Google DeepMind’s AlphaGenome Atlas covers roughly 9 billion single‑nucleotide variants, the full set of possible human DNA letter changes, as of its 8 September 2026 release.
Scope of the Atlas
The MarkTechPost article dated 8 September 2026 confirms that the Atlas contains pre‑computed molecular‑effect predictions for each of those ~9 billion variants and introduces a single AlphaGenome Variant Impact (AVI) score that ranks variants by predicted impact.
“Google DeepMind has released AlphaGenome Atlas, a catalogue of precomputed predictions for the molecular effects of every possible single‑nucleotide variant in the human genome. That is roughly 9 billion single‑letter changes.” – MarkTechPost
Each variant is annotated with thousands of predictions across many cell types, plus per‑variant feature attributions and a genome‑wide motif collection. The AVI score is a new numeric metric designed to simplify downstream prioritisation for researchers.
Size and technical footprint
The same source notes that the full dataset occupies about 1 petabyte of storage. No conversion to other units is performed, as the packet provides the figure in petabytes.
Providing a petabyte‑scale resource via a free web portal for academic users, and an API for commercial access (announced as “coming soon”), represents a substantial engineering effort. The dataset’s size alone makes it the largest genome‑wide variant‑effect resource publicly disclosed to date.
For academic labs, instant access to pre‑computed predictions eliminates the need to run costly in‑silico simulations for each variant, potentially shortening discovery cycles. Biotech firms developing gene‑editing therapies can use the AVI score to triage candidate variants early, reducing experimental overhead.
Cloud‑based AI services, such as Google Cloud’s genomics offerings, stand to integrate the Atlas as a built‑in data layer. That could create a new revenue stream once commercial licences are opened, while also reinforcing Google’s position in the AI‑driven life‑science market.
Analysts note that the combination of breadth (all possible SNVs) and depth (multi‑cell‑type predictions) is rare. Existing variant‑effect databases typically cover only a subset of known variants or focus on a single tissue context. By contrast, AlphaGenome Atlas delivers a uniform, genome‑wide baseline that can be layered with disease‑specific data.
Company context and next steps
Google DeepMind, headquartered in London, is an artificial‑intelligence subsidiary of Alphabet Inc. According to Wikidata, the organisation employs roughly 10,000 staff and was founded in 2010. The packet does not list a chief executive, and the research note advises confirming that detail against DeepMind’s own site before publication.
The release follows DeepMind’s earlier AlphaGenome model announced in June 2025, which demonstrated the feasibility of large‑scale variant‑impact prediction. The new Atlas builds on that model, scaling predictions to the entire SNV space and packaging them for external consumption.
Commercial access is described only as “coming soon”. The packet does not disclose pricing, usage limits or licensing terms, leaving a key unknown for biotech firms planning to embed the Atlas in proprietary pipelines. DeepMind has not commented on whether the API will integrate with existing Google Cloud genomics services beyond the academic portal.
From a market perspective, the Atlas could pressure other players—such as Illumina’s BaseSpace or the European Bioinformatics Institute—to accelerate their own variant‑effect releases. If DeepMind’s commercial API proves affordable and performant, it may become the de‑facto reference for variant impact, reshaping how genomic data are queried in both research and drug‑development settings.
In summary, the AlphaGenome Atlas delivers pre‑computed predictions for ~9 billion SNVs in a 1‑petabyte public dataset, introducing the AVI impact score as a new prioritisation tool. Its immediate availability to academia and pending commercial rollout position it as a potentially transformative resource for genomics research, biotech pipelines and cloud AI services, though details of commercial terms remain to be clarified.