Bloomberg Audio, Studios, podcasts, radio news.
Investors woke up this morning to seismic losses for US tech stocks.
Just creating a massive drag hundreds of billions of dollars in market cap loss.
The Nasdaq fell as much as five point two percent in overnight trading. Is Naszak futures and global tech stocks extend their losses on fear of the AI race may
no longer be America's to lose. By the close of a wild trading day, and Nvidia's stock fell by seventeen percent, The chip maker saw five hundred and eighty nine billion dollars erased from its market valuation, breaking its own record for the biggest single day loss of market value on the US Stock Exchange, and energy firms expected to profit from AI demand sank two, led by a twenty one percent drop in the share price of Constellation Energy Core.
This is a really stunning development, kind of coming out of aware for the AI world.
This stunning development came in the form of a new AI competitor, a Chinese company called deep Seek. Jackie Devalos, who covers artificial intelligence for Bloomberg, says deep Seek has disrupted the AI world by building a model that rivals its competitors using a lot fewer resources.
Tech companies spend billions of dollars at this point on training their AI models, And you had Meta last week saying that it would spend upwards of sixty five billion this year on AI development, So it's a lot of money.
Deep Sea claims it's spent closer to six million dollars to train its most recent model, and it's made its system open source, meaning anyone can access its code and build on it in a matter of days. The company's chatbot, which functions similarly to chat GPT, has rocketed to the top of App store charts, and its sudden dominance has some investors questioning the strategies and the valuations of the US tech companies that are developing AI and the technology that powers it.
And I think deep Seek is now putting a really big focus on what can you come up within the short term or are we just wasting our money here?
This is the big take from Bloomberg News. I'm Sarah Holder today on the show The Rise of deep Seek. How it could upend the USAI industry's competitive advantage or push it to innovate. It's not what deep seek does that's so revolutionary. Functionally, it's actually very similar to a lot of other AI chatbots.
Yeah, it's a large language model that can handle reasoning, and you can put in a question, it gives you answer. Anyone can go to the app store and download it. If you're a student looking for help on your essay, or if you're a researcher looking to help on your PhD dissertation, the same people that kind of go to a chat GBT or an anthropics clod or a Google Gemini. It's general purpose. And what you're also going to start
seeing is the coding community really delve into this. They're kind of one of the bigger cohorts that are early adopters of this because it is so crucial for coding. It takes a lot of the mundane tasks from coding out of the picture.
Why was deep Seek's rise seemingly such a surprise to so many people.
Part of the surprise was that this startup sort of came out of nowhere. It was founded in twenty twenty three by a guy that used to run a hedge fund, and it was kind of this spin out out of this hedge fund and it really didn't come onto the scene until this year when it released this model that when it performed on benchmarks, it was pretty comparable to what we've seen out of open AI and Thropic and
other leading models. And so part of the surprise was how were they able to do this so quickly and once we started seeing reports that not only was it faster, but it was actually done at a.
Fraction of the cost.
Remember, it's taken billions of dollars for these huge AI companies to build and train and release these AI models that power the chatbots that we all play around with.
Deepseek has said it spent only five point six million dollars to train its latest AI model, though it also cautioned that that figure didn't include all their costs, and since it needs less computing power, it could have significantly lower energy needs.
How did this.
Young startup develop this product so cheaply, Well, part.
Of it has to do with its roots as coming out of this quantitative hedge fund, so it has this deep knowledge in this computational framework. And not to get too into the weeds, but deep Seak has focused on what it calls maximizing software driven optimization. Basically, what is able to do is create an architecture that ran more efficiently, that didn't need these really intensive chips that are essential for the models that run open Ai and Google and
Meta Islama. And it did this because there's been export controls on chips going to China to prevent fresh AI startups from coming about and challenging US dominance. And so what we saw deep Seek do is kind of get around these export controls by focus its models to run in a different way, kind of in a different sort of architecture. And Nvidia's H one hundreds obviously have propelled the company.
To new heights, but those are really.
Expensive, so the fact that deep Seek doesn't have to rely as much on them kind of takes a huge chunk of that cost out.
When was deep Seek released and then when did it start getting so popular.
Deep Seek was founded in twenty twenty three, but we really started to see it make a splash in the AI world when it released a report that outlined its performance against other AI companies.
What stood out to investors in those reports.
You know, it performed really well.
So its latest model reasoning capabilities were similar to that of chat GPT. Deep Seek also said that it's better at various tax including coding and math and and chat GPT is sort of this leader in conversational AI. It's pretty versatile, but deepseeak really targets more specific application that developers might be looking for that are more customize and cost effective.
You still lack some of that, you know, polish.
It's not as polished as you might see a chat GPT, but in terms of its speed and just the way that it's able to handle a complex question or query is really surprising given you know, this company was just founded about a year ago.
So when deep seak rose to the top of the app charts, when it released this information showing how it worked, how it's performing, it had the seismic impact on US markets. What do experts think about the market reaction? Was it overblown appropriate?
I think you're seeing some market experts issue and I told you so that there was reason to doubt that so many billions being poured into.
The space was justified.
You also have other experts kind of saying, hold up, let's wait and see what this startup is all about.
We need to look under the hood.
These are just the first initial days of getting to compare what deep Seeks model does how much more runway has to go as well.
If you remember the early.
Days of CHATGBT felt like we've never seen anything like it before, and so I do think people are still trying to wrap their heads around how good is this model.
They're also trying to figure out whether deep seeks reports of lower costs and greater efficiency are as impressive as they initially seen.
You look at the other AI companies open Ai and Meta and Google, their models take a longer time to come out.
They're big tech giants.
They're naturally going to move a little bit slower when it comes to releasing their technology to the billions.
Of people it has captive already.
And that's the benefit of being a small startup that you get to release things quicker and make a big splash.
But whether that splash really has legs, whether it lasts, is.
Still yet to be seen, and I think that's what many people in the AI field are wondering about.
On January twenty fourth, deep Seat got a high profile not of approval when Mark Andresen posted on x that the company's latest version is quote one of the most amazing and impressive breakthroughs he's ever seen once.
He tweeted that coming from a tech titan like himself, he is very familiar with the AI space his firm, and Resent Horowitz is a big investor in AI companies, and so kind of getting his shock and awe at how good this Deep Seek model was, I think sent further reverberations throughout the.
Market coming up. What is the emergence of this Chinese firm mean for US dominance in artificial intelligence? And could Deep Seek's ultra low cost model put pressure on soaring American AI budgets. The US has been a clear leader in the artificial intelligence space, a position it's tried to protect by restricting the sale of AI chips and technology abroad. So I asked Bloomberg's Jackie Devalos how American policymakers might react to the China based Deep Seeks rise.
I think it's an opportunity for people to take a step back and evaluate.
Kind of where we are now, what.
It's taken, the level of investment, and ultimately what people want to see going forward. A lot of the attention right now, especially going into this new administration with President
has been around national competitiveness against adversaries like China. And it really is shocking when you consider the fact that the US has done a lot of work to make sure it can restrict the supply of these high powered chips to China, citing these national security concerns, and then you have deep Seak coming onto the scene, which means that they were able to achieve this low cost model without relying on these high power chips.
So clearly there was a work around.
Here that the AI community missed, but also the US missed well.
Right the fact that deep Seek was developed so cheaply for millions, not billions, What kind of pressure does that put on US companies and their spending. Are they going to try and adopt any of the strategies that deep Seek used to bring those costs down.
That's an open question because we've already seen a lot of concern about the supply of chips. Are we getting enough, are we getting them fast enough? Who are they going to in Vidias coming out with new models. Then you have in Vidia competitors coming out with new models, and so this deep Seek development really raises the question of well, wait a second, do we need this many chips to
begin with? You might see AI companies get more creative in how they're building these models, perhaps changing up their architecture and framework. You might see investors starting to perhaps dole out less money and demand that you know, they find other cheaper ways to make this. So I think it's all going to start kind of coming together with a bit more urgency now that we know that clearly this Chinese competitor is giving US companies a run for their money.
Could this be good news for the AI industry overall if it pushes this kind of innovation and investment, or does it feel like more of a threat.
I think it depends on who you are.
If you're the US government, and this definitely feels more like a threat. If you are an innovator who is enthralled by reading through some of these technical reports that Deep Seek put out, it's a really exciting time, and I think if you're an AI company, then this is a good time to perhaps like light a fire under everyone and say, hey, how can we be more creative. I think investors are always far more sensitive to these things than.
Perhaps the industry itself.
But on the whole, I think we're going to see far more urgency in releasing products, trying out new features, kind of keeping people engaged, because the more people use this, the more data goes in, the better these models get.
By Monday afternoon, and Video released a statement calling deep Seek an excellent AI advancement, one that would require a lot of in video products in the future. Microsoft ce Satya Nadella also reacted with a post on LinkedIn, saying as AI gets more efficient and accessible, we will see it's use skyrocket, turning it into a commodity we just can't get enough of. So Deepseek has this edge, the edge being it can do all this a lot cheaper
than competitors. Did you expect to see such a strong reaction in the markets, this intense of a reaction.
This intensive a reaction, is a huge surprise, because the evidence that this AI startup might dethrone the leading AI
tech companies isn't completely solid yet. This has just kind of been some early signs that they've gotten traction on the app Store, that they're performing while on benchmarks, and sure that you know that this can be built faster and more cheaply, But I think the overall concern that we're clearly seeing, especially going into tech earnings over the next couple weeks, is question whether are the billions of dollars that are going into these models that companies have
been telling investors it's worth it. We are going to monetize this and all of this investment will eventually.
Be fruitful.
That's coming into question, and what we're going to see over the next couple of days is wondering what is going on kind of behind the scenes, how is this being monetized?
Does it have to be this expensive?
Jackie? There are still a lot of unknowns here right What are you going to be watching out for over the next couple of weeks.
There's so much excitement that goes into a newcomer really plowing into the space. When you have this landscape already really cemented, It's almost this like David and Goliath energy coming in and I'm going to be looking out for improvements and deep seeks apps how it uses this momentum. I'm going to be looking for any reactions from President Trump.
You know, Chat Gibt didn't exist when President Trump was in office the first time, and I'm curious if this is going to inform how he goes about policy with China in other ways that could have reverberations inside the tech industry and the AI community. Those two worlds, the policy in Washington and the AI community and technology in Silicon Valley, are far more intertwined today than they ever have been.
This is the Big Take from Bloomberg News. I'm Sarah Holder. This episode was produced by Alex Tie. It was edited by Tracy Samuelson and Lynn Dwong. It was mixed and sound designed by Alex Sugia. It was fact checked by Adrian A. Tapia. Our senior producer is Naomi Shaven. Our senior editor is Elizabeth Ponso. Our executive producer is Nicole Beemster. Border Sage Bauman is Bloomberg's head of podcasts. If you liked this episode, make sure to subscribe and review The
Big Take wherever you listen to podcasts. It helps people find the show. Thanks for listening. We'll be back tomorrow