When OpenAI joined the GSA’s OneGov marketplace in August 2025 for $1 per year, it was less a business decision and more a statement. Now that era is over. As reported by Yahoo Finance, the U.S. General Services Administration has announced a new 27-month agreement with OpenAI that takes effect October 1, 2026, replacing the flat-fee arrangement with consumption-based pricing set at 50% below OpenAI’s standard commercial rates.
The mechanics of the new deal are worth paying attention to. There’s no platform-access fee, no minimum order, and no spending commitment. Agencies pay only for the tokens they consume. That structure is more honest than the $1 headline ever was. A $1 annual fee sounds dramatic, but it told you nothing about what it actually cost to deploy these tools at scale. Token-based pricing does.
Coverage under the new agreement extends beyond the federal executive branch to include legislative and judicial agencies, plus state, local, and tribal governments. That’s a meaningful expansion. GSA says its OneGov marketplace has generated roughly $1.68 billion in cumulative cost savings, with about $1.4 billion of that tied specifically to AI agreements covering around 3.5 million federal employees.
OpenAI CEO Sam Altman framed the transition as part of a broader push to give public servants access to better tools while improving government efficiency and cybersecurity. GSA Administrator Edward C. Forst described it as a “secure, governmentwide framework” for mission-critical work. Both statements are standard issue for a federal contract announcement. The more interesting context is competitive.
When OpenAI signed its original $1-a-year deal, Anthropic matched the terms within days. That parallel track has since gone sideways. Defense Secretary Pete Hegseth flagged Anthropic as a supply-chain risk, and President Trump directed agencies to stop using its tools. Anthropic sued the federal government. A California judge issued a preliminary injunction blocking the ban, and GSA reversed a directive to pull Anthropic from its platform. Litigation in both California and Washington is still active.
That dispute puts OpenAI in a structurally stronger position in the federal market, at least for now. The company separately reached a deal to provide tools to the Pentagon in February, which further cements its foothold across defense and civilian agencies. Anthropic’s legal fight with the administration creates uncertainty for procurement officers trying to plan long-term AI deployments. OpenAI doesn’t currently have that problem.
For competitors watching the federal AI space, this sets a new pricing benchmark. A 50% discount off commercial rates, with no minimums and no platform fee, is a model that other vendors will be measured against. Microsoft, Google, and Amazon all have existing federal cloud and AI relationships, but none of them have locked in a GSA-structured AI agreement with this kind of visibility right now.
The shift from symbolic pricing to usage-based billing is a sign that federal AI adoption is maturing. The $1 deal was about access and optics. This new structure is about scale and actual consumption. That’s a meaningful step forward, and it’s one that will tell us a lot more about where AI really fits inside government once the usage numbers start coming in.
