DeepSeek lights a fire under Silicon Valley

DeepSeek, DeepSeek, DeepSeek. We couldn’t escape the headlines around the Chinese AI lab this week. The startup -- which claims to have built its models more efficiently and at a fraction of the cost of competitors -- lit a fire under Silicon Valley after releasing its R1 “reasoning” model and displacing ChatGPT as the App Store’s top app. Today, on TechCrunch’s Equity podcast, hosts Kirsten Korosec, Margaux MacColl and Max Zeff are digging into the latest news, including the reactions from both tech giants and the U. S. government, which is increasingly concerned that China is pulling ahead in the AI arms race. Listen to the full episode to also hear about: Tesla earnings -- from promises made to robotaxi updates, and questions left unanswered. Don’t worry, Kirsten brought her bingo board if you want to play along. Nucleus Genomics’ $14 million raise for DNA sequencing and analysis. The demand for fusion, and Helion’s $425M raise to build its own fusion reactor for Microsoft. Equity has a special DeepSeek deep dive on the way for those who want more, so stay tuned! Equity is TechCrunch’s flagship podcast, produced by Theresa Loconsolo, and posts every Wednesday and Friday. Subscribe to us on Apple Podcasts, Overcast, Spotify and all the casts. You also can follow Equity on X and Threads, at @EquityPod. For the full episode transcript, for those who prefer reading over listening, check out our full archive of episodes here. Credits: Equity is produced by Theresa Loconsolo with editing by Kell. We’d also like to thank TechCrunch’s audience development team. Thank you so much for listening, and we'll talk to you next time. Learn more about your ad choices. Visit megaphone.fm/adchoices

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Anthony Ha?

  This episode of Equity is brought to you by Schneider Electric. Schneider Electric, your energy technology partner.

• 0:17Kirsten Korosec?

  Hello and welcome back to Equity, TechCrunch's flagship podcast about the business of startups. Today is Friday, January 31st. I'm Kirsten Korsak, Transportation Editor at TechCrunch.

• 0:29Speaker 02

  I'm Mark McCall. I'm a Senior Reporter at TechCrunch. And I'm Max F., another Senior Reporter at TechCrunch.

• 0:35Kirsten Korosec?

  I don't know about you all, but I was so happy this week to, once again, explore the fun world of Elon Musk through the eyes of Tesla or through the sort of foundation of Tesla. We had earnings and, you know, it's been a minute since I've enjoyed a Tesla earnings call. Were any of you tuning in at all?

• 1:00Speaker 02

  I was listening in a bit. It's fun to see Elon be Tesla CEO. Like, I agree. Like, you see him in all these different roles and then you see him doing this and it's like, oh, this is the thing that we know you originally for. Like, this is what you really, like, brought you up here.

• 1:16Kirsten Korosec?

  So are you, I don't know if you're familiar, but so for like, I don't know, a decade or so ago, when I first started writing about Tesla, during earnings calls, we used to have a Tesla bingo card. And that would be like all the common phrases that Elon would say in, you know, an earnings call. So, "Order of magnitude" was always one. "People don't understand this" is another one. "Manufacturing is hard." "Production hell." "Bottleneck." You know, things like that. So, happy to say that he stayed on a trend and said a number of those. And also on trend, did a lot of hand-waving and theater and big announcements with zero detail.

• 1:58Speaker 03

  Are there robo taxis coming? That's what I feel like I want to hear you guys talk about.

• 2:03Kirsten Korosec?

  All right, so we do have to get to the deals, but I will say this. So in traditional fashion, Elon did drop the news that self-driving vehicles from the Tesla fleet, which importantly, aren't owned by consumers, right? These are from the company, will be operated presumably by the company. There will be this ride hailing, driverless ride hailing service that is going to launch in Austin in June. And that is where the details end. So we don't know if this will be in a constrained environment. We don't know if Tesla will be the primary operator. We assume so, but we don't know. We don't know how many vehicles.

• 2:43Kirsten Korosec?

  We know none of this, but we do know that it's going to be existing Tesla vehicles using presumably, again, an unsupervised version of its branded full self-driving advanced driver assistance system. And unsupervised is the important word there. Right now, people who have that upgraded system have to keep their hands on the wheel and have to be ready to take over at any moment if the system disengages. So this is an advanced driver assistance system. What Elon seems to be talking about or suggesting is going to happen is going to be, you're getting in the vehicle, it's driverless, there's no expectation of you take control. So we will see, stay tuned.

• 3:25Kirsten Korosec?

  Again, yeah, it's a little bit of a distraction to like their financial results.

• 3:30Speaker 02

  Just before, what were their actual financial results? Like, how is the company doing? You know, or was this another kind of distraction to make investors not pay attention to the fact that the company isn't selling enough cars?

• 3:43Kirsten Korosec?

  So I would say this, like Tesla is going through a very interesting phase right now. When you really think about this, they haven't had any new models. Mass volume models, Cybertruck is the one exception here for years now. And so they've only done refresh models. So what we saw year over year in terms of sales, specifically on their automotive business, was actually the first decrease year over year. So 2023, they had 1.81 million vehicles delivered and in 2024 is 1.78. And so as you might expect, it was reflective on revenue. Revenue was fairly flat year over year. And so then the question is, what other

• 4:27Kirsten Korosec?

  parts of the business will then, will we see growth? And we've seen growth on the energy storage side actually did quite well. But there's a lot of promises around the future in terms of AI, autonomous vehicle technology, optimist, you know, the robotics piece. And that is all really in research, testing, and very, very much dipping the toe in stage. It's not a volume business by any stretch. There is no sort of optimist robot that is being sold to customers right now. And Elon and Tesla have had this track record of being able to convince investors and sell them on the vision of the future. Whereas honestly, any other legacy automaker that would do this

• 5:11Kirsten Korosec?

  would have been crushed, like in the stock market aftermarket close. So people believe, you know, believe in his vision and that he's going to pull it off and they get excited about it. And so they've stuck with them. So that's my Tesla earnings in a nutshell. We should probably get on unless you have any more final questions, Max, for me.

• 5:30Speaker 02

  I just need to, I mean, they've been promising this for so long. And like you've covered Waymo try to do this in many cities and now Tesla is going to do it with consumers' cars. I feel like there's so many questions about like, what if you have? What if I leave a sweater in the back of my car and then it goes off to like pick someone else out like I, I, and there's a whole app that needs to happen. I'm just curious from your perspective, like, what is the thing that could really hold them back from actually doing this? Or do you think these are all small details and the self thriving part?

• 6:04Kirsten Korosec?

  Oh, no, these are big details. These are big details. And this is, this is a classic Tesla, like, we'll just figure it out, folks. But you're right, we do need to move on to deals. But I will say this, there is a lot of details to work out. Clearly Tesla is holding some of those details back. I have big reservations and questions about their ability to do so, because to your example, Waymo and others have been working on this and testing these things for a long time. That's not to say that it is impossible, but I think that what will happen is we'll see a very limited, very small ride hailing service sort of geographically

• 6:44Kirsten Korosec?

  in Austin, in which really the testing happens, and it will be opened up to the biggest fans of Tesla first and employees. So that's my prediction. But we have two deals of the week, one from Nucleus Genomics and another from Helion. And then we're going to focus on one theme, and that is going to be all things DeepSeek. Margot, please tell me about Nucleus, because I find this company so interesting.

• 7:14Speaker 03

  It's so interesting. Definitely fits into like the kind of longevity themes that I think we've kind of touched on here before. It's essentially 23andMe, but instead of your saliva only getting tested for like ancestry results, it gets tested for, they sequence your whole genome, and then takes that data, runs it through what sort of its model that compares it to a million studies linking genes to a million different things. You get a dashboard, it shows you things like your risk for certain cancers, certain diseases, and then more controversially, traits that are more complicated, like it's not just one gene and you have it, like IQ, which they offered in a service called Nucleus IQ. They just raised $14 million. Obviously, they're backed by Founders Fund. This is

• 7:58Speaker 03

  a very Peter Thiel-esque company to back. And yeah, it raises a million questions about future preventative medicine, about like the ethics of it. Max, what's your gut reaction to this company?

• 8:11Speaker 02

  I mean, I think it's terrifying. I think my gut reaction is an awful one. I think it sounds like Eugenics, where it would allow people to customize, you know, what kind of kids they have. And I think that raises like so many just ethical concerns. But at the same time, I mean, the fact that this technology is becoming more mainstream to and the costs of it seem to be going down a lot really makes me think that this could be something that could maybe help a lot of people. I mean, Margo, what are people really using this to kind of figure out about their kids? And is there any way that this is actually just a healthy thing and not like, I want to make my kid have the exact hair and eye

• 8:56Speaker 02

  color I want?

• 8:58Speaker 03

  There's actually better startups for that, if that's what you want to do. There's a startup called Orkin, where they do like very extensive genetic testing on your embryos. And you can select the best embryo for like whatever you are optimizing for. I actually haven't talked to like parents that have used Nucleus to do that. Still like a fairly new company. But it's really, I mean, to your point, like, it's really interesting, because I think there's a line that has kind of existed where I think we can all agree that for genetic diseases, where it's like, you have this gene, you are going to get Huntington's disease. That's knowledge I think everyone wants to have. And then you get into these things called polygenic traits that are much more complicated, like

• 9:38Speaker 03

  IQ, like your personality, like mental illness, like Nucleus will tell you theoretically what your predisposition for like schizophrenia is. There's not like one gene that gives you schizophrenia. So it's a weird future, but also to your point, in 2007, this cost about a million to do. And now it's $400, the Nucleus. How is that not just going to become a standard part of like preventative medicine?

• 10:05Kirsten Korosec?

  So one, are you going to take this test? But more serious question, I guess, is are there any implications for medical research? Like how maybe these results are shared? Or have you thought about that at all?

• 10:19Speaker 03

  Definitely. One, if tech rental expense it for me, I will happily do it. And then two, another interesting part of this is that, and I can't totally speak for, I did not get into like the nitty gritty of the data safety angle of it, which is obviously very relevant. 23 me famously had their data leaked and that's pretty awful. So yeah, those are questions that the CEO is going to have to keep answering for. But it is interesting because the more people that use nucleus, like the better the models will theoretically get, right? With all these people handing over genetics. I mean, I was talking to Delian, who's partner at Founders Fund, who invested in this company. And he had a quote where he was like, yeah, and like one day, maybe we'll get like the

• 11:00Speaker 03

  phenotypes of these people, like, you know, what color hair they have, and then all this other data and their IQ. And we can make these really extensive models. Depending on your views, I think that's either an exciting future, a scary future. But regardless, I think we are on like a crash course to a future where we all have way more information about our genetics.

• 11:24Speaker 02

  Yeah. I mean, I think it's kind of paints a frightening kind of future, but we actually have another deal that also paints kind of a similarly kind of maybe dystopian future. Maybe this is actually a really cool idea. I don't know. I think I want to hear your guys' thoughts on Helion, which is our next deal. Helion just raised $425 million to help build a fusion reactor for Microsoft. I mean, importantly, the startup is hoping to build the first fusion reactor to generate electricity. I think that's like an important note here is that, you know, they're trying to build this reactor for Microsoft by 2028 to help them power AI.

• 12:05Speaker 02

  But this is a system that has not really worked in practice to practically generate electricity for data centers or anyone yet. But, but they're raising big money and they're backed by Sam Altman and Duskin Moskovitz. So I'm curious to get your guys' take on this. Maybe Kirsten, I was curious, like, What is kind of the the big challenges here? I know you helped Tim edit the story.

• 12:28Kirsten Korosec?

  Yeah. So first of all, kudos to the segue between genetic testing and building a fusion reactor. I've never seen that happen before. Being scared, that's the thing.

• 12:41Speaker 02

  I'm scared. I'm scared. That's the takeaway.

• 12:43Kirsten Korosec?

  So so I think the interesting thing about this company, besides, by the way, a lot of its existing investors, really high profile folks and their post-money valuation, which right now is north of $5 billion, is their approach to fusion power. So I'm not going to get into extreme details here, but there are two basic approaches to fusion power and Helion is not doing either one of those. They are doing something called field reversed configuration reactor. And the important thing to note here is that they are trying something new. And this is going to take a long time and a lot of effort. Like this isn't just a done deal.

• 13:24Kirsten Korosec?

  So more money will need to be raised here. To me, what I find really interesting is there's been a lot of discussion around fusion reactors for a long time, but they're clearly gaining traction now because the reality of power needs for data centers, specifically around the rise of AI, is really coming to fruition here and it is sparking a ton of deals. And it is also encouraging investors to seek out alternatives to renewables, natural gas and other things that, you know, probably won't actually be functioning for another five years or so. So to me, that's like the big takeaway here is like,

• 14:06Kirsten Korosec?

  this is very much being driven by interest in in just like the appetite for power needs. But, you know, I should say like Heliana didn't just like spring up out of thin air. They've been here for a while working on this.

• 14:18Speaker 02

  Right, and they're one of the more promising companies working on this. But Margot, I'm curious, you know, Sam Altman backs this. Microsoft is going to use this energy to power OpenAI's models. We got to peek into kind of OpenAI's economic zones, this kind of vision that they laid out. Can you talk about how this ties into it?

• 14:38Speaker 03

  This is crazy. I have like very few details for you, but OpenAI released this big, it was like a 15-page like plan. And in that plan, they have one little paragraph that described its recommendations to the government, which included AI economic zones created by local, state and federal governments together to speed up the like AI infrastructure, including nuclear reactors. I've talked to nuclear founders before that are really into like states being able to regulate their own nuclear reactors. I have never heard of this idea of AI special economic zones where Sam Altman can regulate the the nuclear reactors. It's fascinating. And it was one paragraph, which is crazy.

• 15:21Speaker 02

  Right, I mean, I think we've seen a lot of, you know, talk in the Trump administration with Elon coming in with Doge. And now it's Sam Altman about just like, and Marc Andreessen, everyone in tech seems to be just talking about like, we need to build faster and take down these regulations. And I mean, nuclear is such a good example of a place where there's so many regulations to building these huge systems. I mean, Margot, what could you say about like the shift away from these heavy regulated environments towards, you know, building infrastructure like this? Like, like, what are your conversations with founders and investors about that shift?

• 15:56Speaker 03

  Yeah, I talked to one founder in the space and they talked about how like a question they always ask people is, if I put you on like a desert island and I gave you all the resources in the world, like, how quickly could you build a nuclear reactor? And if they say anything like over a year. He just like, he's like, oh, they don't, they're not my kind of guy, which is really interesting. And the point that they're trying to make, right? Is that if there was no regulation and you just put a bunch of people that knew how to build this stuff, they could actually build it really fast. I don't know if that's true, but it's how these guys are thinking.

• 16:30Kirsten Korosec?

  Yeah, well, I would say that regulations around nuclear reactors are a good thing. And that is something that I personally support. But you know, this conversation keeps throughout the show so far, circulating back to discussion about AI, but also a little bit of American dynamism and bringing AI home, US-based, I mean, and that is the perfect setup to our theme, which we will get to right after the break.

• 17:03Anthony Ha?

  Every company today is looking for an edge. Better insights, faster decisions, greater resilience. But as AI drives new growth, it also creates new demands on energy and infrastructure. Schneider Electric is connecting energy, automation, and digital technologies to help organizations make smarter decisions and operate more efficiently. Because the future won't be shaped by AI alone, it will be shaped by energy and industrial intelligence. Schneider Electric, advancing energy tech.

• 17:29Speaker 02

  So DeepSeek, I mean, just to, you know, catch people up on the last week's madness very quickly. I mean, DeepSeek is Chinese AI lab, and they have an app that reached the top of the Apple and Google App Store, Google Play Store charts this week around the world in like 100 countries. It was the number one app. And really, I mean, they've been releasing models for the past couple of months that are impressive, but it was this week that something just clicked and everyone lost their mind. Nvidia's stock tanked 17% on Monday. A lot of American tech companies' stocks plummeted. And really, the fear was around, DeepSeek created a model

• 18:11Speaker 02

  that really competes with OpenAI's O1 reasoning model, which is their cutting-edge model right now. DeepSeek claims to have built this model, which is called R1, for Very little money. They claim to have just used a small amount of money. They claim to have done this very quickly and they're selling this model, offering access for a fraction of the cost that OpenAI charges. So there's a lot to unpack here with this. First off, I just want to say that there's a lot of allegations right now that DeepSeek is maybe saying it did a little bit more than it actually did here. There's a lot of skepticism in the AI community about

• 18:50Speaker 02

  the amount of chips that it says it used. They claim to, for a previous model V3, they claimed to have only used 2,000 GPUs, but it's been reported that they have over 50,000 GPUs. There's also allegations from OpenAI that DeepSeek distilled these models from an OpenAI model, which basically means they use OpenAI's models to train there. So I mean, there's kind of a lot of skepticism around DeepSeek's claims, but that didn't stop the AI world and the technology world from absolutely freaking out this week. I mean, this was still a pivotal moment regardless of, you know, those two allegations around training compute and just how they trained the model altogether. But Margot, I mean,

• 19:33Speaker 02

  obviously, we can't ignore that this is a company from China, so that's going to make everyone freak out 10 times as much. What is, have you been talking about with people in high roles, in government advisory roles, just kind of, what have you talked to them about China's race against the US on AI?

• 19:52Speaker 03

  Yeah, they, I mean, they view it like existential. Being hawkish on China is just like the new norm at this point, and I think the Silicon Valley people are no different. I mean, you had Alex Karp writing in, I think it was the New York Times, calling this like our Oppenheimer moment, being like, the US needs to be the guys with AI. They, yeah, they view it totally as an existential threat. They're incredibly hawkish. I also just love the irony of OpenAI getting their IP stolen. I think that's kind of funny.

• 20:19Kirsten Korosec?

  That is great. Max, do you think it is an Oppenheimer moment in the AI world? Or is this being blown out of proportion?

• 20:27Speaker 02

  So I've heard this be called an Oppenheimer moment. I've heard it's called a Sputnik moment. I've heard this called all different kinds of just insane technology moments throughout history. Here's what I'll say about it. I think that the tech industry, a lot of people in the tech industry have a really big incentive right now to be calling this a big moment. Because I think that if the tech industry can get American government officials to worry about Chinese AI, I think that's going to promote a lot more investment in AI right here. So that's just what I'll say on the incentives for these people to be saying this. Do I actually think that this is a big moment for the AI industry outside of why these people are saying it?

• 21:07Speaker 02

  Yes, I do. I think that it represented, one, some really big gains in how people are training AI models. I talked to a lot of researchers who said that the actual methods they used did move the needle. I mean, it just did. And also, you know, the fact that this is an open, I mean, it's not open source technically, but it's an open kind of model. They released the model weights and more information about it. It's a big deal. It helps more people kind of learn how these advanced AI models work. I think that this has really changed how a lot of people are thinking about the AI space right now. I think it's lit a fire under the big tech companies who are creating this stuff. I mean, I do think that people are certainly incentivized to be calling this a big moment to spark kind of fear.

• 21:51Speaker 02

  But I think it truly was in some ways.

• 21:53Kirsten Korosec?

  Yeah, when you say lit a fire, I think that that was very appropriate. And, you know, I started thinking about, actually, Margot, I was thinking about you the other day when I was thinking about this, because when I think of China Hawks, I think of Margot. But there's some interesting tension that's being created. I think that there's, A, you're right, Max, and that there's sort of this momentum now. In terms of the investment community encouraging and putting more money like no one's waiting around for this, they weren't really before, but it was like this wake up call. But I also start to see some conflicts, particularly on the China front. And I was thinking about Elon and I was thinking about Margo.

• 22:33Kirsten Korosec?

  And I was thinking, you know, Elon is, you know, pretty tight over in China, also sitting in the White House, also very much connected with a number of the people in Silicon Valley who are China hawks and also feel like there's a sense of urgency around development around AI here in the US. And I'm wondering if you are hearing on the ground in DC, people mentioning this, like this point of tension at all, either around like conflicts with Elon, but also just general points of tension between where money goes. And we brought this up actually in a show a couple of episodes ago about how some founders are asking where investments are coming from.

• 23:17Kirsten Korosec?

  So I'm wondering if you're seeing more of that or hearing more of that.

• 23:21Speaker 03

  I've always heard a lot from defense tech founders. I also think that there's been a lot more like posturing amongst VCs to like not take money from China. I had a really interesting conversation with, I bring up Jacob Helberg on this pod a lot because I think he's a very important person. He's a very important bridge from Silicon Valley to DC. And I was like walking around with him in the hill and I asked him about, at the time, Sam Altman was doing a lot of business or like taking a lot of trips to like Saudi. And I was like, well, what do you think about like your pal Sam taking these trips to this country that obviously has horrible human rights and Iraqi history with the US? And he was like, I think that we need to put the regulations and guidelines in place and Sam Altman will follow them. You know what I mean? He was like businessmen are going to do what they have to do to make money and it's

• 24:06Speaker 03

  up to the government to put the, oh my God, what's the like bumpers?

• 24:10Kirsten Korosec?

  The guardrails, the guardrails.

• 24:12Speaker 03

  Yeah, the guardrails.

• 24:13Kirsten Korosec?

  Yeah, yeah.

• 24:13Speaker 03

  The guardrails will talk for them.

• 24:15Kirsten Korosec?

  Okay, but you know, visiting a country in which there is some conflicted relationships and issues with is different than, you know, in Elon's case, actual big factories in China. And I will note that Tesla, which is run by Elon, and he's the largest shareholder, was the first company to be able to build a factory without a joint venture with the Chinese, you know, state government. That was huge. And so he has a relationship there. And I'm just wondering, you know, not to get too far away from the AI conversation, but I do wonder if that is going to become a problem and a conflict, particularly because his hands are in everything, including XAI, right?

• 24:58Kirsten Korosec?

  So I wonder how that relationship is gonna come about.

• 25:02Speaker 03

  Totally. My husband's like a foreign policy writer, and they did a great column. And their whole thing is like restraint abroad. And they did a column being like, what if Elon Musk is like our savior? Like, what if he's the only thing standing between us and war with China? Which like, he is, as far as I can tell, the only one from the Silicon Valley camp that's really made inroads in DC that like, is probably advocating a lot for us to go.

• 25:25Kirsten Korosec?

  For China.

• 25:25Speaker 03

  For a little easier on China, for sure.

• 25:27Kirsten Korosec?

  Right, right. Yeah, there's some vested interest there, but, you know, again, we don't really totally know what he's thinking on that topic. But, you know, actually, Max, not to pivot too far away, but you will be talking to someone who could give us some insight into what's happening on the AI front, and that's Databricks co-founder.

• 25:46Speaker 02

  Yeah, talking to Jan Stoichek. And I talked to Jan a decent amount. He is one of the, you know, very local advocates of American open source. I mean, of course, Databricks is a company that would benefit from America investing more in open source, so he's biased. But I think that he can offer kind of a good perspective on, you know, the arguments that a lot of Trump officials are really hearing right now from the, he's one of the experts who is talking to the Trump administration a lot about this stuff. So it's interesting to hear how they're pressing government officials to really compete with China on this stuff. It's becoming a much more popular take between these Silicon Valley investors and and government officials. So you should listen to that interview.

• 26:31Kirsten Korosec?

  Okay, great. Well, before we wrap up, I'm just wondering if Max, you can just give the audience just a sense of where things go from here with DeepSeek specifically, because you're pretty deeply involved in the coverage there. What should we be looking for in the next couple of weeks? Or are you expecting any more news on that front?

• 26:51Speaker 02

  I think the things to watch are, I think OpenAI seems like it is positioning itself to bring a claim that DeepSeek has inappropriately used its own models. They kind of released a statement alleging some misbehavior from DeepSeek and using its models to generate data. I mean, the dirty secret is that a lot of AI companies use each other's models to make their models better. We've reported on it, others have, but I think they're going to highlight that a lot because it's a Chinese AI company. But also, I think that we need to pay attention to DeepSeek now. I mean, I think that's, that's just a main takeaway here is that this is a lab worth paying attention to now. And Microsoft's hosting them on Azure now. A lot of companies are, you know, kind of buying into it.

• 27:35Speaker 02

  So I would say keep an eye on them, look out for more model releases and look out for how OpenAI responds to this. So I don't think they're happy.

• 27:43Kirsten Korosec?

  All right, well, on the next episode, I'm sure that you will have another update for us. We are out of time, so Equity will be back next week. But until then, you can find Equity under the handle at EquityPod on X and Threads.

• 28:00Speaker 04

  Equity is produced by Teresa Loconsolo with editing by Kel. We'd also like to thank TechCrunch's audience development team. Thank you so much for listening and we'll talk to you next time.