Ever since ChatGPT launched at the end of 2022, conversations about what AI might mean for the economy and our portfolios have been hard to avoid. Even a major geopolitical shock like the 2026 Iran conflict proved to be a distraction measured in weeks.

I don’t claim to be an expert on AI, but I use it every day in my work and have spent time with two of the leading large language models: OpenAI’s ChatGPT and, to a lesser extent, Anthropic’s Claude.

I also follow AI developments through a handful of dedicated Substack publications, a podcast, and more traditional financial media sources.

Economists seem to broadly agree that, over the long term, if AI follows the path of previous technological advances, it should boost productivity, increase economic growth, and raise living standards, similar to the Industrial Revolution, the automobile, and the internet.

Of course, just as those technologies were ultimately a net positive, they were also highly disruptive for many workers and professions along the way, sometimes displacing entire industries tied to earlier ways of doing things. Disruption typically comes before stability.

Jobs are displaced, new ones emerge, and the transition can take years or decades to fully play out. The concern now is that AI may accelerate this process significantly, given how quickly the technology is advancing and how broadly it can be applied.

At the same time, it is worth approaching today’s narrative with a degree of skepticism. Silicon Valley has a long history of confidently predicting timelines that prove too optimistic. Self-driving cars, the internet of things, and 3D printing were all presented as imminent transformations.

There has been progress in all those areas, but slower and more complex than initially advertised. And a great deal depends on trust. Using large language models every day, I’ve found them to be powerful tools, but they are still prone to factual errors, limited memory, and reasoning that can become circular or inconsistent.

Although I use AI to augment and speed up some work (including editing this article), I’m not even close to trusting it with investment decisions or trade execution. Just last month, a prominent law firm had to apologize to a judge in a bankruptcy case after submitting filings that contained significant errors caused by AI hallucinations.

For investors, however, the more immediate question is not whether AI will reshape the economy. It is how much of that future is already reflected in today’s portfolios.

Many investors who ask about increasing their exposure to AI already have significant exposure through broad equity allocations. Companies like NVIDIA, Microsoft, Amazon, and Alphabet are among the largest weights in major indices and are deeply embedded in the development and deployment of AI.

Looking ahead, that exposure may grow even further. If companies such as OpenAI, Anthropic, or SpaceX become public in the near future as expected, they are likely to enter indices at a meaningful scale. Each could do so at valuations approaching $1 trillion, making an already large exposure even larger without any direct action.

For those interested in additional exposure, we have solutions, of course, but the challenge is understanding how much exposure is enough.

I’m not a strict efficient markets adherent, someone who believes prices fully reflect all available information, but I believe enough that I don’t feel compelled to make large wagers beyond what is already embedded in the market.

In investing, the risk isn’t usually missing the story; it’s overpaying for it or owning too much of it.