All of Us hits a half-million genomes and a credibility test: what the US biobank actually proves

At 22:00 UTC on 15 July 2026, the National Institutes of Health's All of Us Research Program quietly crossed a threshold that population genetics has spent two decades arguing about. According to a write-up of an end-of-June NIH announcement, more than 500,000 participants have now had their genomes sequenced, blood samples drawn and electronic health records linked into a single federal dataset, turning what began in 2016 as a moonshot into the largest, most diverse open-access biobank in operation anywhere in the world.

The draw is scale and representation in roughly equal measure. Where earlier reference cohorts over-indexed on people of European ancestry, All of Us was built specifically to recruit from communities historically under-represented in biomedical research. Its sequencing factory, run through a network of academic genome centres, has now produced roughly half a million short-read whole-genome sequences, plus the associated longitudinal health data, and is releasing the aggregated results to registered researchers through a controlled-access portal. The bet behind the programme, articulated in plain language by NIH leadership over successive administrations, is that a research base this broad will power discoveries on diseases where the existing literature has run out of signal.

What half a million genomes is actually worth

A single short-read sequence is not, on its own, much of an answer to anything. The value of a cohort this size is statistical: enough people, with enough ancestral diversity, to find rare-variant associations for conditions where smaller studies have washed out. The programme's published use cases, tracked across NIH's All of Us portal and partner institutions, now span polygenic risk scores recalibrated for non-European ancestry, screening for pathogenic variants in genes such as BRCA1 and BRCA2 across populations previously under-tested, and pharmacogenomic flags that flag safer dosing of common drugs. Each of those applications gets more reliable as the participant base grows past the half-million mark.

The economic logic is equally concrete. A research-grade whole-genome sequence at scale costs a small fraction of what it did in 2015, and a single biobank can substitute for dozens of smaller, single-disease cohorts. European and UK biobanks, the UK Biobank at half a million participants, the FinnGen study at hundreds of thousands, already proved the model. All of Us is the US answer, with the explicit twist that recruitment, sample collection and return-of-results are run through a network of community health centres and federally qualified providers rather than academic medical centres alone.

The counter-narrative: what a biobank does not solve

The standard critique is not that All of Us is useless. It is that the programme conflates two different things: genetic discovery, where scale helps, and clinical translation, where the bottleneck is somewhere else entirely. Most drug pipelines still fail in phase 2 and phase 3 trials because the biology does not replicate, not because the cohort behind the original finding was too small. A bigger biobank accelerates the front end of the funnel without altering the back end. Critics in the genomics community have also argued that the answer to under-representation in medicine is healthcare access, not a wider reference panel; a person with a polygenic risk score but no insurance still cannot act on it.

There is also a documented worry about participant protection. The programme collects genomic data, electronic health records and survey responses and links them under a single participant identifier, then releases de-identified data to researchers inside a controlled-access framework. Whether that framework is robust to re-identification attacks at half a million sequences is a live research question. The NIH counters that its tiered access tier and the prohibition on certain re-identification techniques make the dataset substantially harder to abuse than the typical hospital record. Both claims are testable; neither is settled.

A structural shift in who runs human genetics

Putting half a million US-resident genomes under one institutional roof is a quiet change in the geography of the field. For most of the post-Human Genome Project era, the dominant reference cohorts were European, the sequencing capacity sat in a handful of Anglo-American academic centres, and the publishing record reflected that geography. All of Us, with its explicit ancestry targets and its return-of-results pipeline, is a US-government attempt to rebalance who supplies the reference material that the rest of the world's genomics infrastructure quietly runs on.

That has spillover effects elsewhere. The All of Us dataset is now cited in studies produced by groups with no NIH funding at all, including researchers in Latin America, Africa and South Asia who use the controlled-access portal to calibrate findings from their own smaller cohorts. The structural objection is that this still routes a large share of global genomic infrastructure through one funder in one country, with the corresponding concentration of influence over which scientific questions look tractable and which methods get diffused. The corresponding defence is that no other funder has shipped a comparable cohort at comparable scale in comparable time. Both points are true at once.

What to watch over the next twelve months

Three concrete signals will tell readers whether the programme is delivering on its promise. First, a count of peer-reviewed papers in mainstream journals that depend directly on All of Us data, which NIH already tracks and which is rising but remains modest in absolute terms. Second, the return-of-results tally: how many participants have received clinically actionable findings back, and how many of those findings have translated into documented clinical care. Third, the cohort refresh: whether NIH continues to back-fill recruitment in under-represented communities or quietly shifts the programme toward maintenance mode as the political weather changes around large federal datasets.

None of this settles whether the bet was right. A biobank this size is a fifty-year instrument; the first decade is just plumbing. What the half-million milestone does confirm is that the US has decided to keep building the plumbing, while other countries with comparable ambitions are still drafting the permits.

This publication treats the All of Us milestone as a logistics and governance story, not a scientific breakthrough, the breakthrough, if it comes, will arrive in journals that read very differently from a press release.