- Telecommunications operators rank to extend optical networks for AI, but the analyst warns against short -term opportunities
- AI workloads are about to go beyond the data centers, with the edge like the next border
- Companies can be reluctant to enlist telecommunications operators for AI Edge tasks
The age of AI is on our doorstep, and telecommunications operators want to go into action by widening their optical networks for hyperscalers. The request for interconnection of the Data Center (DCI) is indeed massive, but it could also be a short -term opportunity, said research firm STL Partners.
This can be a shock, because operators like Lumen, Logix, Zayo and others double on the creation of longer long fiber roads specifically for data centers and their AI needs.
“We do not dispute that demand exists,” David Martin, principal analyst and telecommunications chief at Stl Partners, in Fierce Martin. The question is how long it will last when the generative AI continues to advance at a rapid pace.
Genai is about to move “more and more towards the edge,” he explained, just like the inference of the AI which is “non-critical”. This means that large data centers will not be the main hot spots in AI activity.
To be clear, Martin does not say that the opportunity of the Optical networking AI will disappear anytime soon, or that telecommunications operators should stop investing. But they should not put all their eggs in a single basket.
Looking at the next 2-3 years, this opportunity “can be much less than people think,” he added, because we will have exceeded the initial stage of the development of AI models.
“It will not be necessary of these vast concentrations of construction capacity of models in interconnected data centers in the same way as today,” said Martin.
Tellcos and The Ai Edge: a difficult battle
The IDC vice-president, Dave McCarthy, has also noted that success in AI will require going beyond the hyperscal data center, because the interconnection of private data centers and on-board locations “will be necessary to deliver the inference of AI in an efficient and secure manner”.
The potential in AI and the edge even lies in fiber deployments to the first. The future could see more workloads sensitive to latency and large -scale executed on FTTP networks, according to the main Avidthink Roy Chua.
But Martin is skeptical that the operators are ready to execute the workloads of the AI of soon.
It is not “an easy evolution of a game of pure connectivity to do something like that,” he said. “They can make the connectivity bit that connects these bits, but that’s not what we are talking about here.”
Not only do telecommunications need many specialized skills and updated infrastructure to successfully execute the workloads of the AI, but there is no guarantee that they will obtain enough request from companies. Companies may prefer to keep their workloads and data on the premises rather than on the telecommunications infrastructure, which has more than its fair share of security risks.
“This could be another version of the media shot that there was around multi-finish IT access a few years ago [with] Telecommunications thought in a way that it is an area in which they could settle profitably without having to invest too much or to increase too much, “said Martin.” And I think we tend to disagree with this. “”
The data circulating in telecommunications networks also have a fundamentally different structure of EDGE computer data. The first implies more “high data plan processing flows”, while the workload workloads do not move on the network in the same way, he noted.
“It’s a different business and you need different rack configurations and different cooling and all kinds of things,” concluded Martin. “But it is simply not the basic field of telecommunications.”
