Speaker 1: There are moments in history where, when nobody expects it, something happens that makes everything jump out of the blue. And once it has happened, then everyone seems to ensure how predictable it was. For example, until the 2008 crisis, everyone bought housing like crazy. But when the bubble exploded, then everyone said how obvious it was that there was a bubble. Because, of course, how were the houses going to be so expensive? This is popularly known as judging the past bull. But of course, as economists, we are a little sly, we gave it a more technical name. The black swans. Friends, a black swan has just set off the alarms of Wall Street and the future of the American economy. The, until some time ago, largest company in the world, NVIDIA, has plummeted almost 20% in the stock market in a single day, and the rest of the companies of the Nasdaq have followed its trail. Many are already talking about the explosion of a financial bubble that has unfoundedly fed the US technological sector. And in this story, in this particular black swan, there is a great culprit with names and surnames. The People's Republic of China. China has just launched DeepSeek, a new artificial intelligence that threatens to dismantle the entire market of American giants like Microsoft, Meta or Google. These companies have invested tens of billions of dollars in their artificial intelligence models, but DeepSeek, China, has achieved the same with only 5 million. And, attention, with less than 200 employees in its team. What does this imply? Well, that only the salary of a Google executive is higher than what it has cost to develop DeepSeek. But also, thanks to being open source, this new artificial intelligence can be optimized to be even much faster than the American versions. Conclusion, the Chinese have just humiliated the best American technology. They have dismantled the oligopoly, but on top of that, they have done it by offering their application for free to everyone. In just a few days, DeepSeek has already become the most downloaded application in the entire App Store, and it has been such a surprise that even the CEO of Microsoft has surrendered to it.

Speaker 2: Seeing the new DeepSeek model is super impressive in terms of how they have really made an open source model that does this inference time calculation and is super efficient on the computer.

Speaker 1: And you will say, okay, Oscar, Chinese artificial intelligence is better than American, but is it as serious as talking about the explosion of a bubble in Wall Street? Well, yes. These are the most valuable companies in the United States. The first nine are technological with billion-dollar investments in artificial intelligence, and the tenth is Warren Buffett's insurance company, which, in turn, has billion-dollar investments in these technological companies. The real problem has come when China has appeared with an artificial intelligence just as good as the Americans, but that costs less than the salary of a CEO of these companies. That is, this allows the Chinese to offer IAEA services at tiny prices with which the United States cannot even compete. Conclusion, all these investments from the big American companies to the trash. Wall Street explodes and uncomfortable questions begin to arise, because if the Chinese have been able to get a competitive artificial intelligence for the cost of basically a churro, where has all the money from American technologies gone? Are Silicon Valley programmers a big fraud? Do the American big tech have a future or are they going to collapse? Well, look at what Nayib Bukele himself has commented on.

Speaker 2: So more than 95% of the cost of developing new IAEA models is pure superfluous spending?

Speaker 1: Due to the hype of artificial intelligence, right now American stocks are trading with hyperinflated pairs. These multiples were justified by the great benefits that artificial intelligence promised to generate. But of course, China has changed everything and the disaster can be capitalized. Well, a moment, a moment, because it is possible that you have not heard anything of what I just said. You already know that in Visual Economic we do not usually use technical words of investment such as but or multiples, among other things, because it is that in institutes financial education is not explained. And if we use it, many people would not understand it. However, this is wrong. We are a channel of economic and financial information with a team quite well prepared in finance and in the world of investment. And, well, we have realized that we have the ability to transmit our knowledge to all of you. It is practically a moral duty, so how are we going to do it? Well, at the beginning, in the Visual Factory club we have prepared a series of free content for essential financial education. Come on, the ABC of what everyone should know to at least understand what we say when talking about the stock market or simply to understand the language of investment platforms. Because of course, if one wants to invest their money so as not to depend on public pension, then at least, I say, he will have to understand what the investment platforms put or simply what is put in the reports on companies that are on the internet. In the link in the description of this video and in this QR code you can sign up to the club and there you will receive the following deliveries of the financial education program that we have prepared. When you sign up, you will receive a dossier where we will explain in an understandable and simple way the most essential investment concepts. What are the multiples au pair, what are the indices, the ETFs, the derivatives, the funds, the BPA ... Well, in short, what you can imagine. And by the way, the sooner you sign up, the better, because as we send the episodes of the series by email and each delivery is unique, if you are not signed up at the time we send an email, it is possible that you will miss it and that you will no longer be able to have it. So the sooner you sign up, the better. But having said that, let's go with the questions of the video. How did DeepSeek manage to humiliate American big tech? What is going to happen with the stock market and the Yankee economy? Is this a game changer between China and the United States? The company that has fallen the most on Wall Street and that is feared the most is NVIDIA. Why? Well, because until now, artificial intelligence needed a brutal computing capacity to run their models, and NVIDIA was practically the only company capable of producing the necessary chips for this task. However, DeepSeek has managed to break this need. DeepSeek is so efficient that only a 50th of the computing capacity required by conventional IAs is needed. And what does all this mean? Well, on the one hand, it takes fewer chips to do the same, but on the other hand, that now not only NVIDIA chips are valid to train artificial intelligence, but the less powerful and cheaper chips of other companies, Chinese companies, will also be able to enter the market. Conclusion, a colossal outburst of demand, price and income from NVIDIA. And therefore, a blast in the bag. But do you want to know the best? The funny thing is that this could have been the direct consequence of a policy of shooting in the foot of the United States itself. Some time ago, the United States banned the export of NVIDIA chips to China. Why? Well, precisely because they thought that if the powerful chips did not arrive in China, then China could never develop an artificial intelligence that would compete against the Americans. But as you already know, sometimes it is better to have magic than strength, so the Chinese said, hey, if we do not have brute force, then we will have to use ingenuity to be more efficient. And well, they have achieved it. DeepSeek has changed everything. The question is, how have they achieved it? Well, although I am not an expert in technology, there seem to be three fundamental keys to this success. The first key is that DeepSeek uses less memory. That is, instead of using processes based on 32-bit quantization, they have done it with 8 bits. This translates into less precision in each process, but much more efficiency in terms of hardware use, which allows you to do 4 times more processes for each unit of memory. The second key to DeepSeek is that instead of reading each word, the AI reads phrases. That is, it uses a multi-token system of whole blocks that is twice as fast and almost as precise. But the third key to this model is already a usual one for economists. Specialization. Instead of making an AI that covers everything, the Chinese have made specialized modules that are activated only when necessary. This means that DeepSeek modules, which are not necessary at any given moment, are not activated, and therefore do not consume computational power either. The progress is clear. The question is, will this be enough to puncture Uncle Sam's most powerful sector? Well, it wouldn't be the first time. Back in the 90s, something very similar to what is happening now happened. However, the protagonist was not NVIDIA, but Cisco. Cisco was the company that produced everything necessary for internet connections. Routers, switches, servers equipment... And they did it at a time when the internet was the technological leap equivalent to artificial intelligence today. As you can see, as with NVIDIA, their actions were like foam. However, on March 27, 2000, their quotation collapsed, and since then it has never been able to regain its value before the fall. Why did Cisco fall? Well, for something very similar to what could happen to NVIDIA now. Everyone thought that Cisco was going to sell so many products that it invested massively in the company, but in the end, the expectation was met against reality. After the puncture of the .com, the demand for network equipment was not as high as expected. Competitors emerged, inventory excess, and well, what happened happened. Will NVIDIA have the same luck? Well, although everything indicates that yes. I'm afraid, friends, that in all this story there is a cat locked up.

Speaker 2: CHINESE STORIES?

Speaker 1: China has a long history of lies about technological advances. In 2021, it announced advances in its nuclear fusion reactor, presenting it as an artificial sun. I'm sure many of you remember. A sun that was capable of reaching extremely high temperatures. And while the project was real and important milestones were achieved, many experts pointed out that the advertising was very exaggerated. In 2021, Chinese state media also promoted a magnetic train capable of reaching 620 km per hour. However, it was discovered that the project was still in the experimental phase and that it did not have a viable infrastructure. With the IAEA, for example, the technological giant Baidu carried out demonstrations that were later found to be false of its autonomous driving system when pre-recording parts of the test. Therefore, what guarantees us that the information from DeepSeek is true? Come on, the Chinese practically say that they have trained their artificial intelligence, I don't know, with the chip of a toaster and a Russian washer. But it is known that its founder, Liang Wenfeng, accumulated up to 50,000 NVIDIA chips before the United States banned its export to China. What's more, there are reports that point out that, in reality, DeepSeek would have trained with 200,000 of the most advanced NVIDIA chips. In fact, this is the theory to which Elon Musk gives more verisimilitude. Although be careful, because here there are also conspiracy theories, such as that all this is a game to make money in the stock market. That is, DeepSeek's $ 7 billion matrix coverage fund could have invested in NVIDIA shares in the short term and have leaked false information to blow it up and make money. Yes, don't worry, because in the financial education glossary that you have in the description we also explain what it is to invest in the short term. The question, does all this mean that DeepSeek is a big fraud? Well, no, neither. As DeepSeek's source code is open, the engineers who have reviewed its code have seen significant improvements in efficiency. Now, the source code of training, which is the most important, remains hidden. That is, DeepSeek does propose some improvements in efficiency, but surely it is not much less cheap or as powerful as they promise. But even, well, even if we assumed that everything was true, that this artificial intelligence was so cheap, would it really be so catastrophic for companies like NVIDIA, Microsoft or Meta? Well, no. The truth is that no, not necessarily. I mean, I'm sorry for all the prophets of disaster who love to preach the apocalypse, but before jumping to hasty conclusions, we have to ask ourselves some basic questions, because how is it possible that Silicon Valley engineers did not realize that they could make much more efficient artificial intelligence? Are they idiots? Well, personally I doubt it. I am more inclined to think that if Big Tech has not been more efficient, it has been simple and plainly because they have not wanted to. The CEOs of Meta and Google have admitted on several occasions that they spend a lot more money than necessary to develop their technologies. Do they like to waste money? Well, no. What happens is that they prefer to spend extra money, but thus ensure efficiency, development, being at the forefront, to stay short. And here it is important to point out one thing. DeepSeek is a language model, but artificial intelligence goes much further. In fact, having language models is not even the goal of Big Tech. There are much more important, much more profitable artificial intelligence programs, such as those that are intended to be used in robots or autonomous cars. That is, this comparison would be like saying that Apple is going to the gallows because a Chinese company has released much better headphones. Well, yes, very well, the Chinese company has done a great job, but the iPhone is more important and still unbeatable. Better headphones will not make Apple fall apart, far from it. And that, friends, not to mention the elephant in the room, which also sometimes seems that we have gone crazy. Is DeepSeek really at the level of XGPT? Well, it is not so clear either. Most comparisons have been made against the O1 version of XGPT, and it does not seem that DeepSeek is clearly improving it, but it is that XGPT is about to release the O3 version, which is much better than the O1. Again, DeepSeek will be more efficient, but the avant-garde, the best, at least for now, will still be from the United States. But it is also true that DeepSeek has introduced improvements as groundbreaking as it promises, which prevents American companies from copying them and improving their models even more. That is, DeepSeek is open source, but the Meta or Google IAs are not open source. They can have the best of both worlds. So, well, it is not clear that neither Google, nor Meta, nor Microsoft are going to suffer a lot from this. And you will ask yourself, well, okay, maybe you are not so much, but what about NVIDIA? On the one hand, it is true that if what DeepSeek says is true, less computational power will be needed to train each artificial intelligence, but I ask you, are they not the most efficient computers every year? Are not mobile phones more efficient? Does greater efficiency mean less use of a resource? Because think about it, combustion cars have become more and more efficient, they have burned less and less gasoline, but has that caused the oil demand to plummet? No, quite the opposite. But I'll give you another example. Electric homes are spending less and less electricity, but has that caused less electricity to be consumed? Well, no, neither, quite the opposite. The cheaper a resource becomes, the more accessible it is, and therefore, the more we use it. The cheaper it is to create artificial intelligences, the more artificial intelligences will develop and the more powerful the current ones will become. And for this, even more chips will be needed and more powerful than those currently in demand. This is something that economists already know very well and that is known as the Jevons Paradox. And do you want to know my opinion? Well, frankly, I think that the collapse of Wall Street, well, it makes sense, it is reasonable. It is possible that many people have invested more in the account simply because of the hype of technologies, and that now the market is suffering a correction. However, this does not mean that companies are going to collapse or lose business. In fact, if I have something clear, it is that all this is going to be excellent news. If the cost of making an artificial intelligence has really been reduced so much, then the market will be more and more democratized. Small companies with few resources will be able to develop their own models, each of them adapted to a thousand different needs. And friends, that can be a huge economic leap. In any case, in the next video we will analyze more in depth the technological war between China and the United States, and above all, what the future may bring. That said, remember that in the link in the description and in this QR code here, I leave you access to the document where we explain basic concepts of investment and the series of financial education that we have prepared. Don't miss it, it will be great for you. And now the questions are for you. Do you think China will destroy the United States in the artificial intelligence market? Do you believe the data from DeepSeek? What will happen with the stock market? You can leave me your impressions in the comment box. And as always, don't forget that here, in Visual Economic, we release new videos every week, so subscribe to this channel and hit the bell so you don't miss any of our updates. If you liked this video, give it a like and see you in the next one. Greetings and see you soon.