The paper, accompanied by a letter from GLF Board Chair Enrico Bagnasco, argues that AI is doing more than driving traffic growth. It is restructuring traffic topology entirely, accelerating inter-data centre and cross-border flows concentrated on a limited number of high-demand corridors connecting hyperscale and regional AI hubs.

As those hubs mature, the GLF says a new “hub-to-hub” connectivity layer is emerging — characterised by high utilisation, bursty replication flows, and stringent latency and packet loss requirements.

The GLF’s central proposition is that carriers should not attempt to compete with hyperscalers on GPU compute or replicate their platforms. Instead, they should productise existing structural advantages, global subsea and terrestrial reach, multi-operator relationships, and trusted regulatory standing into repeatable, performance-assured corridor services.

The market opportunity is framed as substantial. GLF estimates the telco-addressable AI infrastructure pool will reach USD 120–150 billion by 2027, with international carriers collectively positioned to capture 20–30% of that figure, implying a serviceable addressable market of US$25–45 billion. The paper stresses this requires disciplined focus, noting that not all spend is accessible and that value will be driven by converting corridor positions into assured products rather than chasing aggregate market size.

The paper identifies five layers of potential carrier product offering: AI corridor connectivity services with deterministic performance guarantees; selective edge and sovereign infrastructure enablement; platforms supporting deployment and integration across hybrid environments; network intelligence embedded as a differentiator; and targeted vertical solutions where carriers act as trusted integrators.

On the relationship with hyperscalers, GLF advocates a deliberate co-opetition posture. Hyperscalers are simultaneously the largest buyers of high-capacity interconnect and active builders of their own private backbones. Carriers, the paper argues, must partner to serve growth while sustaining differentiation through capabilities hyperscalers are unlikely to replicate everywhere: cross-border regulatory execution, neutral interconnection, and compliance-by-design with auditable controls.

To underpin the AI Backbone role at scale, the GLF calls for industry-wide collaboration across six priority areas: carrier-to-carrier APIs and commercial automation; standardised AI-grade SLAs, KPIs, and telemetry; federated edge and compute interconnect blueprints; trust, sovereignty, and compliance frameworks; corridor collaboration and low-friction end-to-end integration; and risk-shared investment models including consortium, joint-venture, and neutral-host structures.

The paper flags material constraints on delivery, including data centre power availability, multi-year subsea cable build timelines, and geopolitical fragmentation accelerating regional AI hub development rather than a single global ecosystem. It also warns that any collaborative activity must remain competition-law compliant, limited strictly to interoperability enablers and excluding any coordination on pricing, commercial terms, or customer allocation.

GLF is inviting carriers, vendors, and industry bodies to engage with the framework, with the goal of ensuring AI connectivity infrastructure is built as open, interoperable, and resilient — enabling workloads to move securely and compliantly across borders and between hubs and edge sites.

Download the full report here: GLF – Becoming the AI Backbone – GLF

RELATED STORIES

GLF Pulse Survey reveals industry leaders are shifting focus from ambition to execution in genAI era

GLF at BATIC 2025: Driving partnerships, collaboration, and connectivity