OpenAI-GSA OneGov Deal Extends ChatGPT Access Through 2028

What happened

OpenAI has renewed its OneGov agreement with the US General Services Administration, extending ChatGPT's availability to federal agencies through 2028. The renewed deal also marks a shift in how the tool is commercialised across government: pricing moves from a flat-rate model to a consumption-based structure, meaning agencies will pay according to actual usage rather than a fixed subscription fee.

The agreement continues the OneGov arrangement that first brought ChatGPT into federal government use, giving agencies a longer runway to plan deployment and budgeting around the tool rather than negotiating short-term renewals.

Why it matters

This is fundamentally a public-sector digital transformation story. Locking in access through 2028 gives US federal agencies multi-year certainty to build ChatGPT into workflows, procurement cycles and staff training programmes, rather than treating generative AI as a pilot that could lapse. Multi-year GovTech commitments of this kind tend to accelerate internal adoption, because agency IT and procurement teams can plan integrations, governance and change-management around a known horizon instead of quarter-to-quarter uncertainty.

The shift to consumption-based pricing is arguably the more consequential detail for how AI gets embedded in government operations. Usage-based billing aligns cost with actual demand, which can lower the barrier for smaller agencies or pilot teams to start using the tool, while also making spend more variable and harder to forecast at scale. It signals a maturing commercial relationship between AI vendors and large public-sector buyers, moving away from blanket licensing toward metered consumption — a pattern likely to spread as other governments and large enterprises negotiate their own AI contracts.

The Renascence take

Most coverage of this deal will focus on the multi-year horizon and treat the pricing change as a footnote. That's the wrong emphasis.

The move from flat-rate to consumption-based pricing is the real signal here, not the 2028 end date. It shows AI vendors and large institutional buyers converging on a model where cost tracks usage rather than seats or subscriptions — which fundamentally changes how agencies design workflows, because every prompt now carries a marginal cost that didn't exist before. Public-sector leaders adopting this model should treat usage data as a design input from day one: understanding which teams, tasks and query types drive consumption isn't just a finance exercise, it's the basis for deciding where AI genuinely improves service delivery versus where it's being used out of habit. Agencies that build that discipline early will get far more value out of the 2028 runway than those that simply renew access and hope adoption sorts itself out.