AI's new bottleneck isn't chips — it's permits, fiber, and supply chains

Lumen's chief public policy officer Melissa Mann told a Technology Policy Institute Aspen Forum panel this week that the AI build-out's rate-limiter is no longer chips or models. It is permitting.

"It's not just an engineering question. It's really a policy question and our ability to meet these demands," Mann said. Lumen, which recently sold its residential fiber business to AT&T, expects to grow from about 17 million fiber miles (the total length of optical cable in its network) at the end of 2025 to 58 million by the time it exits 2031 to meet hyperscaler contracts, the largest cloud companies operating AI at scale. "If we're actually going to do this and double our fiber capacity across the industry, we've got to fix permitting," she added.

Cisco's forecast at the panel: AI-related network traffic is expected to triple within three years and could rise further. AT&T's Giulia McHenry, SVP for public policy, said the carrier has seen a 15% increase in overall data traffic since 2023, "ensuring that we're ready for AI to cross our networks" even as agentic AI (bots that take multi-step actions) reshapes traffic patterns.

Panel commentary and the article's framing identify three constraints on the AI build-out: permitting, fiber capacity, and supply chains. Mann did not name a fix. Recent Light Reading "The Divide" guests pointed to "shot clocks," or mandatory time-limited deadlines, and cost-based fees. The supply-chain tail of the panel was not in the published excerpt. What remains: whether the US can permit fiber at the pace hyperscalers want.