Can AI Trade Stocks?

0:00Today on the AI Daily Brief, can AI trade stocks? Before then in the headlines, a bunch of indicators that AI is making oodles of money. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

0:18All right, friends, quick announcements before we dive in. First of all, thank you to today's sponsors, KPMG, Blitzy, and Banta. To get an ad-free version of the show, go to patreon.com slash ai daily brief. And if you're interested in sponsoring the show, shoot me a note at nlw at breakdown.network. Welcome back to the AI Daily Brief Headlines Edition, all the daily AI news you need in around five minutes. If you've had your ear to the ground, you might have heard an increasing dint of concern around bubble-type dynamics around AI. And this has a lot to do with broader market things that are happening right now.

0:49SPACs are back on the table. Crypto treasury companies are all the rage. So there's a general sense, I think, of heatedness creeping into markets. And given how much AI drives a lot of market narratives, it's natural for it to get caught up. However, if there is one takeaway from this set of headlines, it's that bubble or not, AI is making a lot of money. Yesterday, Meta did their quarterly earnings and just absolutely crushed analyst expectations. They reported 22 % revenue growth and$18 billion in quarterly income. And what they're going to do with all of that is very clear. The company reiterated plans to double infrastructure spending this year to reach up to$72 billion in CapEx.

1:26They forecast a similarly large increase next year, writing that they've decided to, quote,

1:37CFO Susan Lee noted that although external financing is being considered, Meta expects to fund most of the build-out from cash flows. Now, one of the things that's really important about this is that this is not just Meta doing well over here and doing AI over there. There is a direct connection. On the earnings call, Zuckerberg said that the ads business is already seeing a meaningful amount of revenue from new-gen AI features. And even more than that, overall, he said that the strong performance this quarter was largely thanks to, quote, AI unlocking greater efficiency and gains across their ad system.

2:06In other words, this has met his business doing better, not all of a sudden people showing back up on Facebook. Now, Bloomberg is still giving column space to analysts with concerns about AI investment, but you're seeing more and more qualifiers being added to the comments. Gabriela Santos, chief strategist for the Americas at JPMorgan Asset Management, said, If a company is saying maybe we'll see an AI benefit in five years, we're no longer at the point where that will get a pass. Valuations are mattering more and more, and valuation especially matters if a company can't grow sales as fast as expected or as quickly as CapEx.

2:37And yet, that's the key difference for the hyperscalers this year. Sales are growing alongside CapEx, and most of these firms are seeing near-immediate returns on their investment. Regardless, the market has spoken, with Meta surging 10 % in after-hours trading. Rogue Google Cloud's Brian Beal, Meta can afford to hire as many billion-dollar people as they want to. Microsoft also gave a banger of an earnings call. Now, Microsoft was going to be one of the interesting AI companies to watch this quarter. Over the first half of the year, the narrative was that Microsoft was pulling back from excess cloud contracts, suggesting that demand projections were a little soft.

3:11Executives brushed off the rumors, explaining that a handful of canceled contracts were just volatility on the margins rather than a big strategy shift. Well, turns out there was absolutely nothing to worry about, as Microsoft delivered a huge outperformance to close their fiscal year. Company-wide revenue grew at 18%, and income rose by 22%. Now, those are huge numbers on their own, but the even bigger news came from the Azure cloud division, which was reported separately for the first time. Azure sales grew by 39 % for the year to reach$75 billion. That's around a quarter of the company's overall revenue, and now within striking distance of AWS, which was a$112 billion business over the past year.

3:46Said CEO Satya Nadella, cloud and AI is the driving force of business transformation across every industry and sector. Microsoft saw an 8.5 % move in overnight trading, enough to make it the second company in history to reach$4 trillion in market cap. Now, just for some comparison, the chart you're looking at is quarterly cloud revenue added, basically how much cloud revenue grew by each quarter. This past quarter saw Microsoft add more than twice as much as any previous quarter in history. Now, also as an aside, someone pointed out that if you're wondering if Apple's AI strategy is hurting it, looking at 12-month stock performance, and this is even before those overnight results, while Meta is up 50%, Alphabet is up 15%, Amazon is up 26%, and Microsoft is up 21%, Apple is down about 5%.

4:31So yes, the market is punishing Apple for, among other things, its lack of an AI strategy. And yet this whole money-making thing is not just in the public sector. With everyone talking recently about how fast Anthropic has been growing, sources have leaked OpenAI's revenue as well. The company has hit$12 billion in ARR, meaning they're on a billion-dollar-a-month pace, which is up from$500 million at the end of last year. That's also a big jump from the last revenue check-in we had from OpenAI when they claimed$10 billion in ARR at the end of May. In addition, OpenAI said that they're now seeing 700 million weekly active ChatGPT users, up from$500 million in late March.

5:08That puts OpenAI on track to beat their revenue forecasts from the beginning of the year. They had forecast$12.7 billion for the full year, so with five months left, they're comfortably on pace to blow out that estimate. Keep in mind that when those numbers first emerged, the prevailing take was that the forecast was somewhere between optimistic and delusional. OpenAI had one of the best years for any startup ever in 2024, bringing in$4 billion in revenue. Tripling from such a high base was basically unprecedented, but as we've seen, we are living in totally new times. On the other side of the balance sheet, OpenAI is also ramping up much faster than expected as well.

5:40They've increased their cash burn projection to roughly$8 billion this year, an increase of$1 billion. And yet, even with all of this, the story that a lot of the chattering class is talking about is Anthropic closing the gap. Saster points out that Anthropic is at 4x growth in seven months, which actually I think the number is higher than that. I think it's actually 5x, versus OpenAI's 2x growth in that same period. The bottom line they write, this is one of the most dramatic catch-up stories in enterprise software history. Anthropic has closed a 20x revenue gap to 2x in just three years, growing 10 times faster than OpenAI.

6:13While OpenAI maintains leadership through consumer scale, Anthropic's enterprise-first strategy and superior growth velocity suggests the revenue race will be incredibly tight by 2026 to 2027. Also, they point out, the market is clearly big enough for multiple$10 billion plus AI revenue players. Now, we barely scratched the surface on what the actual revenue profile is going to be. But I think that the thing here is, yes, the catch-up story is impressive on Anthropik's part, but what it's come from is not them out-competing OpenAI on things that OpenAI was already leading in. It's them carving out the central place in the biggest growth area, which is agentic coding.

6:49This massive inflection point that has happened over just the last couple of months is being driven in large part by the massive expansion of coding use cases. and currently, Anthropic is really leading that pack. And this is what makes the stakes of the GPT-5 release incredibly high. Now, we're still getting rumors left and right about this. Yuchen Jin writes, Heard from a little bird at OpenAI, GPT-5 is finally better than Claude at coding. Not just on benchmarks, but in real internal use over the past few days. If that's true, Anthropic can't stay quiet for too long. Claude 5 has to be released sooner.

7:22We'll see. For a lot of developers right now, the devotion to Claude is very, very high. But this battle is far from over. It is certainly something we will be watching. For now, though, that's going to do it for today's AI Daily Brief Headlines edition. Next up, the main episode. Today's episode is brought to you by KPMG. In today's fiercely competitive market, unlocking AI's potential could help give you a competitive edge, foster growth, and drive new value. But here's the key. You don't need an AI strategy. You need to embed AI into your overall business strategy to truly power it up. KPMG can show you how to integrate AI and AI agents into your business strategy in a way that truly works and is built on trusted AI principles and platforms.

8:02Check out real stories from KPMG to hear how AI is driving success with its clients at www.kpmg.us.ai. Again, that's www.kpmg.us.ai. This episode is brought to you by Blitzy, the enterprise autonomous software development platform with infinite code context. Blitzy uses thousands of specialized AI agents that think for hours to understand enterprise-scale code bases with millions of lines of code. Enterprise engineering leaders start every development sprint with the Blitzy platform, bringing in their development requirements. The Blitzy platform provides a plan, then generates and precompiles code for each task.

8:40Blitzy delivers 80 % plus of the development work autonomously while providing a guide for the final 20 % of human development work required to complete the sprint. Public companies are achieving a 5x engineering velocity increase when incorporating Blitzy as their pre-IDE development tool, pairing it with their coding co-pilot of choice to bring an AI-native SDLC into their org. Blitzy is providing a limited-time, 30-day free proof-of-concept for qualifying enterprises. The team will provide a 5x velocity increase on a real development project in your org. Visit Blitzy.com and press Book Demo to learn how Blitzy transforms your SDLC from AI-assisted to AI-native.

9:16That's B-L-I-T-Z-Y dot com. As a founder, you're moving fast towards product market fit, your next round, or your first big enterprise deal. But with AI accelerating how quickly startups build and ship, security expectations are higher earlier than ever. Getting security and compliance right can unlock growth or stall it if you wait too long. With deep integrations and automated workflows built for fast-moving teams, Vanta gets you audit-ready fast and keeps you secure with continuous monitoring as your models, infra, and customers evolve. Fast-growing customers like Langchain, Writer, and Cursor trusted Vanta to build a scalable foundation from the start.

9:52And look, as someone who lives in the world of enterprise procurement, I love how Vanta makes it easy to get compliance right. The last thing you need when you're trying to win that big deal is to have it scuttled by something that Vanta has solved for over 10 ,000 companies. Go to vanta.com slash NLW to save$1 ,000 today through the Vanta for Startups program and join over 10 ,000 ambitious companies already scaling with Vanta. That's V-A-N-T-A dot com slash N-L-W to save$1 ,000 for a limited time. Welcome back to the AI Daily Brief. Basically ever since ChatGPT was released, one of the things that people have been intrigued by is the possibility that it could help them make more money.

10:34Now, obviously some of that is in the form of new entrepreneurial endeavors. heightened productivity, more ability to output things than you were before. Alternatively, there have been folks who have recognized that a paradigm shift like this creates a lot of need and opportunity for people to be Sherpas of that change, to create courses, prompting lessons, anything like that that help people acclimate to this new environment. But for some, the pursuit of AI's opportunity to make them more money has been a little bit more simple. And the premise is if these tools are so smart, can they pick stocks and make investments?

11:08There are effectively infinite experiments to this degree. The one we're looking at now is from back in April and March of 2023, where Finder.com used Chatshiept to put together a dummy portfolio of 38 stocks that gained 4.9%, while 10 leading investment funds had an average loss of 0.8 % in the same period. In July of 23, Mustafa Suleiman, who was at the time still at inflection, but obviously would end up running Microsoft's AI endeavors, and he wrote a post for the MIT Technology Review about an alternative Turing test that wouldn't just be about whether AI could successfully imitate humans, but whether it could actually make money.

11:45Specifically, his proposal was that the new Turing test would be to see if AI can make a million dollars. Now, Mustafa was more thinking about an AI that could design a business from the ground up. He wrote, to make a million dollars, it would need to go far beyond outlining a strategy and drafting some copy, as current systems like GPT-4 are so good at doing. It would need to research and design products, interface with manufacturers and logistics hubs, negotiate contracts, create and operate marketing campaigns. It would need to, in short, tie together a series of complex real-world goals with minimal oversight.

12:14But of course, the first AI to actually become a millionaire was an unhinged bot released on Twitter that got Marc Andreessen to send it$50 ,000 in Bitcoin and then created its own meme coin that at peak had a fully diluted value of$600 million. Now, if you know anything about crypto and fully diluted value, you'll know that the simple math by which we get that$600 million valuation, i.e. price of token times number of tokens, doesn't hold up to the real world. And of course, in retrospect, Truth Terminal was mostly an interesting experiment on the path to broader endeavors. And yet still, that idea that maybe AIs could pick stocks and invest and actually make serious money has never really gone too far away.

12:53One new experiment that I noticed recently is called Comet Portfolio. It comes from Morgan Linton, the CTO of an e-commerce services company called Bold Metrics, who is giving Perplexity's new agentic browser a mission to make as much money in the stock market as possible. Morgan set up Comet Portfolio with$1 ,000 in a Robinhood account and some general instructions on how to research stocks. Importantly, Morgan basically said that he has pretty much no idea when it comes to investing, so it's not like he's putting his thumb on the scale here. Linton is documenting everything on X via a new account called Comet Portfolio.

13:26As part of the setup, the agent first figured out how to do research, real-time news checks, and present trade ideas directly in the Robinhood window for Linton's approval. This was all getting set up over last weekend. And on Monday morning, the grand experiment commenced. The first day was a little bumpy. Morgan writes, Okay, first agentic trading session with Perplexity Comet is done. It didn't go very smoothly, unfortunately. There seems to be a bug where the Comet assistant thinks it has filled out the amount field, but it hasn't. Also, it kept forgetting that it was doing the trading agentically and would switch to giving me instructions, and I had to remind it that it is my agentic trader.

13:59There's also a weird bug where it says it's going to buy one stock, but is actually navigating to a different page to buy a different stock. That being said, I'm a patient guy, and I know these are still the early days. I think I'm probably the first person to trade stocks with Comet, so I'm ready for all the bumps in the road. In the end, it did deploy close to$1 ,000 and built an initial portfolio. Linton was also able to create a custom dashboard in Perplexity Labs to track the portfolio. And right now, I think it's too early to really tell how this experiment is going to shake out. In some ways, in fact, there's really two totally different experiments going on here.

14:30One is an experiment, yes, in stock trading and how an AI would construct a portfolio. But the other is just about agentic capabilities. Those problems that Morgan shared from that first day weren't about the AI making weird stock choices. It was about figuring out the right modes of interaction between humans and agents in the context of this new perplexity browser. Now, when it comes to where the portfolio's allocations are so far, it is a pretty standard tech-heavy portfolio. It's concentrated in Amazon, NVIDIA, Microsoft, Meta, and Google, has a little bit in Berkshire Hathaway, and interestingly, it has about 1 % in ETH and 2 % in Bitcoin, which reads almost entirely like the agent went out and read some very, very standard conventional wisdom financial advice and constructed the portfolio as such.

15:14And so far, that seems like one of the limitations here. Basically, unless you prompt the agent to really go out there and do something different, it's just going to be an instantiation of conventional investing wisdom, which is going to put together a pretty standard portfolio. Now, there is another interesting experiment, however, going on over Reddit, where a high school kid named Nathan Smith gave ChatGPT$100 to trade with. However, instead of giving the chat bot the goal of picking the best stocks overall, he gave it the constraint of only trading smaller companies with less than$300 million in market cap.

15:44Interestingly, the experiment initially was something different. At first, it was a six-month experiment to see whether ChatGPT or DeepSeq could beat the market with a micro-cap portfolio. However, in the first week, DeepSeq really underperformed. DeepSeq underperformed the Russell 2000 by 20%, while ChatGPT in the same period actually beat the Russell 2000 by a little under 7%. And so the experiment refocused to just concentrate on ChatGPT. Now, part of what makes this experiment more interesting is the difficulty of investing in small cap companies. As you've already guessed, the major index for this segment is called the Russell 2000, which is sort of the small cap equivalent of the S &P 500.

16:21The range of companies in the 2000 small caps that make up the index is enormous, from biotech firms to manufacturers to cosmetics. Quality in this area is wildly varied. It's not uncommon for stocks to go to zero. And sometimes companies in this range can turn out to be complete frauds. All of which makes it kind of a more interesting place to have an experiment to see how AI can do. Now, after four weeks, ChatGPT is performing really well. The chatbot is up 23%, while the Russell 2000 Index is up only 3.9%. That is significant outperformance, suggesting that this isn't just luck. The chatbot chose four stocks to start its portfolio.

16:57It actually took a major drawdown early on in the process, losing almost 7 % on the Friday of the first week. After that, however, it reassessed and did research on 25 different stocks it might want to switch to. It determined that none of them beat the risk and reward ratio of its current holdings and decided to sit pat. Smith didn't like the choice, writing, Honestly, I thought it would be a complete breakdown after Friday, but it still feels confident in its thesis. Not many traders can lose 7 % of their account in a day and make zero changes. I was a little worried it would fall into a cycle of picking companies and giving up after like a week, but that hasn't happened at least yet.

17:28Now, the choice to hold was vindicated the following week as the portfolio regained the loss, And again, since then, the chatbot has made a handful of strategic trades and gotten itself up to this nearly 24 % gain. Now, maybe the most interesting part of this is that the chatbot is performing some fairly complex financial analyses to make its decisions. The portfolio is largely focused on biotech firms, which are notoriously difficult to trade. Doing well in that corner of the market requires a huge volume of research to figure out which drug trials are coming up, their likely outcomes, and things like that.

17:56ChatGPT seems capable of figuring out what these catalysts are and making some well-informed guesses about future outcomes. Greg Eisenberg was blown away by this experiment and wrote a long thread about the implications. One of his concerns was actually something that former SEC chair Gary Gensler brought up last year in front of Congress. Greg wrote,

18:27Now, I'm not so sure about this one. I understand the concern, and I think it's a legitimate concern. However, my instinct is that there is a fundamental inflection point that we have not hit yet. That is the difference between an agent using conventional investing wisdom to repeat best patterns and an agent actually throwing out the conventional wisdom to try unorthodox or contrarian strategies. Now, one of the things with markets is that everyone sees everything else that's going on, and so a contrarian strategy one day becomes conventional wisdom the next day. but my instinct is that people are not going to deploy AI en masse and certainly not autonomously until such time as they believe that the AIs are actually going to offer something different than just some carefully managed best practices.

19:08Basically, I don't think that people are going to hand over control to AIs in the phase where they would all just do the same thing. That was just one of the 14 implications that Greg saw from the study though. In another, he wrote, financial advisors will become AI prompt engineers. Instead of picking stocks, they'll craft the perfect instructions for AI systems and charge clients for their prompting expertise. The thing that's interesting about this is that this isn't exactly novel. It's pretty much what we're seeing with all professional industries in some ways. Your expertise still matters, but the way that you monetize that expertise is changing dramatically.

19:37And what's more, usually that's going to involve managing some number of agents that do the thing that you used to do, but with you as the overseer. Now, some of Greg's other predictions that relate to retail absolutely have the ring of truth to me. Greg writes, day trading will be completely dominated by AI within 18 months. Humans won't be able to compete with AI that can process earnings calls, news and social sentiment in real time. Every retail investor will have an AI trading assistant by 2027, not managing their money directly but whispering suggestions in their ear based on news they'd never read themselves.

20:05Investment newsletters will pivot to selling winning prompts instead of stock picks. Why give you fish when they can sell you the AI fishing rod? Someone will build a social network for AI trading strategies. Think GitHub for investment prompts where successful strategies get forked and improved by the community. I think that all of these are dead-on predictions. Retail investing in America is a very interesting space. Anyone who has watched the rise of GameStop and crypto and meme stocks knows that there is more than a little bit of what some might see as a devil may care, screw it, let's go kind of attitude, which others have characterized all the way as financial nihilism.

20:40The point is that this is a group that's going to be extremely receptive to trying anything that they think can give them an edge. And my guess is that there will be enough interesting opportunity in the intersection of AI and trading that those experiments will happen very, very fast. Sometimes when I'm doing this show, I have a feeling of how fast it's going to look extremely quaint based on how quickly the world has changed. And boy, is this show that. I really hope that you listen to this in the next couple of days after I release it, because I wouldn't be surprised if by September, half of these predictions from Greg and half of the things that we talked about today are just completely de rigueur and normalized.

21:16In any case, these are really interesting experiments. Thanks to Morgan and Nathan for publishing your experiments in a way that we can all enjoy and interact with. For now though, that's gonna do it for today's AI Daily Brief. Appreciate you guys listening or watching as always. And until next time, peace.