Federal AI Contract Statistics: 2026 Report

Published October 14, 2026

Federal agencies buy artificial intelligence and machine learning work from outside contractors. We studied federal AI contracts from fiscal year 2020 through fiscal year 2026, using every contract in USAspending.gov whose record mentions artificial intelligence or machine learning, to measure how much agencies commit to these contracts, how much goes to small businesses, and which agencies and contractors account for it.

An obligation is the amount an agency legally commits to pay a contractor in a given year, and it is the measure USAspending uses for federal spending. Federal agencies obligated $504.6 million on these AI contracts in fiscal 2026, more than double the $247.4 million obligated in fiscal 2020.

In the following sections, we total federal AI contract spending by year, show the small business share, break contracts down by agency, list the largest contractors, and compare contract sizes and industry codes.

Federal AI Contract Spending by Year

Seven years of obligations show how quickly agencies have added AI work to their contracts. The AI Services Companies research team totaled obligations by federal fiscal year, which runs from October 1 to September 30.

• FY2026 obligations of $504.59 million were 4.2% above FY2025 and 2.04 times the FY2020 level.
• The number of contract awards with AI obligations fell from 542 in FY2024 to 437 in FY2026 while dollars rose, so the average award grew larger.
• FY2026 figures will rise as late records arrive. The Department of Defense reports contract data to USAspending with a 90-day delay.

Small Business Share of Federal AI Contracts

Federal contract records flag whether the company that won the award counts as a small business under SBA size standards. This split shows how much federal AI work goes to smaller firms.

• Small businesses received 70.2% of AI contract obligations in FY2024, the highest share in the seven years.
• The small business share fell to 52.3% in FY2025 and 44.2% in FY2026. Obligations to companies that are not small businesses rose from $111.03 million in FY2024 to $281.47 million in FY2026.
• Small business obligations peaked at $259.95 million in FY2024 and were $223.12 million in FY2026.

Federal AI Contracts by Agency

Defense accounted for 85.6% of federal AI contract obligations in FY2025. Civilian agencies have a larger share of the FY2026 figures, partly because Defense data is reported 90 days late.

• The Department of Defense obligated $290.29 million in FY2026, 57.5% of the total, down from 85.6% in FY2025. Part of that drop reflects the 90-day delay in Defense reporting.
• Health and Human Services obligations rose from $11.40 million to $51.20 million. CMS's contract with Index Analytics for AI and machine-learned modeling is one of the largest single HHS awards in the set.
• Homeland Security went from $2.93 million to $39.04 million, about 13 times its FY2025 level. The General Services Administration had the largest proportional increase among the eight agencies shown, from $0.02 million to $24.39 million.

Largest Federal AI Contractors

The table lists the ten companies that received the most AI contract dollars in FY2026.

• The five largest recipients took $139.50 million, 27.6% of FY2026 AI contract obligations.
• Enlighten IT Consulting's total came from one Army contract for cyber analytics and detection that includes AI and machine learning integration.
• ECS Federal was among the top two recipients in both years, with $48.13 million in FY2025 and $39.84 million in FY2026. Its largest award is a Defense Contract Management Agency research and development effort with a lifetime value of $120.49 million.
• Carahsoft, World Wide Technology and Four Points Technology are among the ten largest contractors.

Federal AI Contract Size

Award size shows whether federal AI work comes in small pilots or large programs. These figures use each award's lifetime obligated amount, for the 437 contract awards with obligations in FY2026.

• The median award was $953,600, and 213 of the 437 awards (48.7%) were worth $1 million or more over their life.
• 37 awards were worth $10 million or more. The largest was a General Services Administration award to Booz Allen Hamilton worth $243.85 million, for the Enterprise Machine Learning Analytics and Persistent Services 3 (EMAPS 3) task order.
• In FY2025, the median award was $1.00 million across 531 awards, and 30 were worth $10 million or more.

Federal AI Contracts by Industry Code

Every federal contract carries a NAICS industry code that describes the work bought. AI dollars split between research and development and ordinary IT services codes.

• Research and development (541715) fell from 62.5% of AI obligations in FY2025 to 30.7% in FY2026.
• The three IT services codes (541511, 541512 and 541519) together rose from $120.59 million in FY2025 to $196.84 million in FY2026, from 24.9% to 39.0% of the total.

How the Federal AI Contract Figures Were Calculated

The totals depend on how a contract is identified as AI work and on how USAspending records obligations.

We queried the USAspending.gov API on October 13, 2026 for prime contract awards (definitive contracts, purchase orders, delivery orders and BPA calls) matching the keywords "artificial intelligence" or "machine learning" in fiscal years 2020 to 2026. Obligations are the sum of contract transaction obligations dated within each fiscal year, net of deobligations, as returned by USAspending's spending-by-category endpoints. Small business figures use USAspending's recipient business type filters, which come from the size status the contractor reported for that award in the Federal Procurement Data System, based on its SAM.gov registration. Award counts are contract awards with at least one transaction in the fiscal year. Award size uses each award's lifetime obligated amount. Shares and growth rates were computed from unrounded totals.

Keyword matching misses contracts for AI work whose descriptions do not use either phrase, and it can include contracts that mention AI only in passing. Indefinite delivery vehicles are counted through their orders, and grants and other transaction agreements are left out. Recent fiscal year totals rise as agencies file late records, and Defense data arrives about 90 days after award, so FY2026 is incomplete. Recipient names are shown as recorded in USAspending, with parent companies left unmerged.