- Telecom operators have thousands of central offices facilities that they no longer need for copper and DSL services
- These can be either sold or repurposed for new uses like colocation
- AI is changing the way operators think about what to do with these facilities
Telecom companies are sitting on thousands of old central office (CO) facilities sprinkled across the U.S. that could be just right for serving latency-sensitive artificial intelligence (AI) applications. And it seems they’re waking up to this fact as they continue efforts to retire the old copper network gear previously housed in these structures.
"Telecommunications operators have been presented with a unique opportunity to address the growing demand for data center infrastructure by repurposing old central office (CO) facilities as edge colocation centers," Steve Carlini, Chief Advocate, Data Centers and AI for Schneider Electric, told Fierce.
He continued: "With increasing demand for low-latency processing for AI inferencing, these COs are an ideal solution to house edge computing infrastructure, given that they are closely located to densely populated environments."
AT&T, Lumen Technologies, Frontier Communications and Ziply Fiber are among the operators which have started using their old COs for colocation and other cloud deployments.
“Initially, our focus was on assisting carriers looking to connect more closely with their customers,” Frontier’s Chief Strategy Officer Vishal Dixit told Fierce. “Now, building on that successful first phase, we're widening our attention to enabling hyperscalers, satellite providers and AI companies enhance their edge capabilities, moving beyond traditional colocation to a more dynamic edge strategy.”
Lumen, meanwhile, told us it is using COs for enterprise colocation. “Customers who come to us for these services are looking for that deeper presence into markets, as well as low-latency connectivity and proximity to their customers,” an operator representative said.
Seismic shift
But how did operators end up in this position in the first place? Well, demand for legacy copper and DSL services – which relied on active electronics housed in these central offices – has plummeted with the rise of higher speed fiber and fixed wireless access broadband offerings.
As an AT&T representative explained: “One of our Central Offices, built in the 1950s, was built for 100,000 copper landline subscribers and now only has 3,800. This was also used for copper-based DSL internet that was built for 80,000 homes, now only serves 50 customers. This is just one of many Central Offices we have across the country that have experienced the same shift in customer demand.”
Things have been moving in this direction for years now. And so, operators have been left to answer the question whether it makes more sense to repurpose these facilities or simply sell them off, Dell’Oro Group VP Jeff Heynen said.
On one hand, he noted, these facilities have the power, backup power, cooling and security needed for enterprise colocation. On the other, though, certain CO buildings may need retrofitting to make them more friendly to external customers.
According to Carlini, several upgrades may be required to convert a CO to an enterprise-grade colocation facility. These include replacing outdated systems with edge-specific modern architecture and adding enhanced power and cooling systems.
“You can imagine that larger operators serving major metropolitan areas have to find a balance between the additional revenue they might earn through colocation versus the immediate revenue gain they would have through a property sale,” Heynen said.
But AI could be changing this calculation. Analysts have indicated that over the next few years, AI workloads will transition from central training to inferencing. And there’s value in ensuring the latter can run as close to the user as possible.
“I think there are a couple things that have changed the dynamic,” Heynen concluded. “Certainly, there’s a clear demand coming for colocation facilities to support more edge processing for AI.”
UPDATE 10/22/2024: This story has been updated to include comments from Schneider Electric.