The OneGov AI Agreements: What GSA's Deals With OpenAI, Anthropic, Google, and xAI Mean for Contractors

GSA's OneGov AI agreements give federal agencies access to frontier AI tools from OpenAI, Anthropic, Google, xAI, and others at nominal pricing — in some cases $1 per agency for the first year. The license revenue is gone, deliberately. The contractor opportunity moved downstream: implementation, integration, security accreditation, training, and governance services around tools agencies now hold but cannot yet deploy.

I spent eighteen years in federal acquisition as a Contracting Specialist and Contracting Officer at GSA, IRS, DoD, and DOI, and I have never seen the government acquire a technology category this fast. Since April 2025, GSA has stacked up roughly two dozen OneGov agreements covering AI tools from Anthropic, OpenAI, Google, Microsoft, Meta, xAI, and Perplexity. Most contractors read the headlines, saw "$1," and concluded there was nothing in it for them. That is exactly backwards. Here is what these deals actually change.

What are the OneGov AI agreements?

OneGov AI agreements are government-wide deals GSA negotiates directly with AI companies — OpenAI, Anthropic, Google, xAI, Microsoft, Meta, Perplexity, and more — that put their tools on the Multiple Award Schedule at steep, sometimes nominal, discounts with AI-specific terms and conditions already negotiated. Any agency can order without running its own competition or negotiating its own terms.

The mechanics matter more than the press releases:

• The vehicle is the MAS program. These are Schedule contract actions under GSA's Buy AI program, not a new contract type. Agencies order the way they already order software.
• GSA negotiates the AI terms once. Data-use rights, model-training restrictions, and security terms are standardized centrally instead of being reinvented by every agency counsel's office — the same direction GSAR 552.239-71 pushes for cloud.
• Pricing is nominal on purpose. OpenAI offered ChatGPT Enterprise to agencies at $1 per agency for a year. Anthropic and others matched the pattern. The vendors are buying installed base, not revenue.
• The windows are time-boxed. Each agreement has its own enrollment deadline — the xAI deal runs the longest, giving agencies until March 2027 to sign on. Nominal pricing is an introductory rate, not a permanent state.

Why would AI companies give the government their product for $1?

Because the first-year license was never the prize. The prize is becoming the default model inside federal workflows before FedRAMP-authorized competitors lock in, then monetizing renewals, API consumption, and premium tiers once the workforce is trained on their tool. Every commercial software company that won big in government ran a version of this play.

From the Contracting Officer seat, I watched this pattern with virtualization, then cloud, then collaboration suites: the vendor discounts the entry point, the agency builds mission dependence, and the out-year money shows up in consumption and services. The difference this time is speed — DoD alone signed contract actions worth up to $200 million each with Anthropic, Google, OpenAI, and xAI for national-security AI work in mid-2025, while the civilian side was still standing up pilots. The renewal negotiations, when they come, will not be at $1. Agencies know it, the vendors know it, and the contractors who position now will be the ones in the room when real money starts flowing.

One more signal worth reading: GSA's nominal-dollar AI awards drew bid protests from vendors left off the early agreements. Nobody protests a market that doesn't matter.

Where is the actual contractor opportunity if the licenses are nearly free?

In everything the $1 does not buy. An agency with 40,000 ChatGPT Enterprise seats still needs someone to integrate the tool with its data, secure it, build use cases, train its workforce, and stand up governance. That is services work — and services flow through GSA Schedule SINs 54151S, 54151HACS, and 611420, not through the AI vendors' agreements.

Across our 70+ proven GSA contract awards, the services firms winning AI-adjacent task orders right now are not AI companies. They are integrators and consultancies who added AI implementation language to their existing labor categories and technical narratives. The agency already has the tool. It is buying hands.

What does OneGov do to the reseller channel?

It compresses it. When GSA buys directly from the OEM at a negotiated government-wide price, the traditional value-added reseller margin on the license disappears. Resellers who survived prior OneGov actions on software and hardware did it by shifting from license markup to attached services — and the AI agreements accelerate that same shift.

If you are a reseller today, your realistic options rank like this:

1. Become the delivery arm. The AI vendors are product companies with thin federal services benches. Partner into their federal delivery ecosystem and take the implementation work their agreements generate.
2. Sell the surrounding stack. Agencies adopting AI tools buy more compute, storage, identity, and security tooling. That spend still moves through the channel.
3. Aggregate and manage. Multi-model environments — an agency running Claude, Gemini, and ChatGPT side by side — need someone to manage licenses, usage, and cost across vendors. That is a managed-service, and it is billable.
4. Do nothing and quote licenses. This lane is closing. When the customer can order at the OneGov price directly, a quote with margin on top loses every time.

How does OneGov change the AI procurement path for agencies?

It removes the two slowest steps — competition strategy and terms negotiation — from the front of the acquisition. An agency that wanted an AI capability in 2024 faced a market-research, synopsis, and negotiation cycle measured in quarters. Under OneGov, the tool decision takes days and the hard acquisition work shifts to the services task order that makes the tool useful.

That shift changes where you compete. When I was a Contracting Specialist, the requirements documents that moved fastest were the ones where the product decision was already made and the statement of work only had to describe services. Expect exactly that shape: short-turn task orders on MAS, GWACs like Alliant 3, and OASIS+ for "implementation and adoption of [agency]'s enterprise AI tools." If your capability statements and Schedule labor categories do not say AI integration, AI security, or AI training in words an evaluator can map to those SOWs, you will not make the shortlist — regardless of how good your engineers are.

What should you do in the next 90 days?

Position your Schedule and your past performance before the services wave crests. The agencies that took $1 licenses in 2025 are hitting the limits of self-service adoption now — that is the buying trigger.

1. Audit your MAS terms. Confirm you hold the SINs the work will flow through (54151S at minimum) and that your awarded labor categories can cover AI-titled roles without a new offer. If not, submit the modification through eMod now — not when the RFQ drops.
2. Rewrite one past performance narrative. Take your strongest systems-integration or training reference and reframe it around the AI-relevant skills it demonstrates: data pipelines, security accreditation, workforce change management.
3. Pick two agencies that signed OneGov AI agreements and ask their program offices the only question that matters: what is stopping broader deployment? The answer is your pipeline.
4. Track the terms standardization. GSA is building AI-specific terms and conditions into these agreements. Whatever lands there will flow down to your task orders — read the current versions on acquisition.gov and GSA's Buy AI page before you price anything.
5. Watch the renewal cliff. Introductory pricing ends on each agreement's own clock. Agencies facing their first real renewal bill will need cost-optimization and multi-model strategy help — a service line that barely exists yet.

If your firm needs the Schedule itself — or the SIN and labor-category modifications to compete for this work — that is exactly what we build. Start with our GSA Schedule services page and we will map your capabilities to the SINs the AI services wave is flowing through.

What Is the Bottom Line?

• OneGov AI agreements put frontier models in agencies at nominal cost — roughly two dozen deals since April 2025 covering OpenAI, Anthropic, Google, Microsoft, Meta, xAI, and Perplexity.
• The license money is gone by design. The contractor revenue is in implementation, security, training, and governance services around the tools.
• Resellers must convert to services or aggregation. Margin on the license quote is structurally dead when GSA buys direct from the OEM.
• Services flow through your Schedule, not theirs. Confirm SINs 54151S, 54151HACS, and 611420 coverage and modernize your labor categories through eMod now.
• The renewal cliff is the second wave. When nominal pricing expires, agencies will pay real money — and buy real help deciding how.

Frequently Asked Questions

What is a OneGov agreement?

A OneGov agreement is a deal GSA negotiates directly with a manufacturer or software publisher on behalf of the entire federal government, using GSA's aggregate buying power to secure government-wide pricing and standardized terms. The AI agreements apply this model to tools from OpenAI, Anthropic, Google, xAI, and others, delivered through the Multiple Award Schedule.

Can any federal agency use the OneGov AI agreements?

Generally yes — the agreements are designed for government-wide ordering through GSA, though each has its own enrollment window and scope. Agencies should confirm eligibility and current terms on GSA's Buy AI page before committing, because deadlines and covered products differ by vendor.

Do the OneGov AI deals replace FedRAMP requirements?

No. A OneGov agreement sets price and contract terms; it does not waive security authorization. Agencies still need the AI service authorized for their data and use case, which is precisely why security and accreditation services are one of the strongest contractor opportunities around these deals.

I run a services firm with no AI practice. Am I too late?

No. Most agencies are still in early deployment — licenses acquired, broad adoption stalled. Firms with systems integration, cybersecurity, or training past performance can reposition existing capabilities for AI-adjacent task orders faster than AI startups can build federal delivery experience.

How do I add AI services to my existing GSA Schedule?

Usually through a modification rather than a new offer. If you hold SIN 54151S, updated labor category descriptions and a refreshed technical narrative submitted through eMod typically cover AI implementation work. If you lack the relevant SIN, an add-SIN modification requires a technical narrative and pricing support for the new scope.

Will the nominal pricing last?

No. The $1-class offers are introductory terms tied to enrollment windows — the xAI agreement, the longest announced, runs to March 2027. Renewal pricing will be negotiated at commercial-scale rates, which is why cost-optimization and multi-model strategy services will matter in the out-years.

Does OneGov mean agencies stop competing AI requirements?

The tool selection is largely pre-negotiated, but services task orders around the tools remain competitive under normal MAS, GWAC, and OASIS+ ordering procedures. The competition moved downstream from the license to the implementation — which is where services contractors should focus.