In 2001 it cost about $95 million to sequence one human genome. In 2022 it cost about $525. That is a 180,000-fold drop in 21 years, and most of it happened in a single year.

Cost curves that steep usually reorganize an industry. This one did, but not where most people expected. Sequencers got cheap. Everything that happens after the sequencer did not follow. This post looks at one curve and one archive to show where the bottleneck went.

How cheap did sequencing get?

The National Human Genome Research Institute has tracked the cost of sequencing a human-sized genome since 2001. For the first six years the price fell roughly in line with Moore's law, the benchmark NHGRI plots beside it. Then it broke away.

1. 12003: Human Genome Project ends
2. 22008: Next-generation sequencing takes over

Source: NHGRI Genome Sequencing Program, via Our World in Data. The y-axis is logarithmic, so each gridline is ten times the one below it.

View data

year	cost
2001	95263070
2002	61448420
2003	40157556
2004	18519312
2005	13801124
2006	10474556
2007	7147571
2008	342502
2009	70333
2010	29092
2011	7743
2012	5901
2013	5096
2014	4008
2015	1245
2016	1176
2017	1015
2018	1232
2019	606
2020	512
2021	454
2022	525

The curve has two regimes. From 2001 to 2007 the cost fell by roughly 92%, from $95.3 million to $7.1 million. In 2008 it fell another 95% in twelve months, to $342,502. NHGRI attributes that break to sequencing centers moving from Sanger sequencing to second-generation technologies. After that the decline slowed, and the last three data points sit between $450 and $525.

Two cautions. These numbers cover the cost of generating sequence at large sequencing centers, not the full cost of a clinical genome with analysis and interpretation. And the 2022 figure ($525) is higher than 2021 ($454), so the curve is flattening, not still falling off a cliff.

Where did all that cheap DNA go?

Into public archives, mostly. GenBank, the US archive of nucleotide sequences, publishes a release statistic every two months. We took one release per year since 2018.

Source: NCBI GenBank release statistics. Selected releases only. Values are bases in trillions.

View data

release	bases
Oct 2018	0.28
Aug 2019	0.37
Jun 2020	0.43
Apr 2021	0.83
Dec 2022	1.64
Oct 2023	2.43
Oct 2024	4.25
Aug 2025	5.68
Aug 2026	8.24

Look at the record count, not just the bases. GenBank held 209.7 million sequences in October 2018 and 267.4 million in August 2026, an increase of 28%. Bases grew 29-fold. The average record went from about 1,300 bases to about 30,800. The archive is not filling up with more entries. It is filling up with much larger ones.

What changed for the people who use this data?

When a genome cost $95 million, the sequencer was the scarce resource. Teams planned around it. At about $500, a genome is no longer scarce. What is scarce is everything attached to it:

• Storage and transfer for files that now arrive by the terabyte.
• Compute and pipelines that give the same answer twice.
• Phenotype, consent, and clinical context, without which a genome is a string of letters.
• People who can tell whether a result is real.

None of these has an equivalent public cost curve in the sources behind this post. That is the point.

Limits

The cost series we used ends in 2022. We did not add vendor price lists, because a list price is not a measured cost across sequencing centers. GenBank is also one archive among several, and the statistic counts bases, not information.

Sources

1. NHGRI, The Cost of Sequencing a Human Genome. Annual series via Our World in Data.
2. NCBI, GenBank and WGS Statistics, releases 228 to 273.