You want work to be less hard work. You hear an ad for MHR, so you reach out. We connect your department systems, which leads to real-time data sharing that uncovers new insights, which empower your decision-makers and triple monthly sales, which leads to high fives and awkward hugs. You say a big thank you. We say you're welcome.
MHR, the science behind HR, payroll and finance, the science behind a new world of work. Discover more at mhrglobal.com. This episode is sponsored by Wombl Bond Dickinson, an international law firm of more than 1300 lawyers across 37 offices in the United States and United Kingdom.
In today's complex world, new problems need new perspectives. One will bond Dickinson thrives on change, bringing together people with different skill sets and experiences to give their clients a competitive edge. Across a range of markets, they support businesses and private clients on critical challenges, from energy transition, digital transformation and cross-border investment to corporate finance, dispute resolution and personal wealth planning, all with a mix of minds you won't find anywhere else.
Wombelbond Dickinson. A point of view like no other. Discover more at wombelbond Dickinson dot com. From The Times and The Sunday Times, this is the story. I'm Manveen Rana.
Amid all the fanfare of the Trump inauguration, there was one abiding image that could come to define this administration. One postcard to posterity from Trump 2.0.
We're seeing all the big names in tech in that room, the camera lingering on that remarkable picture. It's so notable, especially when you think about the history of many of these individuals with the soon to be president, as well as the relationships between them. I mean, seeing them all in the same room, seemingly kind of on the same side of things here, smiling, chatting, it really is something remarkable. I mean, we just had the cameras trained on.
The saying goes, a picture is worth a thousand words. So, lots of people have seen that photograph of Mark Zuckerberg, Jeff Bezos, Suno Prichai, the head of Alphabet. Elon Musk was there as well. You know, really, this row of tech billionaires being called the Brologarchy.
These tech founders are really running the most powerful, largest companies on the planet, and obviously they're the ones really pushing the envelope and building this new generation of AI tools that they're promising is going to change everything, you know, comparing it to the invention of fire.
And there they were in the front row, put there at the request of Donald Trump. And it was, I think, a very strategic decision. You know, they were standing in front of cabinet members, senators, the most senior people in the new administration, in front of all of them, were his new allies in the tech industry.
And there they were, the Broligarchy, the new masters of the universe, the architects of our future, with unparalleled access to power, they seemed invincible. And yet, just one week later, history was being rewritten.
tech stocks sent the markets into a tailspin today. This is all being driven by a Chinese startup that few people had ever heard of before a few days ago. Deepseak is on par with what OpenAI and Google have made, even though those companies are sinking billions of dollars, tons of years, and lots of development into this space. Deepseak did it for cheaper in a matter of months. Deepseak.
AI from a Chinese company should be a wake-up call for our industries that we need to be laser focused on competing to win. The story today. Deepseak. AI's Sputnik moment.
I am Danny Fortsen, the West Coast correspondent for the Sunday Times, based out here in Silicon Valley. So Danny, there you have the Broligarchy on top of the free world at the inauguration at the right hand of President Trump. But what was happening further afield in the wider world of tech?
Well, so if you step back, if you go back to November 2022, I think of it as almost a big bang moment. That's when Chet GPT came out and everybody was like, Oh my goodness, we have entered a new world, a new age of AI. And there's been a real, just absolute frenzy, the likes of which we have not seen since the dawn of the internet, just hundreds of billions of dollars being invested in data centers, thousands of new startups, the biggest companies in the world spending.
trying to be first, who can dominate this new age of AI, these machines that they say will be better than the best human at any cognitive task in the very near future. What does that mean for the economy, for the world, etc. It's been a real frenzy out here
And there in Silicon Valley where you are, you've got all the biggest firms we know of competing to be there, to be first. At the same time, just as they're all celebrating how close they are to power with President Trump's inauguration, in China, another advanced reasoning model is being launched. Tell us about that.
So it was called Deep Seek, and it was started actually spun out of our hedge fund in China, low hedge fund. No one's ever really heard of it. It's called High Flyer. But they also started really drawing in some of the brightest minds in China, from the worlds of computer science, artificial intelligence. And they started building a model.
to rival the best models of from the likes of Google, open AI, et cetera. And what's really interesting is there's been a ban on the export of certain types of chips, the most powerful types of chips to China because a lot of people see this as, you know, this is a great competition. Think of like the Cold War, we are in a quote unquote techno economic war with China. AI is this new super powerful technology can be used as a weapon.
in lots of different ways. And so there was a real kind of race and then just kind of below, but beneath everybody's nose, there's this little company that no one's really much paying attention to that is building this model that when they come out with it kind of shocks the world because it is according to a whole bunch of benchmarks as good or better than the best models that the best companies in Silicon Valley are developing.
And Danny, that was sort of summed up by a man called Mark Andreessen. Tell us a bit about him. He is kind of a legend out here. If you go back to the dawn of the internet age, he founded Netscape, one of the first browsers, took it public in his 20s. He is a billionaire for the last 15 plus years. He's been one of the most prominent investors in Silicon Valley.
He also sits on the board of Meta, parent company of Facebook, WhatsApp, etc. And is really a trusted advisor to Mark Zuckerberg and the other kind of titans of industry out here. And he's also seen as a real kind of thought leader, a kind of public intellectual.
What's really interesting is he kind of personifies this kind of rightward shift that's happened in Silicon Valley. You know that was manifested in that picture, which is going to go down in the annals of history of all these tech oligarchs standing in the front row of the inauguration.
But he was a lifelong Democrat, and he said, particularly under the Biden administration, he really changed his tune. He thought Biden was really anti-tech. He said, they, quote, unquote, hated us. And very prominently, in the months leading up to the election, he came out publicly for Trump.
And I think he gave a lot of people permission by him coming out and saying that publicly to do the same. And so that's part, he's a central figure in this vibe shift to the right out here in Silicon Valley. He seems to have been rewarded for that. Indeed, Elon Musk is the head of this new Department of government efficiency, Doge. And Mark Andreessen calls himself a quote unquote unpaid intern. So he is very much in the tent, the new kind of Trump tent, as it were.
So as a Silicon Valley thought leader, as a man who has the ear of the American government, what was his verdict on deep-seek? Fascinating. So he said two things. He called it a great gift to the world because if you step back, what's really interesting about where we are with AI at the moment, there's hundreds of billions of dollars of infrastructure being invested into this.
This is a theory I've had for a while, and I think it's playing out now, is that AI is going to become a bit like electricity. It's going to be everywhere. It's going to be trivially cheap. And it's just going to be kind of part of everything we do going forward. And if you have a company that has come out of nowhere and said, we can do this, we can build these very capable models for extremely cheap.
What that does is effectively fast forward the moment where AI becomes trivially cheap and ubiquitous. And again, then you step back. What does that mean for the world? If machines in the very near future are going to be better than any human at any task.
What does that mean for society? So it's big, big questions. And he also called this AI's Sputnik Moment, which is quite a profound way to kind of frame this. If you go back to the Cold War, we were in this great competition with the Soviet Union and they sent up Sputnik, which was a satellite into space. And we thought we were ahead. We thought we were the best, rah-rah America, we're number one. And it was this moment
And again, talk about a collective freak out where, oh my goodness, we are losing. We are losing against our big geopolitical rival. And it was this kind of totemic moment. And that's why he's saying, again, we thought we were the best. We're spending more money than anybody else. And we're keeping China from having the best tools to build this stuff. And yet, they've still created this quite amazing thing.
So this is an extraordinary moment, not just in the history of technology, but in terms of geopolitics too. Yes. On Monday this week, just describe how that big moment for technology was playing out on the market. Yeah, so Wall Street woke up.
If you kind of step back and look what's happened over the last two years, you've had companies like NVIDIA, which are kind of, if you think of a gold rush, they're selling the picks and shovels. They're selling the very high-powered, very specialized chips that power the data centers, that power all of these frontier kind of cutting-edge AI models. And they're based out here on the West Coast, and they've had an incredibly extraordinary run-up. They're now one of the largest, most valuable companies on the planet.
Before this week, there was something like $3.3 trillion, just really extraordinary. Something like 80% of their employees are now millionaires simply because of how much the stock has gone up. It's been a real kind of transformative moment for that company. And within hours, they lost $600 billion worth of value. That's extraordinary.
And Annie, you described how Nvidia has very recently become this huge giant, a colossus. It's worth an absolute fortune. How important have these key companies in Silicon Valley? How important have they been to America's economic boom? I mean, to see the markets crashing for them, what does it mean? Not just for the tech sector, but for the American economy. Well, if you look at what's happened in this AI boom over these last two years,
And you look at the eight most valuable companies on planet Earth, seven of them are on the west coast, because they've all seen their share prices explode with this optimism about what AI is going to mean for their companies and for the world. And of course,
They're profit margins. I can't remember what the last figures are, but it's something like a quarter of the S&P 500, the main index, stock market index in America is comprised of those top seven companies. They've really just kind of eaten up a lot of the market. And when you think about people's retirements, their pensions, a huge amount is wrapped up in a very, very small number of companies.
And to understand why the markets seem to be crashing, why they lost so much value in the course of a few hours, why this was such a big moment for tech and for geopolitics. I suppose we really have to look at the company in the middle of all of this. Deepseek, a company who, as you said, most people hadn't heard of. What do we know about the founder of Deepseek? Who's at the helm?
Well, the short answer is not much. He's 39 years old. He's a hedge fund manager named Liang Wenfeng. And he's been pretty low profile. You know, he was born in 1985 in a poor city in China's southern Guangdong province. His dad was an elementary school teacher.
He studied electrical engineering. He founded a quantitative hedge fund, one of these kind of high frequency trading hedge funds about a decade ago that rely on these really complex algorithms to trade the markets. And he announced on WeChat a couple of years ago that they were going to expand their remit beyond investing to, quote unquote, explore the essence of AGI, artificial general intelligence. And about a month later, that's when DeepSeek was created.
Do we know if he's linked to the government in China? We don't. We don't. Perhaps in recognition of Deepseek's progress, he was asked on January 20th to give a speech at a closed-door symposium hosted by China's premier. And this was according to the state broadcasters in China's CCTV. So it does appear he has kind of gained some notoriety and has certainly been noticed by the powers to be.
But again, if you are a businessman, large or small, you are subject to the rules and strictures of the CCP, which can be quite strict. So it's not clear what direct relationship you may have with the CCP, but of course he has to play by those rules.
So this is a very short timeline. Yes. He founds deep-seak in 2023, fast forward to less than two years later, and he's suddenly disrupting the entire tech world with this model. Have you seen it up close?
I've downloaded. I haven't started playing with it yet. But there's two things that they have created that are really stirring the markets. There's V3, which is like a large language model like chat GPT that we all are quite familiar with.
The other thing it has created is R1, which is a reasoning model. And what's really interesting about reasoning models is they kind of think before they give an answer. So they're able to take on more complex tasks. And again, this speaks to
One reason why people are so taken aback by this is that it was thought before this came out that OpenAI had a huge lead in this market because they have their own reasoning model that they've been pushing out there. And it has been far and away the most powerful. And then all of a sudden, again, you have not only a very capable, large language model, but also a reasoning model, which can be used for a lot more complex tasks and is much more capable.
And for those of us who aren't great at the tech, just give us an example of what a reasoning model is able to do that you don't get with just ordinary chat GPT. Yeah, it's things like if you give them a logic puzzle.
You know, instead of just spitting out an answer that may be nonsense, which we've all done with whether it's with Chad GPT or Claude or Gemini, it'll think it'll give you those three little dots, right? And it could take a few minutes. And then it comes back with something that is reasoned, it explains why it got there and you can kind of have more of a conversation that it just goes a deeper level. If you think about talking to an intern versus talking to a PhD, that's kind of the difference.
And from what you're hearing from people who've been playing with deep-seek so far, what are they saying about it? Well, I mean, I think especially out here in the West, the thing that a lot of people have grabbed onto is the idea that, you know, if you ask it about Tiananmen Square.
or the Uyghur population in China, which are severely suppressed there, it doesn't give those answers, right? There's a certain level of censorship. But again, if you'd step back and you think about, well, what's AI going to be used for? Well, if I'm a little company and I want to create a chatbot, most likely my customers are not going to be asking about Tiananmen Square. They're going to be asking about, you know, how can I return this shirt that doesn't fit?
So it's clearly an effective bit of kid. It also seems to be remarkably cheap. Just talk us through that part because, you know, as you said, this came with a backdrop of people in Silicon Valley in a race to spend more and more on AI. Just give us a sense of what they were planning, why there is such a huge disparity. I'll make an analogy here to the streaming music boom.
If you go back to 90, I think it was 99,000, I'm dating myself here. When Napster came out, it was this kind of pirated software, but it was this sudden arrival of this technology that allowed anybody to share their music, to download and upload whatever they wanted, and it completely upended the music industry for 20 years. It took 20 years for the music industry to get back to a point where they were making enough money, or making the money they were the day before Napster arrived.
Napster died. It was buried under lawsuits and it was replaced by Spotify, by Apple Music, etc. And I think what's really interesting is that a company like DeepSeek, what they have done, it appears, is look at all the best models.
and use those models to train their model. And so it's a really interesting time for all of these tech firms, like OpenAI, who have prided themselves on being first. But being first doesn't necessarily benefit you because you have others who can reverse engineer what you are doing. Figure out all these little ways to do it better, faster, cheaper.
And then all the results is that you have something that is a fraction of the cost to build, to run, et cetera. And that's what's happened right now. Is it any better for the environment? Yeah, so when you talk about AI, the term that uses inference, basically when you ask it a question, it infers the answer. And inference is a cost. And that cost is calculated basically by energy consumption.
because you have to go send it out to the data center. It does its magic and sends it back. And this is another reason why this is interesting is that their inference costs, they have figured out a way to kind of not draw on the entire model, this whole data center, for example, but just the bits and pieces it needs. So they've done some incredibly clever software kind of hacks, if you will,
to make it much cheaper to run. And so for all of these companies, their biggest cost, once you build a data center, which is expensive, is running it, is the power it needs. And so if you dramatically reduce the power a model needs to function properly, that's a huge, huge unlock.
Coming up, how deep-seek isn't just a win for Chinese tech. It's also an unlikely victory in the new battleground for geopolitics. The chipwalls. That's in just a moment.
You want work to be less hard work. You hear an ad for MHR, so you reach out. We connect your department systems, which leads to real-time data sharing that uncovers new insights, which empower your decision-makers and triple monthly sales, which leads to high fives and awkward hugs. You say a big thank you. We say you're welcome.
MHR, the science behind HR, payroll and finance, the science behind a new world of work. Discover more at mhrglobal.com. This episode is sponsored by Wombl Bond Dickinson, an international law firm of more than 1300 lawyers across 37 offices in the United States and United Kingdom.
In today's complex world, new problems need new perspectives. One will bond Dickinson thrives on change, bringing together people with different skill sets and experiences to give their clients a competitive edge. Across a range of markets, they support businesses and private clients on critical challenges, from energy transition, digital transformation and cross-border investment to corporate finance, dispute resolution and personal wealth planning, all with a mix of minds you won't find anywhere else.
Wombelbond Dickinson. A point of view like no other. Discover more at wombelbond Dickinson dot com.
So Danny, I suppose a lot of people will be asking, here's a company deep-seek that most people hadn't heard of. It suddenly appears, disrupts massively changes the world of tech. It's not only released a model that's free to use, it's also open source, which means people can see how it works, what it's doing. Why would they do that? What benefit do they get for releasing all of this for free?
I'll give you the potentially devious answer and then the straightforward answer. So this is launched by a hedge fund. So what an incredible trade if you could spend a couple of years building this model that is going to blow up the stock prices of the most valuable companies on the planet. And before you do that, you bet on those prices falling and you make hundreds of billions of dollars of profit.
We have no idea whether they've done that or not, but that's totally possible, right? Because if people are already saying this AI bubble is about to burst, et cetera, so that's just one kind of Bond villain idea of, well, if we can kind of come up with something that really rivals these tools that are underlying these huge stock market booms, once that collapse hits, we can make a lot, a lot of money. Stepping back.
If you look at the idea of open source, the idea is if it goes back to the electricity analogy, imagine all the things that we cannot do without electricity. So if you are creating a base layer upon which a whole new generation of apps, services, companies, ideas are built on,
You're going to figure out all kinds of interesting ways to make money on that. It's going to benefit us massively, and that is the bet. And should we be skeptical about how much we're told this would have cost for them to put together?
Yes. Yes. So they said it was less than $6 million to train, but that was the final quote unquote training run. There are many runs before that and well, up to two years before that of building up to that point.
You know, they have 200 researchers. Those people are expensive. The data centers are expensive to run. So, you know, the actual number is much higher, but whatever the final number is, it appears it's exponentially lower than the kind of the state of the art that we're building out here on the West Coast.
Danny, as we were saying earlier, this is not just a huge leap for technology. It's also a real moment in terms of geopolitics. The US government in December had intervened in the world of chips. Just explain why chips matter to the production of AI. You described it earlier as the picks and shovels and the gold rush, but just explain the relationship and what the US government was planning to do.
Yes, so NVIDIA, which is based out here in San Jose, California. It was actually started at a local restaurant called Denny's, which is a terrible restaurant. But I read, know the Denny's because I grew up in San Jose. They started as a video game chip company, kind of like a little niche player and then
over time have evolved into this behemoth that has really powering the AI revolution and that really comes down to the chips that they make that allow these big training models to be developed and then to operate.
And the chip band specifically targeted its most advanced models, the ones, as I mentioned, that are really used to train what they call the frontier, the leading edge of the technology. And so there's been a ban on
NVIDIA is selling those into the Chinese market. The idea being that we're in a race with China, we don't want to give them our best stuff to then beat us in this race. Which is why DeepSeek appears to have used like a lower grade model from NVIDIA, which is not subject to the chip band. Which again is why people are so taken aback by this moment, is that they basically use second rate materials to create a first rate product.
And that chip ban came into effect under President Biden. Do we know what Donald Trump was planning for the chip war effectively? We don't. I mean, he's been pretty bellicose toward China. At least in his talk around tariffs, a trade war, AI is right at the center of that. And so surely it will be kind of a chip on the table when it comes to the negotiations with China. And I think a pretty big chip, especially
As we've seen with DeepSeek, the rate of acceleration of this technology really picks up. How has Donald Trump reacted to what's happened with DeepSeek?
He said something to the effect of, oh, it's good. It's cheaper to run. So it's unclear what happens next. You never quite know what the president. But even if a lot of money is being wiped off, indeed. But that said, I mean, we shouldn't cry too many to his friend video. They're still worth over $2 trillion. So they're still doing OK.
And is this almost a form of economic warfare? You know, as you said, Donald Trump has been quite bellicose with all things China. He talks about tariffs. He talks about starting a trade war and really having control of that. And yet here, the week of his inauguration, we see a small company claiming to spend very little in China coming in and really destabilizing the markets. Yeah.
As he's taking power, is this a very effective form of economic warfare by the Chinese? To agree, it's just competition. What's really interesting about what has happened is the only reason Deepseek was really forced to make all of these breakthroughs, which has now knocked more than a trillion dollars off the stock market globally, is because of the chip ban.
because they were forced to use less powerful chips. And so in a way, it's a consequence of this kind of back and forth between the US and China. It certainly changes the balance of power. Before a week ago, it was very clear that the US was in the lead.
Or let's maybe we were all on Hopium thinking like, oh, yeah, we've got this, you know, we're just spending the most money. And to a degree, that is still true. The US is spending more than anybody else that, you know, they announced Stargate, this between $100 to $500 billion data center project on US soil.
That will also matter, scale will matter, because just how much computing you have available to run these systems really matters when you scale them up. Again, going back to the idea of electricity, how many power plants do you have? But again, it's really a shot across the bow of, you know, you're not nearly as far ahead as you think you are.
Out here, they talk a lot about competitive moats. What's your moat? How deep is your moat and are there alligators in it? And they just realized, oh, maybe our moat is quite shallow and we can just hop right over it. And I think that's why people are freaking out.
That was the Sunday Times West Coast correspondent Danny Fortsen, our man on the ground in Silicon Valley. You can read all of his work at thetimes.com with a subscription, and you can also listen to his own tech podcast with the Times Technology Business Correspondent, Katie Prescott, every Friday. The producer today was Sam Chantrasak, the executive producer was Kate Ford, and sound design and theme composition were by Mala Seto.
If you enjoyed this episode, if you think others might enjoy it too, then do pass it on and leave us a review wherever you get your podcasts. Thanks for listening. See you tomorrow.
You want work to be less hard work. You hear an ad for MHR, so you reach out. We connect your department systems, which leads to real-time data sharing that uncovers new insights, which empower your decision-makers and triple monthly sales, which leads to high fives and awkward hugs. You say a big thank you. We say you're welcome.
MHR, the science behind HR, payroll and finance, the science behind a new world of work. Discover more at mhrglobal.com.
In today's complex world, new problems need new perspectives. One will bond Dickinson thrives on change, bringing together people with different skill sets and experiences to give their clients a competitive edge. Across a range of markets, they support businesses and private clients on critical challenges, from energy transition, digital transformation and cross-border investment to corporate finance, dispute resolution and personal wealth planning, all with a mix of minds you won't find anywhere else.
Wombelbond Dickinson. A point of view like no other. Discover more at wombelbond Dickinson dot com.