Block 10, Article 4 — The Foundation Cost $900 Billion. The Invoice Was Never Prepared.
The spectrum and the science didn’t disappear by accident. The apparatus that stopped the watching was built in Blocks 8 and 9. What follows is the balance sheet entry — what was there, what was taken, and what the public received in return.
What the public owned before anyone thought to measure it
In 1927, Congress passed the Radio Act and stated a principle that became the legal foundation for American broadcasting: the airwaves belong to the public. Frequencies are a physical property of nature, produced by no one and owned by everyone. The law that followed — the Communications Act of 1934 — built a regulatory system on that premise. Private companies could use public frequencies. They could not own them. In exchange for the license to broadcast, they accepted an obligation: serve the public interest.
The obligation was real when it was written. What the license was worth, no one bothered to calculate.
The broadcast license was framed as a grant conditioned on public service, not a valuable asset in itself. What the license would turn out to be worth once auctions finally priced comparable spectrum — $200 billion and counting — was not in the frame at the time the condition was written, or in 1987 when the condition was removed.
Sixty-five years of free licenses
From 1927 until 1994, the federal government issued broadcast licenses at zero cost. No rent. No auction. No valuation. A comparative FCC proceeding determined who received the license — the award itself was free. This system operated for sixty-five years, a continuous transfer of public resource to private use, with the public receiving the broadcaster’s promise of service in lieu of payment.
The licenses were not worthless. A secondary market for them was robust and active the entire time. When Congress finally authorized spectrum auctions in 1994, the first round raised $617 million. The AWS-3 auction in 2015 raised $44.9 billion. The 5G C-band auction in 2020 raised $81 billion. These figures are not the value of the spectrum. They are the floor price in a competitive market — what companies were willing to bid in public, on the record. The total value conveyed to broadcasters at no charge over sixty-five years is estimated at $200 billion or more. That is the size of the transfer: two hundred billion dollars in public asset value, issued without payment, to private companies who then lobbied aggressively against any policy that would require them to pay for what they held.
The digital transition added a second installment. When broadcasting moved from analog to digital in the 2000s, broadcasters received new spectrum — additional public frequencies — at no charge as part of the transition. The Congressional Budget Office estimated the value of that additional gift at $70 billion. The broadcasters accepted it, used it, and continued opposing spectrum fees.
In 1987, the Fairness Doctrine — the public interest obligation that had been the condition of the free license — was eliminated. The obligation dissolved. The license remained. The public had issued the asset and retired the rent in the same generation. The bill was never sent.
The knowledge the public funded
Block 1 already inventoried this foundation in full: the Morrill Act’s land grant colleges, the GI Bill, DARPA and the internet, and the NIH’s $900 billion cumulative investment alongside Bayh-Dole’s march-in rights — never exercised once in forty-five years, not even for insulin, still rationed at $300 to $400 a vial for a drug that costs $2 to $6 to make. What Block 1 established as inheritance, this article prices. The land grant college, defunded from 1980 forward and rebuilt on tuition, is now a $1.84 trillion debt instrument held by 42.8 million borrowers — the clearest conversion anywhere in this series of a public good into a private tax on economic participation.
The intelligence we built together
The AI industry’s current valuation is measured in trillions. McKinsey projects cumulative global capital expenditures on AI data centers will reach $5.2 trillion by 2030, with approximately $2 trillion in the United States. This industry was built on a foundation assembled with public money over fifty years.
The AI industry runs on the same DARPA-and-public-university foundation Block 1 already traced — the internet, developed by public universities and transferred to private infrastructure in the 1990s without a royalty — plus one layer Block 1 didn’t have to price yet: the training data itself. Common Crawl, the nonprofit that scraped and organized the public internet into the datasets every major AI model trains on, took $250,000 each from Anthropic and OpenAI in 2023 for datasets worth billions. The Books3 dataset alone scraped 196,000 books whose authors were never paid or notified. OpenAI told the House of Lords outright that training today’s leading models without copyrighted material would be impossible. It used the material anyway.
One hundred and sixty-six active federal copyright lawsuits were working through the courts as of April 2026. The technology industry spent over $60 million on federal lobbying in 2024. The committee chairs who would hold the hearings received some of it. The authors and publishers whose work trained the models had no comparable investment reaching the same committees — no lobbying operation at that scale, no seat at the table where the exemptions get written. A lawsuit, filed after the fact, was the only channel available to them.
The structure is the same structure this series has been documenting since Block 1, Article 1. Public investment built the foundation. Private industry took the value. The public is now being billed — through subscription fees, through productivity displacement, through the concentration of a new commanding industry in a handful of companies — for access to the intelligence it collectively produced.
The unsent invoice
The electromagnetic spectrum: $200 billion. The pharmaceutical knowledge base: $900 billion in NIH investment, converted to private patents. The internet infrastructure: built on DARPA research, transferred to private hands without a royalty. The land grant university: rebuilt as a $1.84 trillion debt instrument. The AI training data: the accumulated written output of human civilization, scraped without payment.
Did you know the federal government has the statutory authority to license publicly funded pharmaceutical patents to competing manufacturers — and has never used it once?
The foundation cost $900 billion. The invoice was never prepared. Block 11 opens on what happens when the industry built on that unbilled foundation starts displacing the workforce that funded it.
Check the auction record — every dollar the government has collected for spectrum since 1994 is public, and so is the sixty-five years before that when the answer was zero.
fcc.gov/auctions-summary
— — —
Steve Sagnotti
is a serious amateur photographer, writer, and technologist based in Oregon. With his camera he tries to capture common images not often seen, leading to common questions not often asked.
© 2026 Steve Sagnotti
— — —
Sources
TechCrunch’s article on OpenSecrets data confirms OpenAI ($1.76M) + Anthropic ($720K) 2024 figures well over $2M combined, and the OpenSecrets AI lobbying explosion piece is the strongest general source. Compiling the full B10 A4 link list now.
Here’s the copy-paste drop-in for B10 A4, one source per line:
Radio Act of 1927; Communications Act of 1934.
https://www.law.cornell.edu/uscode/text/47/301
FCC, “Auctions Summary.”
https://www.fcc.gov/auctions-summary
FCC AWS-3 auction results, 2015 (Auction 97).
https://www.fcc.gov/auction/97
FCC C-band auction results, 2020-2021 (Auction 107).
https://www.fcc.gov/document/fcc-announces-winning-bidders-c-band-auction
CBO auction revenue underestimation pattern, ~$97 billion over 20 years.
https://epicforamerica.org/federal-budget/dont-underestimate-spectrum/
$200B+ spectrum transfer estimate, New America Foundation/OTI.
https://www.newamerica.org/oti/in-the-news/coalition-use-fcc-spectrum-auction-money-to-tackle-digital-divide/
NIH funding for drugs approved 2010-2019, $164B and $187B figures, PLOS ONE / JAMA Health Forum studies.
https://pmc.ncbi.nlm.nih.gov/articles/PMC10370755/
NIH funding for drugs approved 2010-2019, $230B figure (2000-2019 window), Institute for New Economic Thinking.
https://www.ineteconomics.org/perspectives/blog/us-tax-dollars-funded-every-new-pharmaceutical-in-the-last-decade
GAO $97B basic research obligations, FY2017-2021.
https://www.gao.gov/products/gao-23-105656
Bayh-Dole Act, Pub.L. 96-517 (1980), march-in rights context.
https://www.congress.gov/crs-product/IF12582
March-in rights never exercised, GAO-09-742 (2009).
https://www.gao.gov/assets/a293024.html
Insulin manufacturing cost vs. retail price.
https://www.wusa9.com/article/news/verify/insulin-costs-about-10-to-make-but-retails-for-nearly-300-pharmaceutical-companies-eli-lilly-novo-nordisk-sanofi-pbms-insuli/65-73a3cafd-3340-45cd-8324-a5e3e1c78fa5
Morrill Act of 1862, Public Law 37-108, National Archives.
https://www.archives.gov/milestone-documents/morrill-act
GI Bill (Servicemen’s Readjustment Act of 1944), Public Law 78-346, National Archives.
https://prologue.blogs.archives.gov/2014/06/09/on-display-gi-bill-of-rights/
Student loan debt $1.84 trillion, 42.8 million borrowers, Education Data Initiative.
https://educationdata.org/student-loan-debt-statistics
McKinsey AI infrastructure projections, 2025 ($6.7T by 2030).
https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-cost-of-compute-a-7-trillion-dollar-race-to-scale-data-centers
Common Crawl Foundation; $250K donations from Anthropic/OpenAI, 2023.
https://www.newsmediaalliance.org/wp-content/uploads/2026/04/Letter-to-Common-Crawl-4.29.2026.pdf
Books3 dataset, 196,640 books, copied from Bibliotik.
https://torrentfreak.com/anti-piracy-group-takes-prominent-ai-training-dataset-books3-offline-230816/
OpenAI House of Lords written evidence on copyrighted training materials (LLM0113).
https://committees.parliament.uk/writtenevidence/126981/pdf/
Tech industry lobbying spending, OpenSecrets.
https://www.opensecrets.org/news/2024/06/lobbying-on-ai-reaches-new-heights-in-2024/
Leave a Reply