
- Verizon AI Connect marks the start of US telco’s push into the AI infrastructure sector
- Google and Meta have already booked AI Connect capacity to handle AI workloads
- Verizon claims it has a deal pipeline for AI workloads worth over $1bn “simply by leveraging our existing infrastructure”
- Operator has high hopes that it can play a key role in supporting AI inferencing at the network edge
While US tech giants suffered unprecedented share price carnage on Monday, with investors spooked by China’s generative AI (GenAI) newcomer Deepseek, Verizon, which only days earlier launched its AI Connect strategy saw its share price rise unobtrusively by a modest 3% to $40.68.
True, the uplift appears to be mainly because the US telco posted a decent set of fourth-quarter results, fuelled by strong wireless subscriber growth, but investors may well have also been reassured that AI Connect, which is aimed at supporting growing demand for AI workloads deployed at scale, largely involves “re-imagining” existing datacentre assets and fibre infrastructure rather than relying on massive capex splurges on spanking new infrastructure.
AI Connect will inevitably need some extra investment, but Kyle Malady, CEO of Verizon Business, when fielding questions from financial analysts on Verizon’s Q4 earnings conference call (as transcribed by Motley Fool), seemed to indicate that the money already spent (“to tweak assets”) and future capital requirements would not be significant, at least in the wider capex scheme of things.
“[For] some of these [AI workload] deals, what we’ll have to do is spend some capital to edge out our fibre a little bit to go meet people at the right datacentres and/or locations of business and then [retrofit] technical space,” he said. “So, we have money in that, but we’re ready to go with that. It’s all included in the [capex] envelope.”
If Malady was coy about AI Connect’s capex requirements, his AI bashfulness disappeared when talking about potential financial opportunities. “Our best guess at the moment is the TAM [total addressable market] we can sell into… what we have is probably $40bn-plus, but you see that every single day, there are new announcements of hundreds of billions of dollars going into this ecosystem,” gushed Malady. “We have a lot of assets that can play right into this. And frankly, we have well over $1bn-plus in funnel right now, and that's really only with our services that we have today.”
Google and Meta, noted Malady, have already purchased capacity on Verizon’s network with the intent of using it for their AI workloads.
“We are working closely with industry players like Nvidia to reimagine how telco functions can work along with AI workloads. We are starting this development at the far edge in a private network,” he continued.
AI inferencing at the edge
Looking to the future, Verizon has high hopes that demand will soar for AI inferencing at the network edge, which holds out the promise of reduced latency, real-time decision-making and bandwidth optimisation, and marks a seismic shift away from the current AI focus on training large language modules in large datacentre facilities.
Verizon cites consultancy McKinsey, which reckons 60% to 70% of AI workloads are expected to migrate to real-time inference by 2030, “creating an urgent need for low-latency connectivity, compute and security at the edge beyond current demand.”
Malady noted: “As AI shifts from training to deep deployment, the need for distributed computing will become increasingly important for real-time decisions and predictions.”
Warming to his edge AI compute theme, he added: “We have thousands of distributed telco facilities, many of which already have power, space and cooling available for this compute at the edge. As we take stock of our existing assets, Verizon’s ability to be a foundational player in the AI ecosystem is clear. The technical infrastructure required to enable AI is evolving.”
Malady flagged, too, that Verizon has between 100 and 200 acres of undeveloped land, “some currently zoned for datacentre build, and much of it in prime datacentre-friendly areas”.
Through a new strategic partnership with Vultr, a GPU-as-a-service (GPUaaS) provider, Verizon will initially deploy GPUaaS infrastructure in one of its datacentres and tap into its high-capacity fibre network for distribution. “We anticipate helping to broaden their reach and enable our mutual customers with AI training and inference capabilities at the edge over time,” stated Verizon.
Earlier this month, SK Telecom, arguably the world’s most advanced telco when it comes to AI adoption, launched a GPUaaS offering targeted at enterprise and government customers in South Korea from its AI datacentre in Gasan, on the outskirts of Seoul. Expect to see further announcements from telcos around the world this year about new revenue-generating opportunities linked to the AI boom.
- Ken Wieland, Contributing Editor, TelecomTV
Email Newsletters
Sign up to receive TelecomTV's top news and videos, plus exclusive subscriber-only content direct to your inbox.