If your reading of US AI policy assumes that federal financial commitment is comparable in scale to private-sector AI activity and that public investment is shaping the trajectory, that assumption is off by more than an order of magnitude. The US federal government invested approximately $20.4 billion in AI-related activities cumulatively across the 2013-2024 period: $15.9 billion in grants, $3.9 billion in contracts, and $650 million in Other Transaction Agreements (OTAs). Over the same period, US private investment in AI in 2025 alone reached $285.9 billion. The annual private investment is roughly 14 times the cumulative public investment over 12 years. The fiscal weight of US AI strategy now sits with the private sector, not the state.

The composition of US public AI investment is informative. Grants dominate at 78% of the cumulative total. Grant spending has accelerated since 2020: 2024 alone accounted for $5.05 billion in grants, 32% of the cumulative grant total over the 12-year period. Contracts at $3.9 billion are smaller in total but typically larger per-award (median $150,000 vs $304,000 for grants). OTAs at $650 million are the smallest category but have the highest median value at $1 million per award.

The agency distribution shows where federal AI investment concentrates. For contracts and OTAs, the Department of Defense leads at 74.1% of 2024 spending and 73% of cumulative spending, $4.6 billion. Department of the Treasury at 7.2%. Department of Veterans Affairs at 5.1%. All other agencies are below 5% each. For grants, the Department of Health and Human Services (which includes NIH) and the National Science Foundation each account for roughly 40% of total grant funding by 2024. HHS has accelerated sharply since 2020; NSF held the leading position before HHS overtook it.

Three structural observations about the public-private balance follow.

The first observation: the 14x ratio between annual private investment and cumulative public investment understates the actual gap. The $285.9 billion private investment figure is one year (2025) of investment. The cumulative US private AI investment from 2013-2024 is roughly $757 billion. The cumulative public-to-cumulative private ratio is therefore about 1:37. The fiscal weight of US AI activity is overwhelmingly private.

The second observation: the geographic and institutional concentration of public investment is also significant. Virginia received $1.09 billion in contracts and OTAs (driven by ECS, the top contractor by total awarded value). California $0.67 billion. Maryland $0.55 billion. These three states accounted for nearly 60% of total contract and OTA spending between 2013 and 2024. Grants are more broadly dispersed: California ($2.37 billion), Massachusetts ($1.3 billion), and New York ($1.15 billion) lead but represent less than 16% of the total. The contract concentration likely reflects proximity to major federal agencies; the grant dispersion aligns with the federally-funded research footprint across universities.

The third observation: the agency split between defence (74% of contracts and OTAs) and health/science (80%+ of grants) defines two distinct federal AI policy tracks. The defence track is highly concentrated, contractor-mediated, and oriented toward operational AI deployment. The health and science track is highly dispersed, grant-mediated, and oriented toward research capability. These two tracks operate semi-independently: the agencies, contractors, recipients, and policy goals differ. Reading "US public AI investment" as a single variable misses the structural difference between these two tracks.

The data observation: US public AI investment is structurally small relative to US private AI investment, concentrated in defence (for contracts) and health/science (for grants), and geographically concentrated in a handful of states. The trajectory through 2024 shows continued growth but at a pace that does not close the gap with private investment.

Three implications for policy and strategy planning follow.

The first implication: US AI policy that depends on public investment to shape industrial direction operates from a weak fiscal position. The federal government cannot meaningfully shape AI capability concentration through direct funding when its annual capacity is roughly 7% of annual private investment. Policy tools other than direct investment (regulation, procurement preferences, export controls, education, partnerships) are the lever where federal influence can be meaningful. The US AI Action Plan announced in July 2025 leans into infrastructure permitting, export promotion, and procurement framing rather than direct investment, which is consistent with the fiscal reality.

The second implication: the defence concentration of contract investment makes the Department of Defense a structurally significant AI customer in ways the broader federal portfolio is not. DoD AI procurement decisions shape vendor positioning, technical standards, and capability concentration in specific domains (cyber, autonomy, intelligence). The DoD AI customer relationship is now a meaningful strategic variable for many AI vendors. The non-DoD federal AI customer relationship is smaller in fiscal terms but spread across agencies and oriented toward different use cases.

The third implication: the HHS and NSF grant flows define where US public research support for AI is concentrated. Health AI, scientific AI, and academic AI research are the primary beneficiaries. Organisations whose AI work intersects with these domains have access to a substantial grant pool. Organisations operating in domains less covered by federal grants (industrial AI, consumer AI, financial AI) operate primarily in the private investment space without significant federal funding overlay.

The data observation: the US public investment in AI is real but small relative to the private weight. The federal influence on AI direction operates through policy levers other than direct investment. Strategic plans that engage with US federal AI policy should be calibrated against this fiscal reality. Plans that expect federal funding to materially shape industrial AI trajectories are reading the 1960s framing of federal R&D leadership (when public R&D investment was meaningful as a share of total) rather than the 2025 framing where private investment dominates by 14x.

For executive teams planning US AI strategy, the planning anchor should engage with the actual fiscal scale of federal AI investment. The federal government is a meaningful AI customer in defence, a meaningful funder of academic research, and a small fiscal participant in the broader AI economy. Plans that match the actual structure will be more accurate than plans that read federal AI policy as a primary direction-setting variable.

Sources

  • Primary: Stanford AI Index 2026, Chapter 8 (Policy and Governance) 8.5 — hai.stanford.edu/ai-index/2026
  • Federal procurement data: Stanford AI Index 2026 analysis of Federal Procurement Data System (FPDS) — contracts, grants, and OTAs 2013–2024
  • Private investment context: Quid, 2025 (via Stanford AI Index Chapter 4) — US private AI investment 2013–2025
  • Policy framing: US AI Action Plan, July 2025 — federal AI policy direction