The Internet Is Changing Fast

**Jon Quast** (0:02)
The internet might be about to get weird.
Motley Fool Hidden Gems Investing starts now.
Welcome to Motley Fool Hidden Gems Investing. I'm your host today, Jon Quast. Joining me today are guests, longtime Fool contributors, Matt Frankel, and Tyler Crowe. Me and Tyler are switching seats today. But today we're talking about some mining stocks, as well as a $15 billion move from Intel. But first, I want to talk about this bit of news. You know, earnings season, always a busy time. And sometimes things get lost in the commotion. One of the things that I feel like got lost was when Cloudflare reported its second quarter earnings on August 6th. Now, this is a top 10 ranked stock in the Motley Fool Hidden Gems universe. But the report was great, investors responded fine, but buried in the earnings commentary on the call, co-founder and CEO Matthew Prince was talking about AI agentic traffic on the Internet, saying that for the first time, AI agents surpass human activity on its network in May. And the company had projected this to happen already, but it's happening faster than its timeline. And then they went on to say that they actually expect AI agent traffic on the Web to be 1,000 times bigger than human traffic within the next five years. And growth is something that is so important to investing. It's something that caught my eye when I saw this statement. But Matt and Tyler, I want to get your reactions just first here. What do you think about this statement from Matthew Prince?

**Matt Frankel** (1:48)
On one hand, I understand the logic, even though I think the 1,000x figure might be a little bit too ambitious. The agentic AI traffic, it increased by about 18x over the past 12 months, and 1,000 times in 5 years mean roughly quadrupling every year for the next 5 years. So that would actually be a deceleration from what we've seen. There's also the case to be made that human internet traffic has a natural ceiling. We can only look up so much stuff, we can only perform so many tasks at a time.
There's a reason that human-driven internet traffic has only grown at a low single-digit rate for about the past decade. Agents don't have this natural ceiling. It only depends on the cost of doing the work and how many tasks we give them to do. Obviously, there's no limit to the latter there. Costs are falling and usage is rising. A thousand X in five years, as I mentioned, it could be a stretch, but the costs are certainly heading in the right direction. The token costs, they've dropped by an order of magnitude over the past two years, and we've seen roughly a hundred X increase in usage. So if that pattern of decreasing cost continues, agentic AI usage could soar. I don't know about a thousand X, but a lot higher.

**Tyler Crowe** (2:55)
I want to kind of compare this to that. If you double a penny every day for a month, we have like $5.6 million. I mean, yes, the math says it works out that way, and doubling early numbers, like we've seen recently, makes sense, but eventually that pace does become somewhat unsustainable. We see this investing all the time, the S curve of investing, things accelerate, and then they eventually decelerate. We see it with companies, revenue trajectories all the time.
Look, just because agentic web traffic is happening behind the scenes doesn't mean it's without cost. And yes, costs are declining, but there is cost to it. And there will be some inflection point where the compute costs for all this agentic web traffic throttles growth. Someone, somewhere, looking at their AI cost spending will need to see an ROI on this. And, you know, for all this bot traffic, that's what they're doing. Eventually, they're going to have to see a return. If they're not seeing it, they're not going to be spending on it, otherwise, what's the point?

**Jon Quast** (3:51)
Well, it sounds like both of you, then, are a little bit skeptical on the 1000x number. And that's fair. If you believe that Cloudflare is blowing some smoke here, feel free to say so. But I do want to imagine here, I do want to project, I want to look into the future.
Let's say by 2030, if at least this is directionally right, that the share of agentic AI traffic is going to continue to increase at a large rate compared to the human traffic, will the internet look different in five years? And what changes would happen? What would be different from our perspective? What would it look like?

**Matt Frankel** (4:29)