Does fiber-to-the-prem really matter for AI and the edge?

0
6
  • It’s not just long-haul fiber that matters for AI
  • Last mile fiber broadband networks can also serve as creation hubs for personalized AI at the edge, FBA said
  • AI can also be applied within the network to cut costs via automation

It’s fairly obvious by now long-haul fiber and data center connectivity matter for AI innovation. But where do last mile fiber-to-the-premises (FTTP) deployments factor in? 

Network automation is perhaps the most apparent use-case, but fiber operators are also be well-positioned to house future AI edge applications.

According to Jack Gold, founder and principal analyst at J.Gold Associates, fiber has an advantage at the edge as it’s an area where satellite providers like Starlink can struggle.

“Since AI, at least for now, can’t run on local systems and must be sent up to the cloud for analysis, bringing latency down to a reasonable user level is critical for expanding the user base,” he explained.

If AI users have to wait “more than a second or two” to get a response, it can be pretty problematic. That’s true of most apps, Gold said, not just AI. “Starlink has a significant latency problem for things like AI, since the round trip to the satellite, and speed of network are vastly slower than a fiber connection, and more costly as well,” Gold added.

It’s true that latency isn’t a huge constraint for AI applications – yet. According to AvidThink Principal Roy Chua, much of the current AI workload involves chatting, the creation of audio, images and video as well as AI backend inferencing and reasoning, “which doesn’t push the latency requirement envelope much.”

But “over time, as we get more real-time multi-modal interactive capabilities from AI agents or avatars, then we could see the more latency-sensitive, large-throughput AI workloads favor being run over FTTP/wireline networks,” Chua said.

And as fiber broadband continues to proliferate, the Fiber Broadband Association sees plenty of potential for AI models to grow in the home.

“In a sense, as people feed the model with their data upstream, homes will become AI creation hubs for personalized models,” FBA wrote in its latest report on AI and fiber.

AI for telco automation

Then there’s using AI for FTTP network deployment and automation. 

Nokia this week unveiled the Broadband Easy platform, which leverages AI to provide operators “full visibility and control” of the entire fiber rollout process. Nokia said its platform can use AI to, for example, verify and accept the installation of fiber components, generate accurate inventory data, and provide on-site training and guidance to field technicians.

Analysts indicated the technology could be a saving grace for telcos riddled with rising deployment costs. 

“Many network operators are on thin margins as it is,” said Gold. “Anything that can lower infrastructure costs are beneficial to their bottom line, and help to accelerate upgrades.”

Indeed, we’ve reported how digital construction management is becoming a mainstay in deployments, as it can automate routine tasks and potentially place less strain on the broadband workforce (which is currently facing shortages).

Nokia isn’t the only vendor making strides in fiber network automation. Amdocs this month released FiberOne, which uses GenAI to speed up deployments, simplify billing options, among other things.

The autonomous network: fact or fiction?

The prospect of a fully autonomous network is far from a fantasy. On the mobile end, Beijing’s Tsinghua University, China Mobile and Brazil’s Vivo are already running their networks at the Level 4 level – without human oversight. And on the fiber side, European operator POST Luxembourg is working with Nokia to migrate its entire broadband infrastructure to a software-defined access network by the end of 2025.

Despite AI’s promises to bring about more efficiency, it can’t solve all pain points for operators. 

“The real bottle necks, like negotiations with pole owners, supply chain issues, and regulatory requirements cannot be overcome with AI,” said Recon Analytics Principal Roger Entner.

Nokia touted Broadband Easy can accelerate fiber rollouts by 20%, and while that’s “certainly true for a very particular slice of activities,” Entner concluded he’s not sure how big that slice is overall.

LEAVE A REPLY

Please enter your comment!
Please enter your name here