Emily Flippen

Now, you said in this report that

the right mental model is somewhere like three to five years in terms of the timeframe for investors in the sector.

That investor shouldn't get too caught up in things like the present-day cost of LLMs since the intelligence per dollar ratio for models has been doubling roughly every six months.

That goes over my head.

When I hear that coming out of your mouth, can you provide some more context as to what exactly that means and if AI capabilities keep improving at that same pace, where you expect value to accrue in the year ahead?

It's really easy for investors to forget about that cost curve, how quickly it can change.

We saw it change over the past two years.

You can think about how different it'll be in 2028.

We can circle back and have this conversation about the cost of models then.

I think the fundamentals and the impact that it has on a lot of the companies that are, say, building data centers or using the compute will look fundamentally different.

Donato, before we move on though, I do want to also ask how we can apply your technical expertise to an investing framework for our listeners.

You've helped lead the charge with AI changes here at The Fool.

If an investor is looking to evaluate the investments and performance of other companies as it relates to their AI ambitions and capital expenditures, what do you think they should be looking for?

I hope everybody listening does have that long-term mindset.

What I love about your response, Sonato, is it's so incredibly measured.

You're the head of AI here at The Fool, and it's easy for people to say, well, we're in a bubble.

Anybody who operates in the space of AI is probably over-enthused, over-investing, over-indexing, over-hyping.

But the reality is that the way that you speak about what you look for in an AI investment is hopefully exactly the same thing that our listeners look for.

It's something practical and purposeful.

to your last point there, maybe something that looks a little boring.