Email me
Deep tech and money

Economics · · 6 min read · Matt Busel

By , who built the apps, crates and game he writes about · every number links to its source and every dataset is downloadable (methods) · corrections · featured in Kirino, gamebiz, This Week in Rust, New Mobile Life, By-Press

Filed under Deep tech and money Economics AI industry

OpenAI's $1 government year ended. First bills: $4.05M from 5 agencies

OpenAI's $1 government year ended. First bills: $4.05M from 5 agencies
On this page · 10 sections
  1. Key numbers
  2. The $1 year was a sales funnel, and the funnel worked
  3. Who signed the checks
  4. Follow the money: OpenAI 3.7x Claude
  5. Labor is the hedge
  6. Before the $1 year, agencies already paid full price
  7. How we built this, and what we could not see
  8. Why this number matters beyond Washington
  9. What this means for you
  10. Sources

Short answer: OpenAI's $1-per-agency ChatGPT deal with the federal government ran out on September 30, 2026. Federal award records show five agencies placed paid ChatGPT orders worth $4,048,481 between September 30 and October 1. The free year turned into a paying customer base within two days.

TL;DR: Treasury ($2.16M), Labor ($875K), Education ($711K), Commerce ($172.8K, for Codex) and the Pension Benefit Guaranty Corporation ($130K) each signed 12-month ChatGPT orders on September 30 or October 1, 2026. Four of the five went through the same GSA schedule run by the reseller Carahsoft. That is 3.7x the $1.10M that agencies have spent on Anthropic's Claude since the August court ruling, and Labor is now paying for both.

Key numbers

WhatNumberSource, checked Oct 27 2026
Paid ChatGPT orders, Sep 30 to Oct 1 2026$4,048,481 across 5 agenciesUSAspending.gov
Largest order: Treasury TEAM BPA task order 1$2,159,681, through Sep 29 2027USAspending.gov
Labor: ChatGPT vs Claude$875,000 vs $898,539.46, both dated Sep 30 2026USAspending.gov
Orders on Carahsoft's GSA schedule 47QSWA18D008F4 of 5Carahsoft OneGov page
Claude awards since the Aug 27 2026 ruling$1.10MTensorust, Oct 27 2026
Counter rising to $4,048,481, the paid ChatGPT orders five agencies placed Sep 30 to Oct 1 2026
Paid ChatGPT orders after the $1 year. USAspending.gov, pulled Oct 27 2026

The $1 year was a sales funnel, and the funnel worked

In August 2025 GSA announced that any federal agency could get ChatGPT Enterprise for $1 for a year (Nextgov). Coverage this month reported OpenAI replacing it with usage pricing at 50% off from October 1 (Mobile World Live). Nobody had shown what the switch costs taxpayers. The award records do.

We searched every federal contract award on USAspending.gov that names ChatGPT or OpenAI in its description since 2023, then pulled each award's period of performance. Five new orders start on September 30 or October 1, 2026, the exact day the $1 term ended. Every one runs 12 months.

Bar chart: Treasury $2,159,681, Labor $875,000, Education $711,000, Commerce $172,800, PBGC $130,000
First paid ChatGPT orders by agency, Sep 30 to Oct 1 2026. USAspending.gov, Oct 27 2026

Who signed the checks

  • Treasury, $2,159,681: task order 1 of a new Treasury Enterprise AI Models (TEAM) blanket purchase agreement, which the description says "will establish Treasury's ChatGPT" licensing. Treasury already bought up to 3,000 ChatGPT licenses for $3,398,315 in May 2025, before the $1 deal existed.
  • Labor, $875,000: "OneGov OpenAI, ChatGPT FedRAMP environment".
  • Education, $711,000: "a new award to procure OpenAI", starting October 1.
  • Commerce, $172,800: ChatGPT Codex, the coding agent. Commerce paid $117,612 for OpenAI a year earlier.
  • PBGC, $130,000: "ChatGPT Enterprise consumption units", the usage-based pricing in writing.

Meanwhile the FCC's $1 ChatGPT pilot order simply expired on October 11, 2026, with no paid follow-on in the records yet.

Follow the money: OpenAI 3.7x Claude

Earlier today we counted $1.10M in federal Claude awards since the August 27 ruling that let agencies buy it again. OpenAI booked 3.7x that in 48 hours. The $1 year gave it a head start of 3.5 million federal users, per GSA figures reported by Bloomberg.

Bar chart: OpenAI $4.05M in two days vs Claude $1.10M since Aug 27 2026
OpenAI vs Claude federal awards. USAspending.gov, Oct 27 2026

The winner in the middle is Carahsoft. Four of the five paid orders ride its GSA schedule 47QSWA18D008F, the same vehicle that carried the $1 orders. The loser is any agency that let the free year lapse with no usage data: consumption pricing means the bill now tracks how much people actually use it.

Labor is the hedge

The Department of Labor signed $875,000 for ChatGPT and $898,539.46 for Claude for Government on the same day, September 30. Two vendors, near-identical budgets. That is the pattern we expect to spread: agencies buying two frontier models and letting them compete for seats.

Bar chart: FCC $1 pilot, Labor ChatGPT $875,000, Labor Claude $898,539
Labor bought both models on Sep 30 2026. USAspending.gov, Oct 27 2026

Our read, and it is a bet, not a record: the five orders are the first wave. Most agencies run fiscal-year-end buys on September 30, and orders placed that day can take weeks to reach USAspending. We expect the total to keep climbing as late reports land.

Before the $1 year, agencies already paid full price

The $1 deal is often told as OpenAI opening a door. The records show the door was already open. Before August 2025, agencies were paying resellers for ChatGPT Enterprise at list-style prices.

  • NASA paid $108,000 for an annual ChatGPT Enterprise license in July 2024.
  • Treasury paid $107,100 for 150 licenses in September 2024. That is $714 per seat per year.
  • HHS paid $252,683.60 for ChatGPT Enterprise licenses in February 2025.
  • Treasury then paid $3,398,315 in May 2025 for up to 3,000 licenses, about $1,133 per seat per year if all were used (our math).

So the $1 year did not create demand at Treasury. It paused a bill that was already over $3M a year. The new TEAM order brings it back at $2.16M, about 36% lower than the May 2025 order (our math, on the face value of both awards). That gap is roughly what a 50% list discount on a usage model would look like if usage grew a little, but the award does not state seat counts, so treat that as our estimate.

How we built this, and what we could not see

Method, so anyone can rerun it:

  1. Query the USAspending.gov award search API for contract and IDV awards since January 2023 with the keywords "ChatGPT" and "OpenAI".
  2. Drop false matches (an "OpenAirframe" NASA research grant, an Azure OpenAI cloud order inside a larger Microsoft deal).
  3. Pull each award's detail record for period of performance and parent vehicle.
  4. Split by date: before August 2025, the $1 year, and on or after September 30, 2026.

The limits are real. USAspending only sees awards whose descriptions name the product, so ChatGPT bought inside a bigger cloud or integrator contract is invisible here. Defense orders often post with a 90-day delay. And the API matched "xAI" only to explainable-AI research grants, so we found zero federal Grok orders, not proof there are none. Each limit pushes the true total up, not down.

Why this number matters beyond Washington

Government buying is a public price signal in a market where every vendor hides its enterprise rates. Each of these orders is a dated, dollar-stamped record of what a large organization will pay for a frontier model seat. That is data builders can use: to price their own AI tools, to forecast reseller revenue, or to see which agencies are still on free pilots and worth pitching.

What this means for you

If you sell software to government, this is the playbook in numbers: a near-free year, then paid 12-month orders on day one. Download the data.csv below (14 orders, with links) and track your own competitors the same way. For more on these deals see Tensorust News and our deep tech + finance hub.

Want this watched for you? Company Watch ($99/mo) flags every new federal award, filing and lobbying report for the companies you name. We keep fixing it at no extra cost until it does what it promises.

Not financial advice. Our numbers and our read, so you can check them.

Sources

First published here on tensorust.com on Oct 27, 2026. Related: Tensorust News, Deep tech + finance.

Up next

More on Deep tech and moneyAnthropic S-1: Claude was banned. 5 agencies bought $1.1M sinceAnthropic S-1 warns of a federal ban on Claude. USAspending shows 5 agencies signed $1.1M of new Claude awards since, 99.9% after the August court win.Read it next →

Cite this

Free to cite and reuse under CC BY 4.0 with a link back. How it was collected: methods. Spotted an error? Corrections.

Did this answer your question?

Get the next one by email.

What my AI agents tried this week, with the real numbers: what worked, what flopped, what made money. One email when there is something worth reading.

No spam, never sold. Every email has a one-click unsubscribe.

What are people saying about this? Read and post on The Floor →

More: the author, related reading, share this, a free game

Try it yourself · Want the numbers to feel real?

See any amount as a pile of cash

Type an amount, see it as $100 bills next to a person: weight, volume and height.

In The Manual: on this topic

Explainers answer one question, short answer first. Guides walk you through it step by step.

Open The Manual →

Read next

All articles →

Art: Nicholas Roerich, Sanctuaries and Citadels (1925). Public domain, via Wikimedia Commons.

1M+ X impressions in 4 days, run by AI agents · 8,000+ crates.io downloads across our 70 Rust crates, all time · 529 GitHub stars · 8 iPhone apps live on the App Store · O'BLOCK out now, free on Steam

Built by Matthew Busel · code on github.com/Mattbusel · numbers as of Oct 25, 2026, click any to check

Popular in the shop

All 124 tools →

Browse: Bundles and kitsHome, family and schoolMoney and businessAI toolsSports and fitnessMaking and hobbiesFood and drinkIndie dev and launchTrading and quant

Free: Landing page teardown iPhone app cost calculator LLM API cost calculator Paid online: $49 Launch Check App builds Articles FAQ