← All research data

SeqDesk · revised predictions

Did sequencing cost really keep falling faster than Moore's Law?

NHGRI's curve showed the cost to sequence a genome falling from $95M in 2001 to ~$525 in 2022 — 'suddenly and profoundly out-pacing Moore's Law' from January 2008. We rebuilt it and checked it against the data flood it caused: the rare forecast on this page that mostly came true.

$95.3M
cost per genome, Sept 2001 (NHGRI)
$524.62
cost per genome, May 2022 (NHGRI)
~181,000×
cheaper in 21 years — far faster than Moore's Law
<$100
the sub-$100 genome arrived, 2022–24 — the forecast came true
1001.0K10K100K1.0M10M100M2005201020152020
Cost per genome (NHGRI)If it had tracked Moore's Law (halve every ~24 mo)

NHGRI's cost-per-genome curve on a log axis (copper) against where the cost would be if sequencing had merely tracked Moore's Law from 2001 (dashed grey, halving every ~24 months). The cliff at 2008 — the Sanger-to-next-generation transition — is the 'sudden and profound out-pacing of Moore's Law'. By 2022 the realized cost ($524.62) sits ~140× below the Moore line. The catch is the shape: after the 2008–2012 free-fall the curve flattened markedly (≈$1,000 in 2014 to only ~$525 by 2022), so a straight-line extrapolation of the early slope badly over-predicts the 2015–2022 segment.

What held upThe central claim was correct and rare among forecasts on this page: cost per genome fell ~181,000× (from $95.26M to $524.62) and cost per megabase ~880,000× (from $5,292.39 to $0.006); the super-Moore phase from January 2008 is real; and the implied sub-$100/$200 genome arrived in 2022–24 (Ultima UG 100, Illumina NovaSeq X, MGI). The downstream data flood the cheap cost predicted is real and trackable — GenBank WGS bases up ~4,900× since 2002, to ~33 trillion.
What brokeThe decline was NOT a clean continued exponential: costs plateaued from ~2015–2022 ($1,000 in 2014 to only ~$525 by 2022), so straight-line extrapolation of the 2008–2012 slope over-predicts the later segment. And the headline per-genome number is not the true cost to a researcher: NHGRI explicitly excludes analysis, interpretation, labor beyond the center, and long-term storage — the very costs that have not fallen as fast (the Sboner 2011 'real cost of sequencing' caveat).
Forecast milestoneWhen it looked plausibleWhat actually happened
$1,000 genomemid-2010s2014 — Illumina HiSeq X Ten ✓
Sub-$100 genome~2020s2022–24 — NovaSeq X, Ultima UG 100, MGI ✓
Keeps beating Moore's Lawfrom Jan 2008Yes — ~140× below the Moore line by 2022 ✓
Smooth continued halvingstraight-line extrapolationNo — plateaued ~$1,000 → $525, 2015–22 ✗
From $95 million to under $100
  • 12001$95.26M per genome — NHGRI's first data point: $95,263,071.92 per genome and $5,292.39 per megabase, in the capillary (Sanger) sequencing era.
  • 22008Costs out-pace Moore's Law — With the Sanger-to-NGS transition (454, Solexa/Illumina), NHGRI reports cost per genome plunging ~$3M → ~$230k within roughly a year — the start of the 'sudden and profound out-pacing of Moore's Law'.
  • 32014$1,000 genome — Illumina's HiSeq X Ten breaks the $1,000-per-genome barrier — the symbolic target of the super-Moore decline.
  • 42022$524.62 per genome — NHGRI's last published point: $524.62 per genome and $0.006 per Mb — but the curve has visibly flattened since ~2015 rather than continued halving.
  • 52024Sub-$200 / $100 genome — Illumina NovaSeq X (~$200), Ultima Genomics' UG 100 (~$100 raw reagent) and MGI's DNBSEQ-T20×2 (sub-$100 at scale) realize the implied sub-$100 forecast — the prediction broadly comes true.
  • 62024~33T WGS bases archived — The cheap-cost data flood downstream: GenBank's WGS division reaches ~33 trillion bases (Release 264.0, Dec 2024), ~4,900× its 2002 size — the consequence the cost curve foretold.

SeqDesk · revised predictions. The cost curve is NHGRI's own production-cost accounting (reagents, instruments amortized, labor and overhead at the funded centers), per ~30× human genome; it explicitly EXCLUDES downstream alignment, variant-calling, analysis, interpretation and long-term storage — so the headline number is not the cost of an interpreted genome (the Sboner 2011 caveat). Cost is not a SeqDesk tracker metric, so the forecast is checked against NHGRI's published series; only the downstream consequence (the data flood) reuses live tracker metrics. The Moore's-Law reference halves the 2001 cost every ~24 months. Values are NHGRI's published cost-per-genome points; the May 2022 figure is $524.6246837 (~$524.62).

What it measuresA cost forecast checked against reality — sequencing cost per genome vs a Moore's-Law trajectory
The dataNHGRI / Wetterstrand cost-accounting curve (genome.gov), Sept 2001 → May 2022 production costs
Realized milestonesVendor announcements — HiSeq X Ten (2014), NovaSeq X / Ultima UG 100 / MGI (2022–24)
Downstream floodReuses genbank-wgs-bases (~33T, ~4,900× since 2002), with sra-records and ena-read-run
CadenceCurated — NHGRI cost series + live archive count · validated 2026-06-25
Methodscripts/check-revised-prediction.mjs
DateCost per genome
Sep 2001$95,263,072
Jan 2008$3,063,820
Jan 2009$232,735
Jan 2010$46,774
Jan 2014$4,008 (→ $1,000 genome era)
May 2016$1,176
Feb 2020$645
May 2022$524.62

A revised prediction — and the rare one that mostly came true: the cost forecast held, even if the curve flattened and the 'real cost' of an interpreted genome did not fall as fast. Only aggregate counts are published. · back to all research data