15 auctions, 2006–2026 · $216B net nominal = $274B in constant 2025 dollars · every figure reproducible fromanalysis/ in the project repo
Price of spectrum by band, across two decades
Pop-weighted net $/MHz-pop per auction, constant 2025 dollars. Log scale — the low-band-to-mmWave spread is ~125×, a propagation-physics premium: sub-1 GHz signals cover wide areas and penetrate buildings; millimeter wave does neither.
Big-3 share of proceeds, auction by auction
AT&T + Verizon + T-Mobile as % of each auction's net proceeds. No time trend — concentration tracks band desirability: the Big 3 dominate beachfront mid/low-band (C-band 96%) and cede share where designated-entity vehicles or fragmented licensing invite everyone else in (CBRS 42%).
Who has paid the most (constant 2025 $)
Net winning bids across all 15 auctions, by parent company where confirmed — legal filing name otherwise. SpectrumCo (a 2006 Comcast/TWC joint venture) and others stay under their own names rather than being retroactively attributed.
Hottest metros, normalized across every FCC geography scheme
The FCC has licensed spectrum under five different area schemes since 2006 (PEA, CMA, BEA, REA, county) — this site is the only place that maps them to one identity per metro. New York is "PEA001" in one auction and "CMA001" in another; here it's one number. PEA and county resolve to BEA exactly; CMA is a population-weighted best match (it's an independent MSA-based scheme that doesn't nest cleanly in BEA — hover a bar for match confidence). 700 MHz's REA-level licenses aren't shown here — REA is a coarser grouping than BEA and can't be split back down without guessing.
Methodology: pop-weighted $/MHz-pop = Σnet ÷ Σ(MHz × pop) per auction, on each auction's own FCC-scored population basis; CPI-U deflated to 2025 dollars (2025/2026 values provisional). Full methodology and reproducible scripts in the repo's analysis/REPORT.md.