- Verizon announced AI Connect, a suite of networking options for AI delivery
- The operator is one of many looking to cash in on growing AI demand
- But hype doesn't necessarily guarantee success for CSPs, Gartner told Fierce
Ca-CHING. That was presumably the sound bouncing around in Verizon CEO Hans Vestberg’s head this morning when the company announced AI Connect, a new suite of networking offerings aimed squarely at the rapidly-growing artificial intelligence (AI) market.
In a nutshell, AI Connect is the vehicle through which Verizon intends to peddle its fiber, land and 5G assets to data center companies and enterprises hungry for AI connectivity. The idea, Vestberg explained on its Q4 2024 earnings call, is to use its existing assets to generate new revenue streams that can help offset legacy wireline declines.
Scott Lawrence, SVP and Chief Product Officer at Verizon Business, told Fierce Verizon AI Connect isn’t just a repackaging of the edge compute capabilities it launched with AWS a few years back like we'd originally suspected.
Instead, it will build on those capabilities, providing access to the excess space, power and cooling capacity it has at its thousands of existing data centers and retrofitted central office (CO) facilities, as well as access to more than 100 acres of undeveloped land it has in key data center markets. Customers can also tap into Verizon’s local (i.e., Fios), metro and long-haul (see also, OneFiber) fiber assets for dark fiber and wavelength connectivity.
Why is Verizon doing this? Well, while much of the oxygen in AI conversations is consumed by discussions about GPUs and compute, “the role the network has in distribution of AI workloads at scale cannot be underestimated,” Lawrence said. All in all, Verizon is chasing what it sees as a $40 billion total addressable market for telco AI delivery.
If this all sounds vaguely familiar, that’s because it is.
All aboard
Gartner analyst Susan Welsh de Grimaldo told Fierce that by 2027, half of communications service providers will have redesigned their optical transport architectures and added fiber to cater to generative AI’s bandwidth and latency requirements.
And indeed, Verizon is already in good company chasing the AI networking opportunity. As we noted in October, AT&T, Lumen Technologies, Frontier Communications and Ziply Fiber have all also started using their old COs for colocation and other cloud deployments.
Additionally, both Lumen and Zayo are tackling the long haul fiber transport needs of data center operators as AI grows. The former has already closed $5 billion in hyperscale networking deals and has another $7 billion in the pipeline.
Meanwhile, Zayo said this week said it is building 5,000 route miles of new fiber between now and 2030 to meet skyrocketing demand for AI networking. It added it already closed more than $1 billion in AI-related networking deals in 2024 and has another $3 billion in the pipeline.
Lawrence said Verizon is already positioned in a close third place when it comes to third-party data center connectivity, with a $1 billion-plus funnel of interest in its core assets that is “actively growing.” There’s also “tremendous interest” from companies interested in tapping into Verizon’s existing data centers, retrofitted COs and undeveloped land in key data center markets.
Indeed, the operator today announced a partnership with Vultr, through which Lawrence said Vultr is deploying GPU assets at Verizon’s edge locations. It also announced an expansion of its partnership with Meta to include network infrastructure.
“We’ve actually delivered metro fiber solutions that are reflected in our fourth quarter results,” Lawrence said.
“We’re not announcing deals that will take years to build into the future, we’re actually delivering it today. We’re meeting the market where it is today while building for the future," he said.
Vultr is using CO facilities that have already been upgraded, but Lawrence said plenty more can be made available with retrofits within six to 12 months. He added each CO can provide anywhere from 2MW to 10MW of available power, and water infrastructure for liquid cooling can be added as needed.
Money for those upgrades (as well as fiber extensions to near-net locations), he noted, has already been factored into the guidance Verizon provided to investors.
5G differentiator
But there’s one thing Verizon has that Lumen, Zayo, Frontier and Ziply don’t – a 5G network. Lawrence said that will particularly come in handy when the expected transition from centralized training to inferencing at the edge comes to fruition.
“You’re going to have a highly distributed compute environment, but you’re also going to see those models range all the way from the cloud, to edge to the device. So, activating our macro 5G network, specifically with network slicing, will also allow us to create segmentation across our asset base,” he said.
Additionally, Lawrence noted Verizon is working with Nvidia to incorporate GPUs into its private 5G offerings, bringing compute on-prem for those who need it.
No guarantees
Welsh de Grimaldo said for operators it “makes sense to use existing assets to drive new value propositions and revenue – especially as they are going to bear the costs of supporting growing traffic for AI applications.”
The catch, though, is that just because AI is hyped and growing rapidly does not mean operators will be successful in their efforts.
“It’s early stages – you are not going to learn how to drive growth if you are not participating as a key player in the AI ecosystem,” she concluded. “Telcos like Verizon and others are taking their seat at the table, to learn and with aims to lead.