The thesis is one sentence: the entire AI stack — from foundational research to chip fabrication to data-center power — is subsidized by taxpayers at every level, then captured privately without corresponding public-benefit obligations. The audit's key claim: not one major AI company runs on 100% privately-funded infrastructure.
| Layer | Funding | Estimate |
|---|---|---|
| 1 · Federal research | NSF, DARPA, NIH, DOE, DOD grants (1960s–present) | $100B+ cumulative |
| 2 · CHIPS Act | Direct subsidies + tax credits (2022–2032) | $280B authorized |
| 3 · State / local | Tax abatements, land, infrastructure | $3B+ / year |
| 4 · Grid / utility | Rate increases, infrastructure bonds for data centers | $10B+ / year |
| 5 · Hidden | Public data, educated workforce, the internet itself | incalculable |
Layer 1 alone, in a single fiscal year: NSF $2.05B, NIH $3.05B, DOD $2.035B, DARPA $1.41B, DOE $1.54B, NIST $265M, NASA $121M, USDA $216M — $11.1B in annual federal AI/IT investment (FY2025, per the NITRD supplement). DARPA has driven AI since the field's inception; the technologies underneath every frontier model — neural networks, NLP, high-performance computing — were grown on decades of federally-funded research at universities and national labs.
The audit's real work is definitional: it argues AI infrastructure meets the standard criteria for public infrastructure — publicly funded, essential, broadly relied upon — while remaining under private control with no public-benefit obligations attached. That's the asymmetry it names: the public bore the cost and the risk of the foundational research; the private sector captured the returns without inheriting the duties that usually ride along with public funding.
Because it's the macro-scale version of the same grievance this domain runs on: value produced by many, credited to few, with the contribution of the actual source rounded to zero. In Attribution Capture the erased contributor is the individual human author; here it's the taxpaying public. Same funnel, wider mouth. The audit is the restitution question asked at national scale — if the commons paid for it, what does the commons get back? That connects it to David's broader restitution-charter work and to the grievance ledger, where the same math gets applied to a single disabled citizen's $19.28-a-year of ADA enforcement.