The Pentagon has obligated $13,149,999 so far on OpenAI’s frontier AI prototype agreement, a deal it announced in June 2025 “with a value of $200,000,000”, according to federal procurement records compiled by HigherGov. By our arithmetic, that is 6.6% of the headline value.
The record shows the money rising from $1,999,998 at award, and the agreement’s end date moved from June 2026 to June 2027. The figure puts a number on the government business of the OpenAI national security team that Russian pranksters reached on a 17-minute video call, as The Intercept reported on 11 November.
Business Pill · Forecast vs Commitment
An evergreen Business Engineer explainer on the difference between a forecast and a commitment. It teaches the general idea only and does not discuss OpenAI or the Pentagon.
The key insight: The Pentagon announced OpenAI’s frontier AI prototype agreement with a value of $200 million, and federal procurement records compiled by HigherGov list $13,149,999 obligated so far, up from $1,999,998 at award. By our arithmetic that is 6.6% of the headline value, against 14.8% for Google and 7.1% across the four labs that got the same kind of agreement.
The Pentagon’s contract notice of 16 June 2025 says OpenAI Public Sector LLC was awarded a fixed amount, prototype, other transaction agreement, HQ0883-25-9-0012, to “develop prototype frontier AI capabilities to address critical national security challenges in both warfighting and enterprise domains”. It says $1,999,998 of fiscal 2025 research funds were obligated at award, with an estimated completion date of July 2026.
HigherGov’s record of the agreement, last modified on 31 August 2026, lists total obligations of $13.1 million and says the value has “increased 558% from $1,999,998 to $13,149,999”. By our arithmetic, that is about 6.6 times the amount at award. The record says the potential end date was extended from 15 June 2026 to 15 June 2027.
The record lists eight modifications. The latest shown, dated 31 July 2026, added $1.0 million. Another, a change order dated 27 February 2026, shows no added money; The Intercept reported in September on an amended OpenAI contract document of that date with a clause about embedding OpenAI engineers within the military.
OpenAI’s frontier AI prototype agreement with the Pentagon: the $200 million headline value in the 16 June 2025 contract notice, the $1,999,998 obligated at award, and $13,149,999 obligated in the federal procurement record compiled by HigherGov (last modified 31 August 2026). Obligated means committed, not paid. The 6.6% share is our arithmetic.
How the Four Labs Compare
The Pentagon announced three more agreements of the same kind in July 2025, each also with a value of $200 million and $1,999,998 obligated at award, as the notices reproduced in HigherGov’s records show. Google Public Sector’s record lists $29.5 million obligated, 14.8% of its headline value by our arithmetic, after a $20.0 million change order dated 27 April 2026.
The fourth award is recorded under AIQ Phase LLC, with $12.0 million obligated, or 6.0%. The Intercept named xAI as the fourth lab; the procurement record itself names AIQ Phase. Anthropic’s record shows $2.0 million obligated, 1.0%, but was last modified on 15 August 2025, so it may not reflect later actions.
Across the four, the records list $56.7 million obligated against $800 million in headline value, or 7.1%, by our arithmetic. OpenAI accounts for 23% of the obligated total and Google for 52%.
Money obligated on the four frontier AI prototype agreements, each announced with a value of $200 million, as listed in federal procurement records compiled by HigherGov and read on 12 November 2026. Anthropic’s record was last modified in August 2025. Shares and total are our arithmetic.
Connecting the Coverage
On 12 November we reported that three of OpenAI’s senior national security staff spoke to Russian pranksters posing as Ukraine’s foreign intelligence chief, and that OpenAI told The Intercept it had concluded, with the relevant authorities, that the call was a prank. What’s new here: the money the Pentagon has committed to the contract that team works around.
On 2 November we reported that OpenAI hired former Army secretary Dan Driscoll as Senior Counselor for National Security Modernization. On 26 October we reported that Anthropic’s S-1 says a federal appeals court ruled against it 2-1 over the Department of War’s supply-chain risk designation.
The Structural Read
As we read the records, the $200 million in the headlines is a ceiling the Pentagon can fund over time, while the obligations show what it has committed so far: about $56.7 million across four labs, by our arithmetic.
As we read it, the national security staff The Intercept says the pranksters reached work on a government business whose committed money, in this agreement, is $13.1 million. The records do not cover what OpenAI earns from other government work.
US Department of Defense, contract notice of 16 June 2025, on OpenAI Public Sector LLC’s agreement
“Under this award, the performer will develop prototype frontier AI capabilities to address critical national security challenges in both warfighting and enterprise domains.”
Three Implications
FOR OPENAI HigherGov’s record lists $13,149,999 obligated on its agreement and an end date extended to 15 June 2027.
FOR THE FOUR LABS The records list $29.5 million for Google, $12.0 million for AIQ Phase and $2.0 million for Anthropic, whose record was last modified in August 2025.
FOR READING DEFENSE AI DEALS Each agreement was announced with a value of $200 million and $1,999,998 obligated at award, the Pentagon’s notices show.
The framework describes its governance layer as “The rules that determine whether a model can be released, deployed, sold, or used.” and says it is “No longer a passive layer; an active priced variable.”
As we read it, a procurement record is one place that variable gets a price: $13,149,999 committed on OpenAI’s $200 million agreement, by HigherGov’s record. The framework predates these records and does not discuss them.
What Is Not Established
An obligation is money committed, not money paid; the records we read do not show what OpenAI has been paid or billed. HigherGov compiles federal procurement data, and records can lag; Anthropic’s was last modified in August 2025.
The records do not describe what OpenAI delivers under the agreement, and they do not say whether the 27 February 2026 change order relates to the agreement to deploy OpenAI’s models on the Department of War’s classified cloud networks that Sam Altman announced the same day, Reuters reported. We did not contact OpenAI or the Pentagon for this piece.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
The Pentagon has obligated $13,149,999 on OpenAI’s $200 million prototype agreement, up from $1,999,998 at award, according to procurement records compiled by HigherGov. By our arithmetic, that is 6.6% of the headline value, against 14.8% for Google and 7.1% across the four labs.
A note on sourcing. We read the Pentagon’s contract notice of 16 June 2025 in full, and the four HigherGov records of the frontier AI prototype agreements on 12 November 2026 between 09:00 and 09:30 UTC; HigherGov compiles federal procurement data, and we did not have direct access to the government’s own spending database for this piece. The Reuters report on the classified-network agreement was read via The Star.
The prank call, The Intercept’s reporting on the contract documents and the Driscoll and Anthropic details come from our earlier, linked pieces. Shares, ratios and totals are our arithmetic. We did not contact OpenAI or the Pentagon.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
