doesn’t need to go crazy in either direction over the next quarter, and the best thing they can do for themselves is preach some level of common stability.

 

Earnings are coming up in not too long. What are you expecting for this year? What’s going to drive margin expansion or top line growth for tech companies this year? What’s the narrative in 2025? Yeah, I mean, I think ultimately, keep in mind, the vast majority of what we just call tech is not really using AI. So yes, there are AI-enabled products now.

 

Yes, when you even go on Google, you get sort of an AI summary of the search results. But by and large, like the fundamentals of this are still pretty simple, which is you have to make a product or service that people want to pay money for. And I think where both VC and the public markets got off base was this notion that if someone has a lot of customer growth, even without good unit economics and margins, or they’re just in some space that sounds really cool, then we should go all in on it without really using the normal metrics that we would use to evaluate any company and any stock.

 

And the answer is, it’s the same thing, right? So, you know, just buying customers, if you can’t ultimately ever be profitable, is not a good company, and no one should invest in that company. And so I would say that while the last four years of the VC have been pretty awful in the sense that there’s been virtually no liquidity whatsoever, the one thing that I do hope that remains is some level of rationality around the fact that common sense should be part of the mix here. Unit economics and margin are part of it, and it’s not just hockey stick growth at all costs, and it’s not just like buy into a logo or a narrative at all costs.

 

To me, like, that’s dumb money. Okay, well, what’s happening in the VC space? Can you give us an update? What’s hot this year in 2025? Yeah, I mean, I think the first thing is from a macro standpoint, you know, there are two trends that should be helpful. The first started last year and maybe now is reversing, but, you know, the reduction in interest rates.

 

You know, when people through fixed income with a lot less risk can get the kind of returns that they were getting, there’s no weirdo need to take the risk of venture capital, right, and do something so speculative. And as a result, fund formation went way down, new venture funds, you know, coming to light or went way down, and then therefore their VC’s funding of new startups went way down as well. And that’s really bad for innovation long term.

 

Lower interest rates clearly change that pattern. So my hope is that Powell does continue to cut or at least doesn’t raise rates. The other, you know, is change in regulators, right? So the vast majority of tech is actually not regulated by Washington.

 

It’s regulated by state and local governments. But, you know, one of the things that I think the FTC did that was problematic is in their efforts to rein in some of the really big tech companies like Google and Meta and Apple and whoever else, they inadvertently created this sort of wider chilling effect where they just sort of sent this message that all M&A is no longer okay, and that all corporate activity is considered, you know, bad because the left thinks that business is inherently evil. And that just led to a vast reduction in M&A activity.

 

And, you know, some of my company’s IPO, but most of the exits, if there is one, is because of an acquisition. And so I think by having new leadership at the FTC, new leadership at the SEC, I think if you have Doge ultimately actively pursuing adoption of new technologies to make government itself work a lot more efficiently and better, those are all really, really positive signals. And so the combination of all of that combined with massive pent up demand for companies to be able to go public that are ready to go, combined with a more favorable interest rate environment should lead to a better year of VC.

 

Well, this leads to our discussion about Trump’s appointment. So David Sachs, billionaire entrepreneur, was appointed the AI czar. I believe that’s a first for the country.

 

And he recently said that he believes there’s evidence to support the fact that OpenAI’s code was copied by DeepSeek. But anyway, to the broader point, AI czar appointed David Sachs. What do you think of David Sachs? What’s he going to do for the AI industry? David’s a smart guy.

 

I went to law school with David, so I’ve known him for a very long time. He’s also very Republican, very libertarian. And so it will be a very free market approach to both AI and to crypto.

 

And again, that’s not necessarily a bad thing at all. You just have to sort of be mindful of the fact that consumer protection is also really important. So for example, let’s use crypto because it’s also what the cryptos are.

 

Having the SEC be a lot less punitive towards crypto is a good thing. Having the SEC give crypto companies clear guidance about what is or is not allowed, like margin lending or whatever it is, is a good thing. At the same time, it is an industry that is still rife with fraud.

 

And you do need someone looking out for consumers and protecting them from fraud on ICOs and things like that. And so you need a mix. And by the way, you need the mix not just because that’s the job of government to protect consumers, but because the legitimate players in the industry suffer when there’s no enforcement at all of the rules and you’re playing by the rules and the other people are not.

 

And so I think David in many ways will have really good ideas and instincts around growth in both AI and crypto. And I think what he’s got to be able to do though, is just take a half a step back and say, okay, but what is the impact on the broader public? Not just what’s the impact on Silicon Valley or the stock market and my friends, but what’s the impact on regular people? And find the right mix. And that doesn’t mean endless regulation and stagnation, but it also doesn’t mean just everything goes no matter what, because they, both of those result in bad outcomes.

 

Let’s talk about what’s going on with TikTok. So TikTok ban. So it’s facing a ban.

 

Trump has signed an executive order that is going to extend the ban. He wants, Trump said that he wants to make a deal. Somebody buys the company, the US government owns 50% of it.

 

What do you think is the fate of TikTok? Have you looked into this? Yeah, I have. Look, ultimately, I think the real question becomes, it’s not up to Trump, it’s up to the CCP. Question number one, do they want to sell it at all? And two, what are they going to sell? So if you sell TikTok and the algorithms, that’s a very valuable property and there will be lots of US demand for it.

 

And whether or not it makes sense for the US government to own half of it, I’m not sure, but either way, clearly you could execute a deal. On the other hand though, if the CCP says, you know what, we’re going to have a lot of negotiations with Trump over the coming years on AI, on trade, on tariffs, on Taiwan, and let’s hold this back as a bargaining chip. So we’re just going to let TikTok go dark for now.

 

They could choose to do that too. I mean, keep on $50 billion, whatever the price would be for the US subsidiary TikTok, means nothing to China, right? So they don’t need to do the deal just for the sake of doing a deal. Or if they say, we’re going to sell the TikTok brand, but we’re not going to sell the underlying algorithms, that’s much, much less valuable.

 

And it’s an entirely different sales process and auction process. So that’s to me where the real decision-making lies. The other thing though also is, you know, Trump clearly looks to push the bounds of his legal authority at all times.

 

Yes, he had the ability to do the 75-day pause, but ultimately this was a law passed by Congress in a bipartisan way, upheld by the US Supreme Court unanimously. So people from both sides. And ultimately, either a deal has to happen in the relatively near future, or TikTok can no longer be hosted on US servers and available on US app stores.

 

And that ultimately leads to an availability of the product. So my other question is, you know, if and when the 75-day period comes and goes, who is suing to enforce that ultimately the provisions of the ban go into effect? I mean, the company that seems the most position to gain here is Medec, that’s 170 million customers up for grabs. But Zuckerberg also seems absolutely petrified of Trump.

 

And so I don’t know that he would do that. And I’m not quite sure who has standing. I think it might be any American person, in which case, you know, there are consumer groups that will definitely sue.

 

Well, China’s had a change of attitude as well. This is from Axios. China signals willingness to allow TikTok sale.

 

Before, as you know, they turned 180, basically. While a few interested buyers have emerged, China’s unwillingness to let bydance TikTok’s US arm has tied the company’s hands in the face of a US law that required such a sale to avoid a ban. So perhaps China is coming to terms with the negotiation.

 

Does this change of attitude mean anything, or is this just business? I mean, it may mean two. Look, again, to me, the proceeds from a TikTok sale are relatively de minimis if you’re China, right? They matter to bydance shareholders, but they don’t really matter to the Chinese government. And if bydance takes the marching order from the Chinese government, that’s what prevails.

 

Look, China may say, you know what, we want to extend an olive branch to Trump. We want to do something that demonstrates some goodwill, and so therefore we will cooperate here. Or they might decide at the end of the day that they want to hold it back.

 

Either way, I think the decision is going to be geopolitical, not business. Bradley, let’s talk about cryptos now. Donald Trump has announced that his administration is looking into a potential national digital asset stockpile.

 

So we’ve moved on from the Bitcoin strategic reserve that hasn’t been signed into order yet, but now people are talking about a crypto stockpile, not just a Bitcoin stockpile. Trump orders administration to evaluate potential for national digital asset stockpile. What could be in this digital asset stockpile, if we were to speculate? Yeah, I mean, you would imagine it’d be something like an index fund of the best tokens, right? So it’d be a lot of Ethereum and Bitcoin, but there’s a lot of other stuff too.

 

You know, I would be somewhat skeptical if the Trump meme coin were in there. That would seem fairly corrupt. But you know, and look, ultimately to me, this comes down to a broader worldview, which is the reason you’d want to have a digital stockpile of crypto assets is if you really think that the world could come to an end, and that ultimately a lot of the systems of banking and government that we have won’t hold up over time, then you probably, say when people want to stockpile gold, you might want to stockpile crypto, right? If you think that ultimately order will continue to rule, then it may be a waste of money and you may not need it.

 

There are lots of reasons to fear the future, right? Everything from climate change risk, although Trump doesn’t seem to believe in that, to potentially the negative outcomes of AI, to nuclear risk, which only grows to pandemic risk and bioweapon risk, which I think only grows as well. So there’s an argument to say if those things come to pass, we’ll be really happy that we have this stockpile. Or there’s an argument to say, you know, those are all sort of Cassandra-like cries for help that aren’t real.

 

And fundamentally, you know, any U.S. major purchase of the assets is just a giveaway to the industry from the taxpayers to reward people who politically supported Trump. Are you bullish on any particular type of crypto industry this year? Is it going to be payment systems, stable coins? Is it going to be layer twos? I mean, what’s your outlook for cryptos? Yeah, I mean, I think on one hand, I think all of them in the sense that all of them have had a hard time even getting sort of basic guidance out of government from what they can and can’t do. So just having that alone, and that’s, I think, if you look at the crypto executive order, what most of it sort of is geared towards will help.

 

But yeah, I think stable, this could be the year of stable coin, right? You know, we saw FIT21 come out of the house, I think will come out of the Senate. We saw stable coin mentioned prominently in the executive order. You know, Circle is certainly looking to, you know, IPO as soon as they can.

 

So you put all that together. And if I had to pick one, that’s what I would pick. Okay.

 

You wrote a piece about this on Substack on your Substack that Trump Crypto Task Force makes sense. A few minutes ago, Donald Trump signed an executive order creating a crypto working group tasked with developing a clear regulatory framework for digital assets, including stable coins. I mean, this goes back to your point about stable coins.

 

What is this task force going to do? What is the ultimate impact on the crypto industry? I think it’s very significant, right? Because this ultimately sets the direction for the federal government. And crypto is an industry that is regulated almost exclusively by the federal government and not state and local. So if this ultimately has the convening power that includes Treasury, SEC, CFTC, White House, National Economic Council, and there are policies that are coordinated going out across all of them, I think that’s a really big deal.

 

Bradley, excellent. Let’s finish off on Tusk Ventures. Tell us about what you’re focused on this year.

 

Yeah. I mean, I think, you know, we’re always looking at seed and series A companies. But for us, you know, we’re a little different than most venture funds is, you know, we’re looking at the same thing as other funds, the founder, the TAM, the idea, and it could be AI, it could not.

 

But where we’re different is we ask ourselves, A, is there a kind of gating regulatory issue or opportunity that if it were solved, could really drive growth and valuation? And B, if so, can we help solve it? So for example, we invest in FanDuel, then we ran all the campaigns to legalize daily fantasy sports betting on the country. We invested in Lemonade, we got their insurance license, it’s invested in Roman, legalized prescription via tax, and so on. And I started doing all this with Uber back in the early days, and I ran the campaigns to legalize ridesharing.

 

So, you know, it’s really where we think there are regulatory trends emerging that create opportunity. Certainly a lot in AI, but there’s, you know, other sectors that are interesting right now from ed tech to defense, you know, digital health and everything else too. Excellent.

 

Where can we learn more from you and Tusk Ventures? Yeah, you can just find me or Tusk Ventures on LinkedIn. Or if you go to BradleyTusk.com, there’s all kinds of propaganda on there too. Propaganda of the good kind.

 

We’ll put the links down below and we’ll talk about your work next time. Thank you very much. That was great, David.

 

Thanks for having me, man. Thank you for watching. Don’t forget to like and subscribe.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button