Why it matters
The strategic read
Adds a connectivity receipt to the AI buildout, not just another campus-power headline. Lightpath is being pulled into two active gigawatt-scale hyperscale campuses, showing that AI-factory bottlenecks extend into route-diverse fiber, multi-terabit capacity and construction timelines that have to match power and data-hall delivery. The unnamed hyperscale customer limits transparency, but the primary source confirms the AI-driven demand, two 1 GW+ campus locations and delivery windows.
Source-supported terms
What was disclosed
- 01
Lightpath announced on July 8, 2026 that it will build fiber infrastructure for two hyperscale data-center campuses now under construction.
- 02
Lightpath said each campus is planned to exceed 1 GW of capacity.
- 03
The campuses are located in Saline, Michigan and Port Washington, Wisconsin.
- 04
Lightpath said the builds extend its AI-grade network to meet capacity demand driven by artificial intelligence.
- 05
Lightpath will provide triverse fiber infrastructure and multi-terabit capacity to the two facilities.
- 06
The Saline build is expected to be delivered by the end of 2026.
- 07
The Port Washington build is expected to follow in the second quarter of 2027.
- 08
Lightpath said both builds are being delivered in partnership with an anchor hyperscale customer.
- 09
Lightpath described the campuses as gigawatt-scale AI campuses requiring end-to-end connectivity across new construction, existing Lightpath network assets and strategic partner fiber.
- 10
Lightpath said the builds follow recent AI-grade fiber infrastructure expansions across Phoenix, eastern Pennsylvania and Columbus, plus its long-haul Columbus-to-Chicago corridor.
Disclosure boundary
What is known—and what is not
Primary Lightpath company announcement. Public materials disclose the AI-grade fiber-build award, two active hyperscale campus locations, each-campus-above-1-GW scale, triverse fiber infrastructure, multi-terabit capacity, expected delivery windows, anchor hyperscale customer framing and AI-driven demand rationale. Public materials do not disclose the anchor customer name, campus developers, total contract value, route miles, network-map details, service pricing, term length, dark-fiber versus lit-capacity split, redundancy architecture, customer remedies, termination rights, capex budget, permits, utility interconnection details, data-center building count or whether either campus already has signed AI compute tenants beyond the unnamed hyperscale customer.