What Whole Genome Sequencing Costs in 2026, and What the Price Buys

A 30x short-read human whole genome costs between $200 and $600 at a high-throughput research core in 2026. A direct-to-consumer provider that returns your FASTQ or CRAM files charges $400 to $1,000. A CLIA-certified laboratory that sequences, interprets, and issues a signed clinical report charges $1,500 to $5,000. Long-read WGS (PacBio HiFi or Oxford Nanopore) runs roughly $700 to $1,500 at 15-30x. The spread is mostly about what fraction of the total workflow the price includes, and the interpretation end of that workflow costs more than the wet lab. Sequencing chemistry does not explain most of it.

Where the money goes

The published cost analyses are consistent on this point: reagents are the minority of the total. A detailed micro-costing of an academic diagnostic laboratory found that the sequencing consumables for a genome were a fraction of the per-sample cost once library preparation labor, instrument depreciation, quality control, data storage, and the salaried time of analysts and clinical geneticists were counted ¹. A more recent micro-costing in oncology reached the same structure, with the gap between a single-gene assay and WGS narrowing considerably once you account for the fact that the sequencing step scales far better than the interpretation step ². Informatics and curation have been the persistent bottleneck since the technology entered clinical use ³.

So when you compare a $221 core-facility genome against a $3,500 clinical genome, you are comparing a FASTQ file against much more than a FASTQ file. The clinical price also buys alignment, variant calling, and annotation against ClinVar and gnomAD. It buys manual curation of candidate variants by a board-certified geneticist, orthogonal confirmation of reportable findings, and a document that a physician can act on. Both numbers are real. They are quotes for different products.

Depth, and what you get for each increment

Depth is the single largest lever on sequencing cost, because it maps almost linearly onto flow cell real estate. Standard practice for germline small-variant calling is 30x mean coverage, which is where the common callers were trained and validated. At 30x with 150 bp paired-end reads you will call roughly 4 to 5 million SNVs and small indels against GRCh38, with sensitivity above 99% for SNVs in high-confidence regions.

Dropping to 15x saves half the sequencing cost and costs you heterozygote sensitivity, particularly for indels, where allele balance noise at low depth turns real heterozygous calls into no-calls. Going to 60x or higher buys you somatic sensitivity (detecting variants present in a minority of cells) and better performance in segmental duplications, which matters for cancer work but does very little for a healthy person’s germline baseline ⁴. Above 30x, per-genome cost rises faster than useful information for the germline case.

One number to check before you buy: mean coverage is not callable coverage. Ask the provider for the fraction of GRCh38 primary assembly at ≥20x. The mean does not tell you this. A genome with 32x mean and a badly skewed GC distribution from a PCR-based library can have worse callability than a clean 28x PCR-free genome. Run mosdepth --by 1000 --fast-mode sample sample.cram on your own data and look at the distribution. The average is not the number to read.

Short read versus long read

Illumina’s dominance is eroding, and the pricing pressure from Element AVITI, Ultima UG100, MGI, and Complete Genomics is the main reason the $200 genome exists at all ⁵. For a germline baseline, short reads at 30x remain the best value per unit of clinically interpretable information. That is where the variant databases, the pathogenicity classifications, and the benchmark truth sets such as GIAB HG002 live.

Long reads cost more and answer different questions. PacBio HiFi at 15-20x resolves structural variants, phases haplotypes across tens of kilobases without a trio, and reads through repeat expansions and the pharmacogene CYP2D6 region where short reads produce ambiguous calls. If you care about structural variation or phasing, the premium is worth paying. If you want a well-characterized set of small variants and a polygenic score substrate, it is not.

The costs after the invoice

Sequencing is a one-time charge. Data custody is not, and most people underestimate it. A 30x genome produces roughly 90-100 Gb of bases. That is about 40-55 GB of gzipped FASTQ, 80-100 GB of BAM, and 15-20 GB of CRAM referenced to GRCh38.

We would store CRAM and discard FASTQ once you have verified the CRAM round-trips. Convert with:

samtools view -T GRCh38_full_analysis_set_plus_decoy_hla.fa \
  -C --output-fmt-option level=8 -@ 8 -o sample.cram sample.bam
samtools index sample.cram

At S3 Standard prices (about $0.023/GB/month), a 20 GB CRAM is roughly $0.46 a month. In Glacier Deep Archive it is under a penny, with a retrieval delay of twelve hours and an egress charge when you pull it. Keep one copy in deep archive and one on a local disk you control. The CRAM plus the exact reference FASTA is the durable artifact. Everything else, including the VCF, is a derived file you can regenerate.

Recomputing is cheap. BWA-MEM2 plus GATK HaplotypeCaller in --emit-ref-confidence GVCF mode takes roughly 20-30 CPU-hours for a 30x genome, which is a few dollars on spot instances. DeepVariant with --model_type=WGS on a single A100 finishes in well under two hours and gives better indel precision on most benchmarks. Illumina’s DRAGEN, on FPGA or in the cloud, does alignment through calling in about 30 minutes for roughly $5-10 per genome. Reanalysis matters because variant interpretation changes: a variant of uncertain significance today may be reclassified in three years, which is a documented pitfall of treating a WGS report as a permanent answer rather than a snapshot ⁶.

Failure modes that a low price can hide

A few things to verify before you send money. First, confirm you receive raw reads. A report or a filtered VCF alone is not enough. Some low-cost offerings are genotyping arrays or low-pass sequencing with statistical imputation, which produce a file that looks like a genome and is not one. Ask explicitly for FASTQ or CRAM plus the alignment recipe.

Second, ask which reference build. GRCh38 with the decoy and alt contigs is the current standard. T2T-CHM13 is better assembly but has thinner annotation support. Legacy hg19 output in 2026 signals a stale pipeline.

Third, check contamination and sample identity yourself with verifyBamID2, and check sex chromosome coverage against what you expect. Sample swaps happen, and they happen more at high throughput and low price.

Fourth, understand that a research-grade genome is not a diagnostic. You may be sequencing because of a specific clinical question, a family history, or a suspected inherited condition. In that case, order through a CLIA-certified clinical laboratory and work with a genetic counselor or physician. Diagnostic-grade WGS has demonstrated real clinical value, including as a single comprehensive test replacing panels of separate assays in childhood B-cell acute lymphoblastic leukaemia ⁷, and health-economic modeling supports its use in personalized oncology ⁸. That value depends on the interpretive and confirmatory layer, which is precisely what the cheap genome omits.

What we would buy

For a healthy adult building a molecular baseline, use 30x PCR-free short-read WGS with alt-aware alignment to GRCh38. Ask for CRAM and gVCF to be returned with no vendor lock on the data. Pay the extra $100-200 for PCR-free library prep, because it flattens GC coverage bias and improves indel calling more than a few extra x of depth would. Skip the interpretation add-ons from consumer providers, which are mostly ClinVar lookups you can run yourself, and spend that money on long-read sequencing later if structural variation becomes relevant to a question you have.

Questions people also ask

How much does a WGS test cost? Between $200 and $600 for 30x short-read sequencing at a research core with raw data returned, $400 to $1,000 direct to consumer, and $1,500 to $5,000 for a clinical genome with a signed interpretive report from a CLIA laboratory.

Is WGS covered by insurance? Sometimes, and only in a clinical context. Coverage is routine for critically ill infants with suspected genetic disease, common for rare-disease diagnostic odysseys after other testing has failed, and inconsistent for tumor profiling. Elective WGS in a healthy adult is essentially never covered. Coverage decisions depend on medical necessity documentation, so that conversation belongs with your physician.

How much will whole genome sequencing cost in 2026? The competitive pressure from Element, Ultima, MGI, and Complete Genomics has pushed marginal reagent cost per 30x genome toward $100-200 at scale ⁵. Retail prices are falling more slowly, because labor and informatics do not scale the way flow cells do ¹.

Is WGS worth it? If you want a durable, reanalyzable substrate for pharmacogenomics, carrier status, polygenic scores, and future questions you have not thought of yet, then a $400 30x genome with raw data returned is an unusually good value, because you buy the sequence once and reinterpret it indefinitely. If you want a specific clinical answer today, buy the clinical test and the clinician’s time. The cheap FASTQ will not give you that answer.

Footnotes

1. Kirsten J M van Nimwegen, Ronald A van Soest, Joris A Veltman, et al. Is the $1000 Genome as Near as We Think? A Cost Analysis of Next-Generation Sequencing. Clinical Chemistry, 2016. https://doi.org/10.1373/clinchem.2016.258632 ↩ ↩²

2. Clémence T. B. Pasmans, Bastiaan B. J. Tops, Elisabeth M. P. Steeghs, et al. Micro-costing diagnostics in oncology: from single-gene testing to whole- genome sequencing. Expert Review of Pharmacoeconomics & Outcomes Research, 2021. https://doi.org/10.1080/14737167.2021.1917385 ↩

3. Rama R. Gullapalli, Ketaki V. Desai, Lucas Santana-Santos, et al. Next generation sequencing in clinical medicine: Challenges and lessons for pathology and biomedical informatics. Journal of Pathology Informatics, 2012. https://doi.org/10.4103/2153-3539.103013 ↩

4. Phuc H. Hoang, Sara E. Dobbins, Alex J. Cornish, et al. Whole-genome sequencing of multiple myeloma reveals oncogenic pathways are targeted somatically through multiple mechanisms. Leukemia, 2018. https://doi.org/10.1038/s41375-018-0103-3 ↩

5. Michael Eisenstein. Illumina faces short-read rivals. Nature Biotechnology, 2023. https://doi.org/10.1038/s41587-022-01632-4 ↩ ↩²

6. Christopher J Record, Mary M Reilly. Lessons and pitfalls of whole genome sequencing. Practical Neurology, 2024. https://doi.org/10.1136/pn-2023-004083 ↩

7. Sarra L. Ryan, John F. Peden, Zoya Kingsbury, et al. Whole genome sequencing provides comprehensive genetic testing in childhood B-cell acute lymphoblastic leukaemia. Leukemia, 2023. https://doi.org/10.1038/s41375-022-01806-8 ↩

8. Martijn Simons, Michiel Van De Ven, Veerle Coupé, et al. Early technology assessment of using whole genome sequencing in personalized oncology. Expert Review of Pharmacoeconomics & Outcomes Research, 2021. https://doi.org/10.1080/14737167.2021.1917386 ↩