| 
Happy 2025! It may be a new calendar year – but we don’t think “new year, new me” will translate for Wall Street. Instead, we think we’ll see more of the same in ‘25. And that means AI stocks should stay red-hot. Indeed, throughout 2024, the market soared, led by huge gains for AI trades. The S&P 500 rallied about 23%. Yet headline AI stocks like Nvidia (NVDA), Meta (META), Palantir (PLTR), Broadcom (AVGO), and TaiwanSemiconductor (TSM) all more than doubled. Of course, recently, a debate has emerged about whether artificial intelligence has hit a proverbial “wall.” For example, famed venture capitalist Marc Andreessen noted that models like ChatGPT aren’t getting noticeably better. Others agree. But on the opposite side of the aisle, OpenAI CEO Sam Altman has said that “there is no wall.” In fact, just a day ago, it was reported that Altman claims the company “is now confident we know how to build AGI.” Nvidia CEO Jensen Huang has also disputed reports that industry progress has stalled. If Altman and Huang are to be believed, it certainly doesn’t sound like the industry has fizzled. So… who is right? The answer is important because it could determine where AI stocks go in 2025. If the pessimists are correct, AI stocks themselves could also hit a “wall” this year. However, we’re confident that AI models have decidedly not plateaued. In fact, we think they will continue to improve dramatically this year – even more than they did in 2023 or ‘24. Unlocking Unprecedented Capabilities With Logical Reasoning We believe three big factors will drive massive improvement over the next 12 months. First and foremost: the introduction of reasoning or inference into AI models should drive huge performance advancements in 2025. To date, most AI models have not possessed the ability to reason. That’s important because while such an artificial intelligence can be very good at answering questions and completing tasks it is trained to, it lacks the flexible thinking needed to draw meaningful connections between concepts or engage in genuine problem-solving. But an AI with reasoning can adapt its knowledge to unfamiliar contexts, demonstrating a level of cognition on par with human intelligence. And in late 2024, OpenAI launched the world’s first reasoning AI model with ChatGPT-o1. This model doesn’t just recognize patterns and provide pre-programmed responses. It actually considers a user’s question, thinks it through, and then answers. This “thinking” is a paradigm shift for AI. These models are now breaking out of their programming box, if you will, and starting to develop the capability to do more than ever before. That’s a massive deal. It’s the beginning of the journey toward creating AI that can solve problems that humans cannot – the journey toward artificial general intelligence (AGI). And we are expediting that journey here in 2025. As we mentioned, this headway technically began in late 2024 with the launch of ChatGPT-o1. But we believe that progress will pick up at a tremendous pace in 2025. Reasoning should increasingly become the norm architecture for foundational AI models. As it does, most AI will become much smarter, more capable, and more transformative. That is the biggest reason why AI models haven’t hit a wall and, instead, will continue to get way better with time. How Multimodal Data Further Unlocks AI's Potential The second big reason? The transformation and use of multimodal data in AI models. At present, AI is mostly trained on textual data. That is, bots like ChatGPT scrape the internet for millions of articles, comments, posts, etc., to “learn” about the world. But there is a whole universe of non-textual data out there, too. Just think of the myriad information embedded in things like podcasts on Spotify, videos on YouTube, or movies on Netflix. AI models are just now starting to incorporate such material into their training data sets. That means there is a huge opportunity for these models to become significantly smarter and more complex. And we expect this multimodal data expansion effort to start in earnest this year. Additionally, we’re confident that developers will increasingly use private data to improve their AI’s performance and capabilities over the next several months. Just like today’s AI models mostly use textual data for their training, they also mostly rely on public data to learn. But there exists a ton of private digital data as well, locked behind a paywall or proprietary to a certain company (like a social media platform or paid content site, for example). We expect that in 2025, AI developers will increasingly partner with database owners to give their models access to private data. Integrating such data should help AI models become significantly more refined over the next few months, allowing them to solve a much wider array of unique tasks. The Final Word on the AI Industry's Future In other words… Thanks to the creation of reasoning/inference architecture in AI models, the increased use of multimodal data in training data sets, and the inclusion of private databases… We firmly believe that AI models will get significantly smarter and more capable in 2025. That will help AI stocks to remain the biggest winners on Wall Street over the next 12 months. So, while it is a new calendar year, much should remain the same on Wall Street. And AI stocks will likely keep winning big. To help us find some of the best AI stocks to buy for the new year, we’re looking toward the world’s richest man – Elon Musk – and his AI venture for guidance. Learn more about xAI now to get your portfolio ready for the months to come. Sincerely, | 
No comments:
Post a Comment