Doug Clinton

Investors use Claude and ChatGPT every day for investment ideas. We’ve seen at @iafunds that frontier AI models can already act effectively as investors, but as these models become a bigger part of investment decisions, the persistent alpha they generate should become beta — just part of the market. To make AI alpha market beta, we need to redefine the market with an AI-first approach. That’s why we’re launching the IA 500, the latest Intelligent Alpha benchmark for measuring the performance of frontier AI models at investing at https://t.co/0gt3MpDVUn. We test the ability of 10 frontier AI models to build investment portfolios, and the aggregate of AI’s selection makes up the IA 500. It’s the new market benchmark for the AI era. Disclaimer: The IA 500 is experimental and for informational purposes only. Past performance does not guarantee future results. Investments involve risk, including possible loss of principal. Additional performance methodology and disclosure information available on the IA 500 website.

5 Predictions for Markets in 2026 By Investor X, our AI investment committee at @iafunds 1 The 10-year Treasury yield will go toward 5% despite the Fed cutting rates. Our AI’s highest conviction call is that yield curve steepens dramatically with the 10-year living in the mid-to-high 4% range despite the Fed continuing to cut rates. Primarily, AI sees this as a structural supply issue with the government committed to increasing the deficit over the next decade, continued lack of QE, and lack of appetite from foreign buyers without a healthier term premium. 2 It will, finally, be a year where small caps beat large caps. Feels like small caps have been a popular call for the past couple of years. Not much different heading into 2026. AI is going back to the well on small caps, and it’s an even more challenging call given our AI’s view that long rates are going higher which is typically bad for small caps. Our AI thinks that flows will drive small caps as investors look for return beyond the large caps. 3 The AI bubble won’t burst in 2026. After two rough periods for the AI trade in 2025, our AI think the trade sustains in ’26 even though we likely get a 10% drawdown in tech at some point in the year. AI thinks the “danger period” for the trade is in 1H 2027 when we may find more supply/demand balance driven by data centers going online, and the circular financing deals become more challenged. 4 There will be a major cybersecurity breach driven by AI, and it will be a tailwind for cybersecurity stocks. A specific and unfortunate call by our AI, but one that shouldn’t be surprising. It’s more surprising we’ve only seen fairly contained agentic/generative AI-driven cyber breach yet. 5 Own hard assets for the physical economic: uranium, copper, silver. Gold has been the focus hard asset for the last couple of years, but our AI thinks uranium, copper, and silver will be the assets to own in 2026 as demand from AI data centers maintains their momentum.

Nvidia is playing the circular economy game and investing in its biggest customers. That’s a common tactic in bubble cycles, but I think it misses the absolute scale of AI. Nvidia generates $15+ billion in FCF per quarter. They have huge amounts of cash to put to work, and there are few opportunities to deploy that capital. Buying back stock or dividends are the last ditch tactics of mega caps. The charitable and possibly correct view is that Nvidia sees a way to capitalize a major customer, and that incremental capital could not only create a multiplier effect in terms of what OpenAI and its partners might spend with Nvidia, but it might also give Nvidia a massive stake in a company that could be worth trillions in the future. h/t @dylan522p

It’s not the top. It might be a local top, but not the top for AI. Think about what the top means. The top is when the absolute last and most price insensitive dollar has entered the market. Bulls are maximally levered, and they’re fomo-ing and yolo-ing into the most speculative stocks possible. When the last buyer enters, there are no more dollars to enter after that, so sellers take over as the dominant force. The bubble bursts. Do you really think we’re there? OpenAI hasn’t even gone public yet. Top callers point to many of the same things… h/t @itsTarH

It's the part of the bubble where the proudest rational actors say its over. It's not over until those actors are absolutely besides themselves screaming that everyone else is irrational. https://t.co/oXCXFRvthG

It pays little to be skeptical early. It can pay a lot to be skeptical later as long as you weren’t skeptical early.

Apollo is wrong. Technologically driven bubbles don’t depend on zero rates. The 10 year was between 4.5-7% during the dotcom era. 4% during the 60s Tronics boom. Tech bubbles build on grand potential that a modest cost of capital doesn’t impair. https://t.co/XcCUIaWasd

This is why bubbles are a feature not a bug of breakout new tech. Excess capital is required to subsidize expensive things for mass consumption until we’re hooked. Economics get figured out later. https://t.co/bNDr2UP0Kf

What is the company today that no one wants to invest in that will be worth $100b+ in a decade plus? Or is there so much capital available that nothing is going unfunded? https://t.co/LrVaxTNA27

Two kinds of AI apps: a. Require frontier intelligence. Handles the most complex tasks like stock picking, forecasting, scientific discovery. These tasks are where small alpha can be very valuable. b. Can be done with commodity intelligence. Handles simple tasks like in-app navigation/UI, customer service, etc. small alpha doesn’t matter here. Hard to believe models ever get commoditized for the first category. If there is small alpha to be gained, frontier models will unlock that value if in aggregate it’s economically viable. For commodity tasks, old models, small models, etc will rule.