| More Articles | Free Reports | Premium Services Unless you live under a rock, you saw what happened yesterday. The battle for AI supremacy took a wild twist. Chinese AI startup DeepSeek booted ChatGPT from its top spot as most-downloaded free app in Apple’s App Store. Its latest model – dubbed R1 – is as good as ChatGPT’s most powerful “reasoning” model. And it’s not sequestered behind a $20-a-month paywall. Oh, and DeepSeek-R1 cost a fraction of what it cost to build and train ChatGPT. Allegedly, the Chinese model cost only $5.6 million to build and train, while OpenAI spent $80 million building and training its model. That’s like the difference between Warren Buffett’s frugal home in Omaha, Nebraska and Jeff Bezos’ adjoining penthouse apartments on 212 Fifth Avenue in Manhattan. One is practical. The other, excessive. It also uses cheaper, less sophisticated chips than the newest Nvidia chips that ChatGPT, Elon Musk’s Grok, Meta’s Llama, and other U.S.-made AI chatbots use. DeepSeek-R1 matched the performance of these more established models with about 2,000 of Nvidia’s lower-performance, export-restricted H800 chips. That compares with the 10,000 state-of-the-art A100 and H100 Nvidia chips ChatGPT uses. The prospect of more efficient AI models that don’t need Nvidia’s most advanced chips was enough to send its shares down 17% yesterday. This wiped out about $600 billion in market value. That’s the biggest one-day dollar loss of any stock in history. And the game changing nature of DeepSeek hasn’t been lost on one sharp-eyed Freeport Navigator reader. As Peter A. wrote me on Saturday… Hi Charles I was reading about DeepSeek R1 and its amazing capabilities. Potentially, it could change the whole AI game… no extraordinary electricity requirements, no super-sophisticated computer chips, open-source code so no one can capture super-huge economic rents from proprietary IP. Do you have any thoughts on this new development? Well, Peter… in fact, I do. So today, let’s discuss them and what this means for your investments. | Recommended Link | | | | Jeff Clark has spent 40+ years studying market behavior and he’s traded through 11 presidential inaugurations. In that time, he’s uncovered a pattern that could be the key to potentially doubling your money weekly during Trump’s first 100 days. So while Trump’s aggressive orders send shockwaves through the markets… Jeff’s pattern could show you how to profit on both sides of the action. Until midnight tomorrow, catch the replay of his timely event, The Most Profitable 100 Days of Your Life, and see how this pattern could help rewrite your financial future. Stream the event free right here. |  | | DeepSeek Has Massively Democratized AI The first important takeaway is that DeepSeek has massively democratized AI. That holds true even if some of its creators have exaggerated its cost savings. Its model almost certainly cost more to develop than the $6 million dollars its creators claim. We should always take news coming out of China with a grain of salt. But even if the price tag is higher than advertised, DeepSeek has found ways to train and operate AI models that use less electricity and less computing power than “Made in America” AI. I’m no computing expert. Silicon Valley venture capitalist and the guy behind the first commercial Web browser Marc Andreessen is. And he says… DeepSeek-R1 is one of the most amazing and impressive breakthroughs I've ever seen – and as open source, a profound gift to the world. Open source means DeepSeek’s code is freely available to tinker with. That’s great news for researchers, developers, and smaller start-ups who – unlike Microsoft or Amazon – don’t have tens of billions of dollars to throw at AI development. They can leverage DeepSeek’s work and adapt it for their own use. This encourages programmers to work to improve the model. And if developers can build and operate AI models more cheaply, it breaks the grip mega-cap tech stocks have had on this new technology. Smaller companies – even mom-and-pop outfits – will be able to harness cheap AI to boost their productivity. The Real Winners of the AI Boom As I wrote about in these pages last month, the real winners of the AI boom won’t be Microsoft, Google, Meta, OpenAI, and other tech behemoths that are spending hundreds of billions of dollars on AI infrastructure – data centers, hardware, chips, etc. Elon Musk’s AI division, xAI, has spent about $4 billion on Nvidia chips for its gargantuan AI supercomputer, Colossus. But it’s no closer to profitability than its rivals. It’s running up billions of dollars in expenses and generating revenue just tens of millions of dollars. No. The real winners will be the companies that use the technology to build new businesses or run existing ones more effectively. We saw the same phenomenon play out during the Internet boom. The companies that built the hardware and software foundations of the Internet weren’t the big winners. The lion’s share of the profits went to Google, Amazon, Facebook (now Meta), Netflix, and other companies that built business models around this infrastructure. There’s no reason to think it will be any different with the AI boom. But that’s not all. There’s a geopolitical angle to DeepSeek’s insurgency, too. Sputnik Moment Its release isn’t the first time America’s tech titans have been caught with their pants down. On October 4, 1957, the Soviet Union shocked Americans by launching Sputnik 1, the world’s first manmade satellite. The resultant anxiety was compounded a month later when the Soviets launched Sputnik 2, carrying Laika, a stray dog from the streets of Moscow, into orbit. Laika became the first Earthling to go into space. We know what happened next... Faced with the prospect of Soviet space supremacy, the U.S. poured unprecedented resources into research and development. It set itself the goal of being the first country to put a man on the moon. In 1958, it created a new space agency, the National Aeronautics and Space Administration (NASA). Eleven years later, in 1969, two American astronauts, Neil Armstrong and Buzz Aldrin, became the first humans to set foot on the moon. Expect something similar to happen in this race for AI supremacy between the U.S. and China. America cannot – will not – cede AI supremacy to the Chinese. If you think ChatGPT, Gemini, Claude, and other American-made AI models have political biases, you ain’t seen nothing yet. Just for grins, I downloaded DeepSeek and asked it a few questions it really didn’t want to answer… 
DeepSeek doesn’t like questions about Chinese Communist Party atrocities That’s no surprise coming from a highly censored Communist dictatorship known for elaborate propaganda. And if the U.S. government considered Chinese social-media platform TikTok a security risk and a potential source of information warfare, it’s hardly going to tolerate Americans pouring their hearts out to DeepSeek. So, what does this all mean for us as investors? Should You Buy the Nvidia Dip? If we can build AI models more cheaply, with less sophisticated chips, does that mean that Nvidia is about to hit a brick wall? Not necessarily. Nvidia still makes the chips the Chinese have used to build DeepSeek. And the democratization of AI means we’re about to see more demand from smaller players entering the scene. That’s how capitalism works. As the costs to build and train AI models drops, more players will start creating more and more AI models. Some may use less sophisticated chips from Nvidia and its rivals. But its high-end chips will still be in demand from folks who want the best performance possible. So, if you own Nvidia shares, there’s no reason to panic. But this is as good a time as any to make sure your position size is reasonable. If Nvidia has grown to become a disproportionately large chunk of your portfolio, look for ways to spread your bets and diversify. What about buying Nvidia on the dip? The drop in its share price, while gargantuan in U.S. dollar terms, only took the shares back to October levels. Going into yesterday’s rout, Nvidia traded for 31 times sales. That’s roughly 4 times the price-to-sales of the VanEck Semiconductor ETF (SMH), which invests in a basket of the world’s top chipmakers. Any stock priced that aggressively becomes the proverbial bug looking for a windshield… because so much of its growth potential is already reflected in its price. Yesterday, that windshield appeared in the form of DeepSeek’s new model. Nvidia is still one of the great growth stories of our time. But it’s also darn expensive. So, if you’re itching to buy, use moderation.. Average in with a series of small positions rather than trying to call the exact bottom with a big purchase. To life, liberty, and the pursuit of wealth, | 
No comments:
Post a Comment