The $60 genome, and why the cheap instrument sets the ceiling on the models

Ultima Genomics announced a consumable set this week, at the American Society of Human Genetics meeting, that it says brings a whole human genome at 30x coverage to $60.

The change is physical: a denser flow cell that doubles output to 20 billion reads per run, so the fixed cost of a run is spread across more data.

On a single instrument, the company puts the ceiling at up to 60,000 genomes a year.

A caveat worth stating plainly. This is a company press release, so the figures are self-reported, they're quoted as "up to" maximums, and the release doesn't spell out what the $60 does and doesn't include. Treat it as a direction, not a receipt.

The direction is the point, and it works at two levels.

Practical: at that price, sequencing stops being something you ration. Population-scale studies, routine clinical use, repeat sequencing of the same patient over time all move from "can we afford it" to "why wouldn't we."

Structural: every one of those genomes is training data. The AI models now reading biology were built on public collections of real sequences. When sequencing gets this cheap, the commons those models learn from grows faster. The cost curve on the instrument quietly sets the ceiling on the models.

Read the primary source →