DeepSeek's AI Brings Tech Rout
42:42 Share

89% of business leaders say AI is a top priority, according to research by Boston Consulting Group. The right choice is crucial, which is why teams at Fortune 500 companies use Grammarly. With top-tier security credentials and 15 years of experience in responsible AI, Grammarly is how companies like yours increase productivity while keeping data protected and private.

See why 70,000 teams trust Grammarly at grammarly.com slash enterprise. Meta's open source AI, available to all, not just a few. Here's Dr. Cal Clark of Zaron Labs. Meta's open source AI model, LAMA, helps us collaborate with universities to help radiologists catch more errors. Learn more at ai.meta.com slash open. Bloomberg Audio Studios. Podcasts. Radio. News.

Live from New York, I'm Caroline Hyde.

And I'm Mike Shepard in San Francisco. This is Bloomberg Technology. And we have special coverage of the Chinese AI competitor DeepSeek and its impact on tech names, the broader markets at large. Let's look at the Nasdaq currently having its worst day in more than a month as we question the need for compute power to get us ever more efficient, ever more specific generative AI.

We're down 2.6%. Move on and look at the individual names because this is when we question NVIDIA, the compute power, the necessity for ever more efficient chips and the cost of them. NVIDIA has its worst route in history in terms of market capitalization ever on the market. We have never seen almost half a trillion dollars wiped out from one single name. This is global. ASML, chip equipment maker in Europe, down almost 6%.

But Hong Kong stocks do well as we start to see AI as a real champion for China. Let's talk about it all with Bloomberg Intelligence Senior Analyst Manneep Singh. He joins us some more. How does this change the game for LLMs in the US, for chip equipment in the US?

Yeah, look, I mean, we look at it from a stack perspective and with generative AI, so far all the value was captured by the chip makers, the semi-cap equipment guys, and the LLM companies like OpenAI, Anthropic. And the software guys were really missing out because everyone was like their margins will be compressed, they have to pay more for chips, pay for this LLM layer. Well, guess what? If the price can come down and what happens

DeepSeq has shown us is the price per token, price per query can come down drastically. Then suddenly I think the power goes back to the software guys because now they can incorporate AI in everything they want, in their app, in their website, whatever

functionality they have, they can add a layer of AI without having to pay too much and hurting their gross margins. And so I think this really is a big development. And every cloud company, whether it's Alphabet or Microsoft, they'll be looking to leverage their compute more efficiently because they will benefit from the cloud revenue on the infrastructure side, but also on the application side, which is why I think Meta raised their CapEx on Friday.

Mandeep, how does this undercut NVIDIA in particular in its argument for advances like the Blackwell product that it is pushing out now and what may come after it? Is it becoming much harder for Jensen Huang to sell that to Microsoft and its other customers now all of a sudden?

Look, I mean, last three months we were debating about the scaling laws and how big of a cluster we need. We were talking about 100K GPUs in a cluster and possibly 1 million GPUs in a cluster. I think this puts into question mark

how big of a cluster you need actually for training your LLM. And look, all these LLM companies have to keep training their algorithms. You know, this is not a one and done thing. So you need compute capacity, but at the same time, NVIDIA was accruing all the benefits in terms of pricing because of the, you know, advances in their chips. So if you can do more,

with even the older version of their chips and you don't need a cluster of one million GPUs connected together, then probably I think you can do more in terms of the distributed infrastructure. And I think that's what a lot of people are trying to figure out today.

B.I.'s Mandeep Singh, thank you. Let's bring in Tony Wong, portfolio manager of the Science and Technology Fund at T. Rowe Price and a key NVIDIA shareholder for more on this. Tony, we've got a big week and month of earnings coming up. How does this change the way you are looking at companies as they prepare to deliver their results?

Yeah, I think it's a great question. And if you look at the kind of trends in technology, generally when things become more efficient, it increases demand and there's more ubiquity. And so it wasn't that long ago, you know,

you know, a few months that we were all debating that like the cost was too great to deliver. And so where is the ROI? And so I actually feel like this is TAM expansionary for the market long-term. I think that Mandeep, you know, obviously has highlighted what the market is kind of talking about today. I think that's a very valid concern, but you know, Javon's paradox essentially is that when things become more efficient, consumption goes up, demand goes up. And so I think that's, that's necessary. That's healthy. Yeah.

I do think that there's a little bit of sensational reporting perhaps on Twitter and just in terms of how DeepSeek only took $6 million to train. I think that's a nice marketing number. But if you dig in, that doesn't include R&D costs, experimental runs, and the market is

The model is a lot smaller than what a chat GBT is, and it's only text-based. So there's a lot of reasons to think that this is not exactly a kind of direct comparison. They do have a lot of great breakthroughs in the efficiency of the model, obviously, but I think it's good for the industry and it's open source, and the players will, you know, the community will adopt the innovations and then further kind of pursue AI

in a bigger way, I think. To that point, let's just reconfirm to our audience what it is that DeepSeek has done. Because Bloomberg Intelligence, Bloomberg News has been reporting on DeepSeek for months. Bloomberg Intelligence ranked it as the seventh most powerful LLM all the way back in June of last year. But now it gives us the fact that its latest R1 model is as sophisticated, many would say, on the internet and other experts out there, is as sophisticated

as that of Anthropix latest model or of OpenAI's latest model, but they've done it at a fraction of the cost. But 6 million is what is reported to have been cost. We haven't been able to delve into the veracity of that number. Tony, when it comes to the rush to the bottom, ever more cheap,

of applications of generative AI? Do we therefore see Microsoft have to cut its prices? Do we see ultimately revenue and profitability of these LLM providers having to become less?

Yeah, I think that's a good question. I would say that, you know, for one, like we don't know what OpenAI and Anthropic are going to release next. And a lot of times, like, you know, there are so many innovations, the market is moving very quickly. So you can very well that they've got something that's really exciting to be launched. So that's number one. I think like number two is that a lot of times, you know, this is a fast follower press model that,

has a ceiling in terms of what it can do versus larger models. And it's, you know, open AI is going to be multimodal with our model. And, you know, I think there's a lot of innovation that will happen. And then I think that like at the end of the day, like, you know, there's a lot of kind of ecosystem building enterprise aspects that also factor into it. So I think that the market for LLMs can be more diverse than people are giving credit for.

Is this a buying opportunity, Tony, to stack up on Nvidia even more? I think long term, you think about the technology trends. And I don't think that just because there's one model that is more efficient that people are going to say, OK, we're just going to stop here. At the end of the day, everybody's racing towards hopefully AGI and looking to continue to invest. I think that you saw Mark Zuckerberg, even though he knew about DeepSeek,

since it was released over Christmas, you saw that he actually increased CapEx spend. And so I think that it's not just the model, it's also the compute as well. And so I think that as there's more demand of the app layer, this is actually good for filling up the compute capacity that we're building.

Tony, you talked about Mark Zuckerberg and his announcement last week on CapEx. Is there any thought to the way these companies should revisit those spending plans? Are they going too big? Is it too big and maybe not efficient enough?

Yeah, well, naturally, absolutely. I think that everybody's going to their engineering teams and saying, like, hey, what can we do? This is an open source model. Let's take it apart. Let's figure out how we can incorporate the best. And, you know, as costs come down, like, I think that actually increases the number of use cases. And people say, like, oh, well, I wasn't able to afford perhaps the latest, like, equipment because...

of the cost of doing so. We're not sure about the ROI. And so this, I do think that it's absolutely natural, but I don't think that companies are going to say like, oh, like, let's just stop here. And because there are so many like potential benefits, the kind of goal at the end of the day is pretty grand.

Tony, I wanted to follow with a question about the potential policy responses from Washington. Given the hawkish mood toward China with the new administration, do you see any risk of a response that would pose a challenge to the companies that you own and are taking a look at?

Yeah, well, I think that definitely there might be some geopolitical aspects of what's going on this week. You know, I think President Trump announced the Stargate project. And so this was perhaps, you know, DeepSeek is a response to that in that, you know, the question is, like, can you build AI more cheaply and openly versus like building a bunch of AI infrastructure? I think that, you know, my view is it probably you need both.

And each country will continue to race towards AI capabilities. And I think there could be some export restrictions more along the way, possibly, but it does show that

you know perhaps like restricting isn't particularly helpful because then you actually bore innovation out of scarcity. I think that what we've seen is that there is a lot of innovation to be had and you know I think you know competition is good and you know at the end day like you want an open market. Who will win from your portfolio from that innovation or potentially from that shift from hardware more to the software side?

Yeah, I think it's a really exciting time because I do think that at the app layer, this improves kind of the ROI, obviously. And, you know, anybody that has like a really sticky customer base and large kind of apps that they have a lot of data from, I think that's good. You know, I do like, you know, the AI infrastructure still. I think that there's, you know, still steady kind of consistent demand there.

And the history of technology would say that people take that innovation and that improved performance and they put it to good use for developing the next-tier technologies. Tony Wang, Portfolio Manager at T. Rowe Price, one of the most significant NVIDIA shareholders. We so appreciate it. Coming up, we'll be joined by former White House Chief Information Officer Theresa Payton to discuss DeepSeek's cybersecurity implications and more. But there are other earnings upon us.

Keep us afloat with AT&T. We'll bring you what AT&T is currently doing on the market. We're higher in that particular stock. Fourth quarter results came in better than had been expected, driven by seasonal promotions for AT&T and bundled product offerings. This is Bloomberg Technology.

89% of business leaders say AI is a top priority, according to research by Boston Consulting Group. But with AI tools popping up everywhere, how do you separate the helpful from the hype? The right choice is crucial, which is why teams at Fortune 500 companies use Grammarly.

With over 15 years of experience building responsible, secure AI, Grammarly isn't just another AI communication assistant. It's how companies like yours increase productivity while keeping data protected and private.

Designed to fit the needs of business, Grammarly is backed by a user-first privacy policy and industry-leading security credentials. This means you won't have to worry about the safety of your company information. Grammarly also emphasizes responsible AI so your company can avoid harmful bias. See why 70,000 teams and 30 million people trust Grammarly at grammarly.com slash enterprise. That's Grammarly at grammarly.com slash enterprise.

Grammarly. Enterprise-ready AI. Your customers are important to you, but they won't feel that way if they're stuck messaging a clunky chatbot or waiting on hold for a representative. Estimated wait time is 25 minutes.

With Sierra, your company can deploy a branded AI agent that engages and delights customers anytime, anywhere. Sierra agents pick up every phone call and personalize every interaction. No more menus, no more hold times. And if you have an issue, Sierra's AI agents solve tough problems. Whether they're helping your customer pick out the

Always helpful. Always ready. Visit sierra.ai to learn more. That's sierra.ai.

There's no question that this could be a potential game changer. It's a game changer for the MAG7 stocks. Tech has to be okay simply because it starts to lose weight in the market and if tech is not okay, the entire market goes. We're all tied to these seven stocks and in particular, we're all tied to Nvidia. It looks like the story is China is not as far back as what people thought. They are much closer. Now China looks like a very viable competitor and a competitor that perhaps might catch the ire of the Trump administration. This has got to be

a core concern of not just the Trump administration, but all the tech universe that has moved into the West Wing. The US-China trade wars and the restrictions on chips and things like that could get even more heated, which will bring more volatility to this market.

That was what some of Bloomberg Television's guests had to say about the DeepSeek impact earlier today. This is NVIDIA is hitting session lows. Now, DeepSeek reminds us that technology lies at the heart of geopolitical tensions between the U.S. and China, with disputes over AI, TikTok, and cybersecurity. And for more on that, we're joined by Teresa Payton.

She is the CEO of Fort Ely Solutions and the former White House chief of operations officer during the George W. Bush administration. Theresa, thank you for joining us. Obviously, this is going to cause a significant impact in Washington. How do you expect policymakers there who have taken a hawkish approach toward China will respond?

I mean, this tech route and massive market sell-off, it's a wake-up call. It is for me. And it underscores the urgent need for robust policies. We have got to safeguard the United States' leadership in AI. I certainly don't want China setting the gold standard for the rest of the world for privacy, safety, security, resiliency, and ethics around AI.

I believe this is why you saw the technology advisors, Elon Musk included, and President Trump last week speaking very boldly about AI and the United States leadership position that is necessary. I mean, we didn't even have a full week before the Trump administration passed an executive order on AI. So this is very serious.

They've set AI as a priority across Washington and including from the White House itself. And yet previous measures, including export controls by the Biden administration, don't appear to have worked. So what is the formula to achieve what Donald Trump has set as a policy goal, that is U.S. leadership in AI?

No, you're right about that. This sort of this open source method in which DeepSeek was released, it's now the top downloaded free app at the Apple Store, definitely got around sort of the different protocols that were put in place by the U.S. government over the last couple of years.

And so it shows that although regulatory frameworks are incredibly helpful to set the right standard and the right tone, that sometimes regulatory frameworks don't service and they get in the way of innovation. So we're going to have to figure out

how do we go back to the drawing board and look at how we engineer AI in the United States today and ask ourselves what can we be doing differently and it's a race right now. It's a long race right now. We kind of lost one of the heats. So but you know there's time to make up the difference. With your security expertise we bring you this latest in terms of breaking news that

DeepSeek says it's subject to a large-scale malicious attack. If you've tried to download it, you can't. Currently, it seems to be offline in many ways. How do you think this will be able to be responded to? Is it right that we can't access it here in the United States?

Yeah, I think this is a real challenge. I saw that they were having, this is late breaking news, that they were having some resiliency and recoverability issues ostensibly because there was such heavy volume in traffic and interest.

But if they are under some type of a cyber attack, I think that should give pause to everybody who's thinking about testing out the app, perhaps on company data, perhaps using your own personal data. This is an untested app.

I would caution people not to put too much proprietary company information into it until there's been an opportunity to actually do something called ethical hacking or red teaming pen testing of the actual app and learning more about how is your data treated, where is it stored,

How do the algorithms work? It is open source, but things do still need to be put through their paces. - Andresa, this comes at a time when an app that everyone in the United States from a bipartisan perspective thought was a national security threat

that of TikTok doesn't get banned for U.S. users. We see this ongoing push to drive in AI, but actually perhaps pull away from some of the commitments when it comes to ethics, when it comes to safety. Is that something that you're reading between the lines right now? Yeah, I'm definitely watching this very closely. And my interest has peaked here, for starters,

where is CFIUS on this, right? So we've been upset about TikTok. We've demanded TikTok do a lot of architecture changes. They did a lot of architecture changes with Oracle. The United States still said that there were some concerns there. Where does CFIUS stand on this particular app? How do they feel about it for both U.S. citizen data as well as U.S. corporation data?

The origin story is China. It is headquartered in China. And so, again, a lot of questions remain. Are we going to allow this app to be downloaded in the United States? And is it allowed to be downloaded just because CFIUS hasn't had the time to analyze it and give a ruling on it? Teresa Payton, CEO of Fortalece Solutions and the former CIO in the Bush White House, thanks for joining us.

Let's now bring in Bloomberg's Jackie Davalos to discuss the impact of DeepSeek on OpenAI and Stargate. Jackie, we worked on the coverage of the Stargate project unveiling last week at the White House. This is not the kind of moment that President Donald Trump and his administration wanted to start this week so soon after that big announcement. Where do they go from here?

Well, from here, we really have to assess how powerful is DeepSeq? What makes these models so threatening to the ones that we are building here in the United States? One of the things that we can look to is the fact that this R1 model

model family that is really making a splash. They're open weights. What makes this so surprising is the fact that many open weight models have actually kind of been lagging their closed source peers like OpenAI, like an anthropic

And so what the United States government really now has to assess is what are we missing here? How are they able to get past these export controls? It might not even need them. It seems like these models are really efficiently run because they're far more computational in nature than what we might see from an open AI. So this kind of turns the whole premise that

chips were really going to be the key here to getting ahead in the AI race, this really kind of flips that. Necessity is a mother of all invention. They had to do it. They had lack of access to the most sophisticated chips coming from Nvidia. And we understand from Blooming Intelligence that this is about novel mixture of experts, architecture, right? This is how they lower the cost. This is how they bring computational power. It's a different form of running the models.

but others have been bringing it on too. Can you tell us a little bit about how they're managing to compete and win out when it comes to mass challenges, when it comes to reasoning versus the lightest out of Clawed or indeed OpenAI?

You're absolutely right, Caroline. It's no surprise that the computational framework here really has gone a long way. It's a startup that has its roots in a quantitative hedge fund. So this is really their bread and butter, at least it used to be. But when we look at what the numbers actually show, their performance across several benchmarks, the ones that stand out to me are in math and coding. These are the ones, especially when you think about how widely used it is in the coding community, the fact that it outperformed a

and open AI on some benchmarks, DeepSeq here really is a player that is contending with these incumbents. But kind of past that, you're right, there is a difference here when it comes to is this general purpose? How big are these parameters? So far, the latest model has 671 billion parameters

compared to OpenAI. So it has models that are smaller, that they can handle customization, more specific things quicker, can also give you how it's coming to that, similar to an OpenAI's reasoning feature. Jackie Davlos with the detail on the technology. Welcome back to Blue Mag Technology. I'm Caroline Hyde in New York.

And I'm Mike Shepherd in San Francisco. We must get to these markets. The deep seek impact royals across the board. A clear present concern about the cheap

offering coming from China. Generative AI, are you able to compare with that of OpenAI or indeed of Anthropic? But with just $6 million spent on the latest model, that's what we question. NASDAQ 100 off by 3%, worst day in at least a month. But Bitcoin also following down. Risk assets sell off hard. Let's see what drags down the NASDAQ 100. The benchmark crumbles as Nvidia loses 16%. More

than half a trillion dollars has been wiped off in terms of market cap that is a record in terms of a suffering of a single name apple on the higher side earnings of course are going to come thick and fast this week the 30th is when apple comes but the most downloaded app on your app store right now is deep seek which is currently suffering they are saying a attack bloomberg intelligence analyst anna ragrana is here for more on the ripple effects

Did you see this coming? DeepSeek, of course, has been written about plenty by your colleagues in China. They said it's one of the most powerful LLMs and indeed have singled out how cost efficient it is. But should it have such an impact on the market capitalizations of U.S. companies?

Yeah, I've been surprised about the massive impact it's having on the chip guys. But at this point, I think everybody is questioning that how is it that they can run the model at such a cheap price while the U.S. company needs so much infrastructure. So I think that that's the big question mark today. And again, as you said, it is surprising. The follow-through impact is if you look at some of the software names, they are actually responding positively to that because

down the road, it means that AI adoption rate could actually accelerate. So lots going on, lots to digest today. So Anurag, will demand for AI products make this in the long run more of a blip than anything else?

Yeah, but that's, I mean, that is, that was eventually going to happen anyway. The question is, are we actually at that faster pace today? So you look at it, you know, Microsoft's Office Copilot product. Let's just take that as an example. $30 per user per month. They did, you know, come up with some consumption-related stuff as well. But...

The adoption rate for that product, in our view, has not been at that same rate as it should have been if it was only, let's say, you know, Teams edition for $5 a month or $7 a month. So it does have an impact on the adoption rate of people. You know, Adobe selling its Firefly services, ServiceNow. I mean, you look at all these companies that are, you know, spending a lot of money to embed these features. And if they can do it cheaply, that means faster adoption, you know, in the long run.

Anurag Rana of Bloomberg Intelligence, thank you. Let's now bring in Jordan Klein from Mizuho Americas for more. Jordan, thank you for joining us. We have to ask, is this the beginning of the end of the great AI trade or maybe not so much? Well, it's a beginning of probably a consolidation phase and some profit taking for sure. But I don't think it's the beginning of the end of the AI trade. I just think we're up a lot.

after two massive years of outperformance in both tech and these AI winners. And it's really only one month into the year. So I think the size and the scale of the pullback is that who wants to be a hero if you're institutional money manager or a hedge fund?

uh... you know and by the dip on the first day i mean catch the you know proverbial falling night so to speak so i think you know people waiting here from these companies that's the that's the positive is that we're going into the need of earnings season where you're gonna get microsoft and meta

and Apple and eventually others to talk about what they're seeing and what their capex plans are. So I think we'll know a little bit more in the coming weeks, but for now I think yes, it's probably a healthy and needed consolidation phase.

A healthy consolidation phase that sees 16% wiped off of Nvidia, more than half a trillion dollars lost, the lowest it's trading at since October of last year, and we don't get its earnings until February the 26th. How many calls have you had about Nvidia, Jordan? And what do you think it means for the popularity of its very expensive chips? Well, look, people are very concerned, obviously, because it's probably the most owned stock in the market.

I think you have a lot of people that don't fully know what they own, and they're just panicking and selling. I think the real institutional money managers, we're not seeing a wave of selling across our equity desk at Mizuho. I mean, we're seeing some profit-taking. We're seeing some investors rotate some money out of all these companies.

Tech names, that's to be expected, right? But I'm not hearing from people that it's game over for NVIDIA. It's game over for Broadcom, Marvell, Micron, or the big cloud hyperscalers. And if anything, I think people are looking at that price point that DeepSeek threw out of $6 million and questioning if that's even real.

Again, I think the real money managers who are here for the long term and think more months, not days, are going to wait and hear from the companies before they do anything. You echo what Tony Hwang of T. Rowe Price was saying at the start. We called you up immediately that we got these sorts of market sell-off news on our hands, Jordan, because we saw your note. It had a great title, of course, and the fact that we've got a freakout happening. DeepSeat creates a deep freak.

across tech. I'm interested though about the calls you're fielding and as to whether or not you're starting to see a question of buying into Chinese names. Is that something you have access to? Is that something that people want to see as a winner here?

Well, that's a great question. I mean, I'm not seeing people call me and say, these are the list of Chinese names that I want to own. I think this will bring up a discussion point that's probably needed and was overlooked is we can't all own the same trade, right? If everyone owns the same four, five, six names, right?

And then you get news like this which questions the longevity of this, sustainability of this thesis. People just rush to sell and it's painful, right? It's escalator up, elevator down. But the real thing is that people are getting a healthy wake up call. It's like, look, I have to have a diversified portfolio. The other thing that I'm really encouraged by is look at a lot of the green on your screen as it relates to software and some of these larger cap tech names. They're not all getting sold indiscriminately.

You know, a lot of areas of software are up. And I think people that you've had on your show say this could only increase the adoption over time and make people want to deploy and invest more to catch up with China or deploy some of these cost saving measures to build out their own model. So I think it's just too early to know.

Jordan, we're going to hear a lot from companies over the next several days and coming weeks about CapEx. Does this news on DeepSeek unravel the argument that Manhattan Project-like spending is needed to maintain an edge in AI?

I mean, yes and no. I think it raises a lot of questions that were already there. So most of the meetings I have with investors, the question comes up is, are they going to ever see a return on these tens of billions of capex spending that they're deploying? Will it ever monetize? What will the returns be? Or is this just throwing money down a hole?

I think that is going to be an ongoing question. And until we see these companies, for example, Facebook or Meta and Microsoft this week, talk about what they're seeing in terms of monetizing AI, there's going to be some questions and doubts. But again, I don't think this is going to create we're cutting CapEx, we're scaling back because they're all in an arms race with one another.

And China's deep-seek initiative isn't going to change that. That's why I think Microsoft talked $80 billion. Mark Zuckerberg at Facebook talked $60 to $65 billion. The Stargate of $100 to $500 billion. I don't think that, again, one new Chinese app

that makes some aggressive, audacious claims is going to all of a sudden create this pullback effect. If they start to see real savings and they can do this faster, yeah, they might do that. But that's going to take time, and we haven't seen that yet.

Jordan Klein, so good to have your voice on the show. The Mizuho Americas TMT analyst, we appreciate it. Coming up, much more on China AI startup DeepSeek's breakthrough model, what it means for the United States in terms of supply chain. Former Congressman Ken Buck joins us next. This is Bloomberg Technology.

89% of business leaders say AI is a top priority, according to research by Boston Consulting Group. But with AI tools popping up everywhere, how do you separate the helpful from the hype? The right choice is crucial, which is why teams at Fortune 500 companies use Grammarly.

With over 15 years of experience building responsible, secure AI, Grammarly isn't just another AI communication assistant. It's how companies like yours increase productivity while keeping data protected and private.

Designed to fit the needs of business, Grammarly is backed by a user-first privacy policy and industry-leading security credentials. This means you won't have to worry about the safety of your company information. Grammarly also emphasizes responsible AI so your company can avoid harmful bias. See why 70,000 teams and 30 million people trust Grammarly at grammarly.com slash enterprise. That's Grammarly at grammarly.com slash enterprise.

Gramelly. Enterprise-ready AI. Your customers are important to you, but they won't feel that way if they're stuck messaging a clunky chatbot or waiting on hold for a representative. Estimated wait time is 25 minutes.

With Sierra, your company can deploy a branded AI agent that engages and delights customers anytime, anywhere. Sierra agents pick up every phone call and personalize every interaction. No more menus, no more hold times. And if you have an issue, Sierra's AI agents solve tough problems. Whether they're helping your customer pick out the person

Always friendly. Always helpful. Always ready. Visit sierra.ai to learn more. That's sierra.ai.

As DeepSeek's AI potential continues to jolt the tech sector today, what does this mean for geopolitics and President Donald Trump's agenda for AI dominance? Joining us now to discuss all of this is Ken Buck, former congressman from Colorado's 4th District. Congressman, we have to ask you, when it comes to AI, is DeepSeek's breakthrough a Sputnik moment for Washington?

You know, I think time will tell on that. I'm not sure. I think that the critical factor is that America develops its own ship manufacturing here. We don't rely on Taiwan Semiconductor. We don't rely on

in this world marketplace. We certainly could be behind. We recognized a few years ago that we needed to incentivize chip manufacturing in the U.S. We

we passed the CHIPS Act in Congress. It has been a failure. Throwing a lot of money at a lot of different companies has not worked. I think that Donald Trump's policies involving, you know, a combination of tax incentives and perhaps tariffs will be more effective in trying to raise America's productivity in this area. Congressman, you brought up the CHIPS Act and your misgivings and concerns there. How should the president...

adjust maybe the implementation of this law and maybe the doling out of some of the money that still remains in a way that could meet the challenge that companies like DeepSeek are posing from abroad? Well, one of the biggest benefactors of the CHIPS Act has been the

Taiwan Semiconductor. And that's a mistake. They are not producing chips in the U.S. They have a plan to produce some in the future. We don't know if those will be the high-end chips or not. We don't know if Taiwan Semiconductor will actually...

mesh with the workforce here in the U.S. And so I think what we've got to do and what President Trump, I'm sure his advisors are telling him right now, is find those companies in the U.S. that can produce the high-end chips that Taiwan Semiconductor has a monopoly on. Look into the monopoly. First of all, we certainly have antitrust laws in this country that are

prohibit a company from having a 95 percent market share like like Taiwan Semiconductor does, but find the U.S. companies and make sure that we have the tax structure in place and other incentives to to make high end ships in the U.S. This whole issue of deep seek. We don't know exactly whether someone violated our export controls to to China or not.

I'm looking at Nvidia down almost 17%, massive wipeout of market capitalization by $600 billion, Congressman. But to that point, did perhaps China circumvent some of the limitations on Nvidia's exports? Is that something you're thinking through rather than perhaps this was innovation because they couldn't get their hands on the latest and greatest?

Well, this wouldn't be the first time that China stole technology from the U.S. or other countries, and then with their labor market and with their other

production advantages got ahead of the curve on us. And so I think that's certainly one of the issues that we've got to look at is whether they did violate export controls or not. But more importantly, what do we do in this country? That cat is out of the bag right now. What do we do in this country to make sure, and in Western Europe, frankly, to make sure that we stay ahead of the AI race? But the answer thus far has been money.

$500 billion unveiled by potentially Oracle, OpenAI and SoftBank into the future for AI infrastructure in the United States. The creation of more $3 trillion companies, but ultimately that amount of money perhaps has slowed innovation here in the US. Would that be a concern?

I think it's a big concern. Again, I don't think we just throw money at an issue. I think we've got to be very precise in our strategy and how we develop AI. We all recognize AI has these tremendous potential risks.

wonderful benefits, positive benefits. But when you're talking about a totalitarian country like China, you've also got to think about what that country can do with AI advances that will put us at risk and our allies at risk. Former Congressman Ken Buck, thank you very much for joining the show today.

SoFi shares, like the rest of the market, slumping today. In fact, having its worst day since March of last year. The fintech lender publishing perhaps a forecast that seems to be below where the market wanted to see. We've got a broad tech route, of course, linked to deep seek more broadly in generative AI over in China. But here to discuss the earnings, SoFi CEO Anthony Noto. A tough day to report earnings, Anthony. And I ask you about the forecast because, look, your revenue wasn't rising.

record rates, adjusted revenue up 24% for the fourth quarter, but you're pushing us forward in a forecast that isn't as high as the market wanted. Why?

Well, I think two things. One, 2024, we took a pretty conservative approach to the year in terms of our growth and our profitability. We really wanted to make sure that we were able to drive strong profitability in 24 to achieve gap profitability, to ensure we reinforced our balance sheet and a capital cushion. And we did just that.

Prior to 2024, we had committed to a balanced approach to growth and profitability where we would reinvest 70 cents of every incremental revenue dollar, and we call that 30% incremental EBITDA margins. So we took our revenue forecast for 25 up. We're in the best position we've been since I've been here at SoFi for the last seven years. We're calling for 25% revenue growth, strong margins at 26%, just not an expansion in the margins,

because we want to invest in the massive opportunity that still sits in front of us. And so we took revenue guidance up and it's about 10% higher than the street, but that will require more investment and it will help us ensure we have growth beyond 25. And we also took our revenue guidance for 2023 to 2026.

on a compound and a growth rate basis higher as well. So the profitability of the business is there. We could drive more to the bottom line, but we think that's not the prudent thing to do because we just see massive growth in front of us. And the ability for us to keep driving member growth and product growth of more than 30%, which we've done, and revenue growth of more than 25% is what we're calling for through 2026.

Anthony, of the investment that you just described, what portion of it will you be putting toward artificial intelligence? It's the topic at the top of everybody's minds and people are watching how companies are deploying it. Do you see, for instance, an agent driven service perhaps at SoFi at some point?

The vast majority of our investment going into 2025 and 26 will be in building unaided brand awareness, becoming a trusted household brand name. We have great products. The reason why we're driving such strong product and member growth is because the products are very differentiated. And when we make people aware of them and they use the products, they not just use the first one and they gain trust with us, they use the second and third one. 30% of our product growth in the quarter of 34% was from our existing members.

About 40% of our members take out a second product within 30 days. So we'll continue to invest in differentiating the products. Sofa Money is a great product. It has a high APY. You can do person-to-person payments. You could do Zelle, auto pay, two-day early paycheck. So we'll be investing in that. We'll be investing in our Invest product to expand the selection for Invest in addition to other product categories like insurance and credit card and small and medium business lending. Okay.

You want to be able to offer crypto again. And that's something the administration talked a lot about. Look, on a day where we see such volatility that engulfs crypto too because of DeepSeek, is that really a product placement that you want to get into again?

If the regulations change so that cryptocurrency is permissible by bank holding companies, we would absolutely provide not just what we used to provide, which was the ability to safely and securely trade Bitcoin and other cryptocurrencies, but we'll also go into other areas like custodying and clearing in addition to asset-backed lending, etc. But that will all be gated by the regulators, which we think should be coming over the next 24 months, and we'll be ready when it does.

Ultimately, are you feeling more risk on in this environment, the new administration, a way in which to develop your business, maybe M&A, because it doesn't feel like a risk on day to day. We are definitely leaning into 2025. 2024 was a record breaking year. We had record revenue, profits, returns, member growth, product growth. Couldn't be happier with the year that we had. The most proud I've been of the company since I've been here. But I think 2025 will be even better.

We think the outlook is the best environment we've operated in over the last seven years. Our business is bigger, stronger, and more well-known than it's ever been, and we have more resources to go after the opportunities. So we love our competitive positioning. We like the macro backdrop. It's a very different outlook than when we came into 2024, and we're definitely being more aggressive in innovation and driving durable growth and strong returns.

Anthony, back to the regulatory side of things. What is the one thing, one hurdle that you would like to see the new administration and Congress clear for you?

The biggest question is what will bank holding companies be allowed to do with cryptocurrency? What will be permissible? And that clarity is really critically important. The interest rate cycle, I think, has pretty good visibility and transparency to that. I think the economy also has really strong economic indicators. The big question is how much can we innovate? How much can we invest in these different asset classes that our members want to reinforce their ability to not just

borrow better and save better and protect better, but to invest better. Investing is critical to reaching their long-term financial goals. We want to be there for every one of the major financial decisions our members make in their lives and all the days in between. The key is spending less than you make and investing the rest. So the more opportunities we can give our members to invest, the faster they'll get to their financial outcomes. Anthony Noto, SoFi CEO on the day of your earnings. Great to have you on. Thank you so much.

Let's return to the story of the day, though, which is, of course, DeepSeek and generative AI being powered in China for much smaller amounts of money, managing to circumvent perhaps the limitations on compute power, Mike. We see a whopping 16% market cap erosion on NVIDIA. But we all thought we were going to be talking about earnings this week. We're looking towards Meta, Microsoft coming in the 29th, Apple, which is actually in the green, coming on the 30th.

Cara, this is rewriting the narrative for earnings later this week and through the rest of this month. And there is also a wild card, and that is the new president in Washington and how he will react. We still have not heard from him. And he does look at the market as a benchmark of U.S. success.

Also, we see how people try to continue to download DeepSeek as, of course, an app. It was the number one app on Apple and currently unavailable. They say they are currently being afflicted by some sort of attack. But that does it from this edition of Bloomberg Technology. Mike, do not forget to check out our podcast. You can find it on the terminal as well as online on Apple, Spotify and iHeart. This is Bloomberg Technology.

I think a

A lot of people think that you're supposed to be going to therapy once you're like having panic attacks every day. But before you get to that point, I think once you start even noticing that you feel a little bit off and you can't maintain this harmony that you once had in relationships, that could be a sign that maybe you want to go talk to somebody.

There's always a benefit in talking to someone because we can all benefit from improved insight about ourselves and who we are and how we behave with other people. So if you're human, that's like a good indicator that you could benefit from talking to somebody. Find out if therapy is right for you. Visit BetterHelp.com today.

That's betterhelp.com. This episode is brought to you by Intuit Enterprise Suite. Let's talk to the CFOs and business leaders out there. As your business grows, so do the challenges. That's where the new Intuit Enterprise Suite comes in. It's an AI-powered solution that brings together financials, payroll, marketing, and payments processing.

all in one place. And with automated multi-entity accounting and reporting, you'll be doing less manual work and unlock richer insights on your business performance. Learn more at Intuit.com slash enterprise. Money Movement Services by Intuit Payments, Inc., licensed by NYDFS.