Drew Lentz

Chief Technology Officer, Signal Roam

← Drew Lentz

The fiber-AI finding is good marketing and a warning about who gets left behind

Drew Lentz · 2026-08-09

A rural road at dusk with a single utility pole strung with cable, fading into open country while a distant city glows on the horizon.

There's a finding making the rounds that fiber broadband spurs consumer AI usage, and it is starting to show up in broadband funding arguments. The short version is that Recon Analytics surveys 6,000 people per week about AI, and they seem to have found some connective conversation there where the people that have fiber optic connections are asking 28.5 questions per AI session. Fiber sessions are described as 50 percent more intensive than cable network sessions. By that math, that means it's only 14 questions or so per AI session for someone that's using a cable connection.

The fiber providers obviously win from this. It's an interesting read, because I don't know where this research comes from or how this correlates, but I do know the research came from the Fiber Broadband Association. There might be some advantage in why they're mentioning this rather than someone mentioning it at a telco or a cable expo.

So why is a good number for fiber a bad signal overall?

But the idea that if you've got a fiber connection, you will do more with AI is actually a big red flag for me because this just starts to talk about creating another digital divide.

We already know that there's a digital divide in the United States and across the world between people who can afford high speed connectivity and those who can't. Now you end up in a position where if someone can't afford high speed connectivity, it isn't only that they can't get the services everyone else has on a level playing field. It shows that even with advanced types of services like AI, if they're not operating this type of network, that puts them at a significant disadvantage.

When you think about areas like South Texas and rural markets, when you think about tribal lands and places that don't traditionally have fiber optic connections, it really puts the children at a disadvantage, and the students at a disadvantage. The Fiber Broadband Association stands to win from this, because that's good information for them to push fiber connectivity as far as they can. But it's tragic conversations for people who can't get fiber, who can't afford fiber. Even if it was available, they wouldn't be able to afford it.

This is just another signal that the digital divide continues to grow as more technologies and applications are introduced.

Does this change the math for a fixed wireless operator?

It has to. People really have to start considering what it is that their customers are doing on those internet connections. Luckily for fixed wireless, you've got products like Tarana out there that allow connectivity at huge speeds, offering gigabit services over wireless, which is phenomenal. Starlink is pushing higher speeds, and you've got different satellite providers coming into play that will be able to give you these high end speeds. This really seems like it would level the playing field for wireless operators, but we don't know what that looks like yet.

The research in this article was done for the Fiber Broadband Association, so I wonder whether Recon Analytics did any research into fixed wireless operators. For the people running Tarana, what do their numbers look like? What do those sessions look like for AI? I genuinely don't know what the current average utilization per user is. That'd be a great question for someone in that specific space.

Where does a gigabit connection actually fall apart?

Inside the house. It doesn't matter whether your internet connection is fiber or satellite or wireless if you've got poor Wi-Fi connectivity at home. My son recently moved into a new apartment and called me up because he needed to figure out how to get Ethernet cabling into his place. He plays a lot of games, and he wanted to make sure he had great connectivity so he could stream everything he did, because that's his main source of income.

Even though Wi-Fi is pervasive in people's homes, there's still that demand for a copper connection or a fiber connection inside the home. If you're dealing with poor connectivity inside the house, it doesn't matter what your pipe out to the internet is, because you still can't take advantage of the speeds if you don't have a capable setup.

What changes for a venue when guests start using live AI?

Latency is going to become the number one concern, and it already is. For a long time, people were mainly concerned with throughput, how much capacity they could get, how fast their speed test could go, and they really don't look at the big picture, which is the entire experience. In that experience, you have to look at latency as well, and start to understand what that connection looks like.

It's not just about how much of something you can get. It's how quickly you can get it, where it's coming from, and how you can consume it. Products like Orb can do a great job of monitoring your network capability and showing you an overall score, not just what your raw throughput is. Being able to monitor that with new tools is essential, because that's what gives you the big picture.

You definitely have to consider what your AP density looks like, especially if you're trying to serve people in large environments with low latency traffic. You want to make sure your cells are smaller, so that it's a better quality connection for everyone connected to that wireless network in venues and larger scale environments.

Is capacity really the reason to build?

People are looking at capacity thinking that they're doing more, but if you look at most AI sessions, they're all about latency. You're talking about exchanges of a few K, text going back and forth between client devices and services. That's the typical large language model use.

Then you look at people transferring larger files who want to use AI to build images or video or something multimedia, and they want an immediate response. They wanna be able to say, you know, make me a cat picture of a cat playing a piano, and they wanna be able to see that immediately. Not only do you have to have the low latency to handle that conversation, you have to have a big enough pipe to make sure that high resolution image or AI generated video shows up on the person's device immediately. It's a combination of the two, and in this age where people expect everything to happen immediately, you have to be able to provide the network that gives them that response.

Have you seen speed tests look great while applications failed?

That's just about any enterprise network. Sometimes you've got really great connectivity, but application hosting is done in different locations, and people are unable to use the applications they need. In large venues you see people happily getting online and offline while the applications run poorly. You have to make sure that the network is built up to handle that.

It goes back to the methodology I use. When I look at designing a network, I think about the devices, the people using the devices, the services that make those devices work, and the applications running on those devices. If you miss any leg of that survey, you're leaving yourself open for a poor quality network.

What you need is a tool that can monitor everything as best as possible. You're not just looking for throughput, and you're not just looking for jitter or latency. You're looking for true application quality. Once you find that something's running slow, there are a number of things you can do by leveraging SD-WAN or multiple providers to get your data where it needs to go as quickly as possible.

Why does proximity to processing matter as much as the pipe?

This whole conversation about people with fiber connections using AI more comes back to delivery. It isn't only the immediacy of the connection and getting the questions answered, it's the delivery of the service they're looking for, whether that's travel plans or weather updates or pictures of cats playing the piano. It really doesn't matter what the end user is doing, because they're going to use it for everything. We've already seen that as more people adopt AI, they use it for just about everything they can possibly think of.

If there's not that fiber optic connection providing low latency service, or a wireless connection from something like Tarana providing low latency and high throughput, you really are at a disadvantage. There are so many people in this country and around the world who don't have the ability to access those types of connections, and that just means this is going to drive the gap further.

This goes hand in hand with the proliferation of data centers. If you've got more data centers closer to people, then it's less latency to get that data processed. There are groups providing power as well as on-prem processing for individuals, so the AI is happening in your backyard, and Starlink is talking about building data centers in space, where as a customer you've got almost direct access to the data center.

So the latency conversation isn't just about the pipe. It's about how close you are to where that data is being analyzed and where that AI is actually being functional. That's one part of the conversation. The other part is what we're doing to make sure that people with low speed or mediocre connections still have the ability to process AI as quickly as possible so they can take advantage of it.

Do this

I'd encourage anyone who operates a service provider network to make sure they're offering a robust service and giving people the opportunity to do what they want and need to do on it. If your network isn't optimized for handling AI transactions, that's not just the latency and delay, it's also the throughput to deliver the immediacy people are looking for.

Sources