U.S. buyers push generative AI vendors on governance and returns

The global generative AI market was valued at US$67.21 billion in 2025 and is projected to reach US$3,282.75 billion by 2035, with a 47.53% CAGR during 2026-2035. North America accounted for about 42% of 2025 demand, led by the United States, where enterprise purchasing is shifting from budget approval to workflow integration, risk control and measurable returns.

Regulatory and standards activity is shaping procurement. NIST is revising AI Risk Management Framework 1.0 and has released a concept note for trustworthy AI in critical infrastructure. The Federal Reserve, FDIC and OCC issued revised model-risk guidance on April 17, 2026, while noting that generative and agentic AI fall outside its formal scope because the technologies are evolving rapidly.

Vendors are responding with governed agent systems rather than basic model access. Google, IBM, Microsoft and AWS are emphasizing identity, registries, gateways, evaluation, observability, policy controls, consumption limits and multi-agent orchestration. Buyers are also scrutinizing hosted models, retrieval systems, internal data, security layers and workflow software as training costs make frontier model development impractical for most enterprises.

Financial services demand remains significant but requires separate qualification, including inventories, ownership, staged deployment, escalation, output testing and limits on what agents can read, generate or execute. Procurement teams are being urged to negotiate cost per completed task, autonomous retry limits, routing rules, commitments, overage treatment, unused capacity, data-egress charges and exit support before deployment.