AI projects now start with hiring and fiber, not models

What happened

The Social Security Administration posted a request for information on an enterprise AI strategy that explicitly includes agentic capabilities (AI that can take actions), with responses due Sept. 28. New York state signed a three-year enterprise agreement with IBM for AI and hybrid cloud tooling across agencies, expecting $6 million in savings over three years. Kyndryl and Broadcom are packaging VMware Cloud Foundation services as an AI-ready private cloud, pledging to skill up several thousand certified staff. Zayo locked in fiber optic cable supply with Corning to manage execution risk, aiming to add 15,000 new route miles by 2030.

Why it matters

These moves show that AI programs are increasingly constrained by the capacity to deliver and operate them, not the ability to demo them. The common thread is execution capacity: people who can implement and operate platforms, contract vehicles that can onboard large groups, and physical infrastructure supply that shows up on schedules. This suggests a more mature buying pattern for AI in 2026, where operational readiness and infrastructure are prerequisites for effective AI model deployments.

What to watch

SSA's RFI asks for architectural approaches and models, not vendor pitches, and explicitly calls for governance balancing speed with accountability. New York's enterprise agreements are credited with $58 million in annual savings across 14 agreements. Zayo's deal locks in a substantial share of fiber it expects to require through the rest of the decade, with over 8,000 miles of new long-haul fiber built with NVIDIA.