At a glance
- The pact comes months after Verizon signed a $1 billion fiber deal to supply Google with dark fiber linking the hyperscaler's data centers.
- Verizon said the AWS agreement "establishes a commercial framework to scale distributed compute and meet the real-time demands of agentic and generative AI workloads," according to Data Center Dynamics' report.
- For allocators underwriting fiber and edge-compute assets, the disclosed figures are signed contract values, not billed revenue, and the deal cannot yet be modeled per mile or per site.
Verizon signed an AI infrastructure agreement with Amazon Web Services worth more than $2 billion, announcing the deal Oct. 26 alongside its third-quarter earnings. The pact comes months after Verizon signed a $1 billion fiber deal to supply Google with dark fiber linking the hyperscaler's data centers.
Verizon said the AWS agreement "establishes a commercial framework to scale distributed compute and meet the real-time demands of agentic and generative AI workloads," according to Data Center Dynamics' report. The report gives no contract length, no committed volume and no list of sites the AWS workloads would use, which leaves the per-site economics of the agreement unstated.
For allocators underwriting fiber and edge-compute assets, the disclosed figures are signed contract values, not billed revenue, and the deal cannot yet be modeled per mile or per site. What has changed is the customer base: Verizon now has two named hyperscalers buying connectivity it already owns right-of-way for, and a third source of demand — carriers of AI inference traffic at network hubs — that Schulman says is converting into contracts. The financial bridge between those contracts and recurring connectivity revenue is the missing piece; the report does not separate construction or installation work from ongoing service revenue, and it does not say how much of the roughly $4 billion signed book is dark fiber, lit fiber or edge compute.
CEO Dan Schulman told the earnings call that demand spans the product set. "Customer demand across dark fiber, lit fiber, [and] our Edge compute locations is exceptionally strong," he said, describing it as "a significant growth opportunity" and "a rapidly growing TAM."
He said Verizon has "an active pipeline of additional hyperscalers, neocloud providers, and large enterprise clients who are in negotiations to have us supply them with either dark or lit fiber connectivity." Partners are also using Verizon's network hubs for AI inference at the edge, he said, and "that momentum is already turning into signed deals."
Schulman put the carrier's signed contracts at approximately $4 billion, a total the report does not break down by customer or connection type. It also does not say whether that sum includes the AWS agreement. He said the first long-term contracts are expected to generate hundreds of millions of dollars in revenue over their term, and that Verizon expects fiber and edge computing "to add meaningfully to our future growth." That gap between billions in deal value and hundreds of millions in expected revenue suggests the headline figures measure total framework value across multi-year terms rather than revenue near at hand, though the report does not define its terms.
The buildout rests partly on glass. The report notes that Verizon's ability to serve the hyperscalers is supported by a multi-year supply agreement with fiber manufacturer Corning.
The call also marked a year at the helm for Schulman, who outlined an AI-led "full reboot" at the third-quarter earnings a year earlier, shortly after joining. His predecessor, Hans Vestberg, said in 2024 that Verizon was best placed to maximize the potential of generative AI edge computing.
The next checkpoint is the fourth-quarter call, when Verizon could update the pipeline Schulman described, put a number on the edge-inference contracts he says are signed, or disclose further hyperscaler agreements.
What the $4 billion book does not show
Reported by the Data centers & fiber desk from Data Center Dynamics. Published . Corrections