A Dollar a Year
The US government's one-dollar AI year is ending, and the new deals charge for every use. In August last year, a whole agency could have ChatGPT for one year, for one dollar. Now OpenAI's dollar is over, and Anthropic's runs out on the thirty-first of October.
The deals
OpenAI first, at a dollar per agency. Days later, Anthropic, which makes Claude, the model Token Runner is built with: a dollar too, for all three branches of government. Google: forty-seven cents.
As announced: OpenAI's $1 per agency on 6 August 2025, ended 30 September 2026. Anthropic's $1 on 12 August 2025, to 31 October 2026 as reported. Google's $0.47 on 21 August 2025, to 15 November 2026 as reported.
The chart sets the three prices side by side, each with its end date. Each was for one year.
Source: GSA's own announcements (6 and 12 and 21 August 2025; 10 September and 27 October 2026), the companies' own posts, and Nextgov for the extended end dates (as reported), read at source 28 October 2026
Who it reaches
GSA, the government's buyer, says these deals now reach about three and a half million federal staff. In its words, they "expanded access to AI tools for roughly 3.5 million federal employees".
That is GSA's own calculation, and so is the $1.4 billion of savings it counts against list prices. Neither is audited.
The year is up
And the year is up. In GSA's words, it's "maturing from short-term promotional pricing toward direct contracts."
OpenAI's dollar ended in September. Its new deal charges for every use, at half price: "a 50% discount on token-based usage", with "no platform-access fee, no minimum orders, and no spend commitment". Anthropic's dollar runs to the thirty-first of October. Google's, to mid-November.
GSA's Laura Stanton calls "consumption-based access" the "next logical step". Nextgov's reading is that the shift "will likely include higher costs for agencies". It gives OpenAI's list price as $15 per user per month; that is an illustration only, not what any agency pays.
Anthropic's new deal
On the twenty-seventh of October, GSA announced what comes next for Anthropic: in its words, a "40% discount on token-based usage", and "no minimum orders". The discount runs "across Claude models available to the government, including FedRAMP-authorized environments for sensitive unclassified work". Like OpenAI's, it charges for every use, at a discount. Anthropic makes Claude, which Token Runner is built with.
“… 40% discount on token-based usage across Claude models available to the government …”
GSA's notice states no term or dates, and doesn't mention the $1 offer. Its own line on what comes after: "As offers expire, GSA will renew agreements where appropriate." In Anthropic's own post that day, the discount covers Claude Enterprise, Claude for Government and the Claude API, and reaches state, local and tribal governments. Anthropic estimates it will save the government $1 billion; that is its own claim.
The bridge
Days before, the health department said it planned what it calls "a temporary six-month sole-source bridge contract" with Anthropic, without competition, to keep Claude running. Anthropic makes Claude, which Token Runner is built with.
HHS posted its notice of intent on 19 October: "continued enterprise access to Anthropic's Claude", "pending award of the competitive HHS Enterprise Frontier SI Models and Advanced Features multiple-award BPA". It is a notice, not yet an award, and it gives no value. FedScoop's "under $350,000" is its own inference.
The order
This is the same company the President ordered every federal agency to stop using, in February. In March, a federal judge blocked that order. The case goes on. Anthropic makes Claude, which Token Runner is built with.
The directive of 27 February 2026 ordered all federal agencies to cease use of Anthropic's technology. On 26 March, Judge Rita F. Lin of the Northern District of California enjoined it, "pending final resolution of this action or further order of this Court". Only the order and the injunction are stated here, nothing beyond the court record.
The warning
Procurement experts saw this coming. Jessica Tillipman of George Washington University called the dollar price "a nominal cost that makes the transaction sound exciting and low risk, while the true economics… arrive later." Her verdict: "The leverage agencies have today will not survive renewal."
The cut in her first line is "such as lifecycle costs, operational dependency, and reduced negotiating leverage". Greg Barbaccia, then the federal CIO, said in the same FedScoop piece: "I don't think it's going to be $1. I think it'll be transparent at least." FedScoop also reports, from Thomson Reuters, that 27% of government professionals use AI tools their agency hasn't sanctioned.
The next bill
A dollar bought a year. The next bill is the one that counts.
Three things to watch. What an agency pays per use once its dollar is gone; no claim is made here about future prices. Whether HHS's bridge contract still goes ahead after GSA's agreement of 27 October. And how the court case ends.
What we couldn't confirm: Whether the 2026 Claude listing is per user, as Nextgov reports it. The HHS contract's value.
Every figure from the government's own announcements and the companies' own words. Anthropic makes Claude. A disclosure: Token Runner is built with Claude. Token Runner dot A I.
Sources
- GSA, 6 August 2025: "Every participating U.S. federal agency will have access for a nominal fee of $1 to ChatGPT Enterprise for one year." Read at source.
- GSA, 12 August 2025, the Anthropic deal: Claude "to all Branches of Gov for Just $1", "$1 Per Agency" in the body; Anthropic's own post the same day: Claude for Enterprise and Claude for Government "to all three branches of government. for $1"; "This package will be available for one year." Anthropic makes Claude. A disclosure: Token Runner is built with Claude.
- Anthropic's own post, 12 August 2025.
- GSA, 21 August 2025: Google's Gemini for Government at $0.47 per agency.
- GSA, 10 September 2026, OpenAI's new deal: "a 50% discount on token-based usage"; "no platform-access fee, no minimum orders, and no spend commitment"; and GSA's own calculation that "approximately $1.4 billion stems from AI agreements that have expanded access to AI tools for roughly 3.5 million federal employees" (GSA's savings figures are its own calculation against list prices, not audited).
- GSA, 'GSA, Anthropic Reach OneGov Agreement on Discounted, Consumption-Based Pricing for Claude', 27 October 2026: "40% discount on token-based usage"; "No minimum orders"; "As offers expire, GSA will renew agreements where appropriate". Read at source 28 October 2026.
- Anthropic's own post, 'Expanding our partnership with GSA to bring Claude to all levels of government', 27 October 2026: its estimate of $1 billion in savings and the extension to state, local and tribal governments, as Anthropic's own claims. Read at source 28 October 2026.
- Nextgov, September 2026: Anthropic's $1 deal extended to 31 October 2026 (citing GSA's listing; the 2026 listing is reported as 'per user', not established, so neither is said here); Google's to 15 November 2026.
- Nextgov, September 2026: Google extends its Gemini OneGov deal to 15 November; a GSA official's statement to Nextgov, 29 September 2026: "OneGov is maturing from short-term promotional pricing toward direct contracts that secure benefits and give agencies broad choice."
- HHS, Notice of Intent to Award Sole Source, 7571TE27R00001, posted 19 October 2026, responses due 21 October: "NOTICE OF INTENT TO AWARD SOLE SOURCE. continued enterprise access to Anthropic's Claude. pending award of the competitive HHS Enterprise Frontier SI Models and Advanced Features multiple-award BPA"; cites FAR 13.106-1(b)(1); gives no value. A notice of intent, not an award.
- FedScoop, 23 October 2026: an HHS spokesperson, "a temporary six-month sole-source bridge contract" to "maintain continuity of vital" AI services; FedScoop's 'under $350,000' is its own inference.
- Anthropic PBC v. U.S. Department of War, US District Court for the Northern District of California, No. 26-cv-01996-RFL: Judge Rita F. Lin's order of 26 March 2026 enjoins "implementing, applying, or enforcing. The February 27, 2026 Presidential Directive ordering all federal agencies to cease use of Anthropic's technology", "pending final resolution of this action or further order of this Court". Anthropic makes Claude. A disclosure: Token Runner is built with Claude.
- Jessica Tillipman, George Washington University Law School, 'Twenty-Five Cent AI: The Hidden Costs of Promotional AI Deals': "The leverage agencies have today will not survive renewal."
- Jessica Tillipman as quoted by FedScoop, 11 August 2026: the dollar price is "a nominal cost that makes the transaction sound exciting and low risk, while the true economics, such as lifecycle costs, operational dependency, and reduced negotiating leverage, arrive later." FedScoop also carries Greg Barbaccia, then federal CIO: "I don't think it's going to be $1. I think it'll be transparent at least."