The learning curves.
When the world decides to make a thing at scale, the price does not drift down — it collapses on a curve, and this report draws four of those collapses from the primary records, indexed honestly on one chart. Reading a human genome: $95 million in 2001, $525 in 2022 — a 181,585-fold fall, halving every 1.04 years, faster than Moore's law ever ran. Solar modules: 497-fold since 1975. Battery cells: 118-fold since 1991. A kilogram to orbit: from tens of thousands of dollars to the low hundreds, by vehicle. And beside the four price curves, one gate that does not obey them — the FDA approves roughly the same number of novel drugs every year regardless — because the report's honest finding is that learning curves bend prices, not institutions.
published 2026-08-29 · data cutoff 2026-08-29 (series as published by their authorities) · edition 1 — updated only by a new dated edition · sources: NHGRI · IRENA/Nemet · Ziegler & Trancik · CSIS · FDA CDER
Four collapses, one chart.
The only honest way to draw $/genome beside $/kg is to index each curve to its own first year = 100 and let the log axis tell the story. The steepest line on this chart is biology's — sequencing fell five decades' worth of Moore's law in two. Real units are stated at every endpoint; a curve missing years is missing on the chart, never smoothed.
Each curve in its own units.
The same three series, un-indexed — $95,263,070 to $525 per genome · $128.27 to $0.26 per solar watt · $9,210 to $78 per battery kWh — each on its own log axis, because the real units are the story once the shape is established. The computed halving times: genome 1.04 years · battery 4.9 · solar 6.4.
Not a line — a scatter of machines.
Launch cost is not one series; it is 61 vehicles across six decades, each a dot in its class. The frontier is the LOWER EDGE of the cloud — and the modern heavy-lift dots sit orders of magnitude below where the heavy lifters of the 1960s started. This view refuses the false line on purpose: the honest shape of this record is a scatter, and the report draws what the record is.
One number the curves do not bend.
Sequencing collapsed 181,585-fold; the pipeline of biology filled with candidates — and the FDA has approved between 22 and 59 novel drugs a year for a decade, with no trend the record supports. The gate is an institution, not a technology, and it moves at an institution's pace. This is the report's honest counterweight: learning curves bend prices; they do not bend gates.
Where every number comes from.
Four published series, one curated table. Genome sequencing cost is NHGRI's own program record; solar module prices are the IRENA/Nemet series; battery cell prices are Ziegler & Trancik's; launch costs are the CSIS per-vehicle compilation — all four retrieved through Our World in Data's grapher exports on 2026-08-29 and stored in data.json beside this page. The FDA table is curated by this desk from CDER's annual Novel Drug Approvals reports, one citation per row, 2015–2024; 2025 is omitted because an open year is not a record.
The index is a transform and says so. First year = 100, per series, log axis — the only honest way to co-plot dollars-per-genome with dollars-per-kilogram. Real units live at the endpoints and in the frontiers view. Halving times are OLS on log₂(value) against year over each full series — the same discipline as every pace figure on this site.
What this page does not do. It fits no line to the launch scatter — that record is per-vehicle and is drawn as it is. It forecasts nothing — every curve ends at its last published point. It does not claim the FDA should move faster; it states that the record shows the gate flat while the curves collapsed, and leaves the argument to the reader.