Verizon announced a multi-year dark fiber agreement with Google worth more than $1 billion, marking the telecom's formal pivot to wholesale AI infrastructure. CEO Dan Schulman told investors the company expects to announce additional AI-related infrastructure deals by year-end, collectively worth multiple billions of dollars. The Google contract delivers dedicated optical fiber routes connecting data center campuses without shared public internet capacity.
Dark fiber agreements lease unused physical fiber strands to customers who then light them with their own transmission equipment. Google pays for exclusive access to the raw cable infrastructure, avoiding congestion from Verizon's retail broadband network. The arrangement lets the hyperscaler control latency and throughput between AI training clusters and inference endpoints without negotiating service-level agreements tied to Verizon's commercial offerings. Schulman declined to specify route geography but emphasized the deal addresses Google's need for deterministic network performance across multiple metro regions.
The structure signals a shift in telecom economics. Verizon historically monetized fiber through retail broadband subscriptions and enterprise WAN contracts, both requiring operational overhead for customer support and service guarantees. Dark fiber arrangements strip that complexity: Verizon provides conduit and cable maintenance while the customer manages everything above Layer 1. Gross margins compress but capital efficiency improves because the company no longer provisions active equipment at scale. For Google, the deal solves a capacity problem tied to model size. Training runs for frontier AI models now require synchronized data movement across thousands of GPUs, making shared infrastructure untenable. Leasing dark fiber lets Google deploy Dense Wavelength Division Multiplexing hardware tuned to its exact traffic patterns, a capability unavailable through standard transit agreements.
The $1 billion figure likely reflects a three-to-five-year commitment with minimum capacity guarantees. Verizon's existing long-haul fiber footprint spans roughly 47,000 route miles in the continental United States, much of it underutilized following the collapse of enterprise office demand during the pandemic. The company now repositions that stranded asset as AI infrastructure, competing directly with wholesale providers like Zayo and Lumen. Schulman's reference to a multi-billion dollar pipeline suggests active negotiations with other hyperscalers. Microsoft, Meta, and Amazon each operate private backbone networks but rely on leased capacity to bridge geographic gaps. If Verizon closes even two additional contracts in the $500 million to $1 billion range before December, the infrastructure segment could exceed 15 percent of total revenue by fiscal 2026.
Operators should watch for Verizon's third-quarter fiber backlog disclosures, expected in late October. If the company reports contract commitments above $3 billion, it confirms sustained hyperscaler demand and likely triggers competitor response from AT&T and Lumen. Allocators tracking the AI infrastructure trade should note that dark fiber agreements carry minimal churn risk once lit—Google will not abandon capital-intensive DWDM deployments after 18 months. The agreement also insulates Verizon from the margin pressure affecting retail broadband, where cable operators continue to underprice fiber-to-the-home offerings.
Google does not pre-announce data center construction timelines, but the company filed permit applications for four new campuses in Virginia and Ohio between March and June. Those projects total roughly 2.8 million square feet of raised floor space and align geographically with Verizon's densest fiber routes.