Imagine a storage medium you could hold in your palm that would hold every film ever made, every photograph ever taken, every book ever printed — and would still be readable in a hundred thousand years, without power, without maintenance, without a single moving part.
That is the promise of DNA data storage, and it is not a fantasy conjured up by science fiction. It is a serious engineering program with real money behind it, real prototypes, and a handful of numbers that are usually left out of the story.
Start with what genuinely happened. In 2016, Microsoft bought ten million long oligonucleotides from Twist Bioscience to run DNA encoding experiments. In 2020, Microsoft, Twist Bioscience and Western Digital formed an alliance. In 2026, a Twist spin-off called Atlas Data Storage published a roadmap with a headline figure that travelled fast: thirteen terabytes in a single drop of water. That works out to roughly 260 terabytes per cubic centimetre, which the company claimed is a thousand to fifteen hundred times denser than an LTO-10 tape cartridge.
That figure is worth reading carefully, because it is a target, not a shipping product. There is no drop of water holding thirteen terabytes that you can buy, or borrow, or read.
Here is the first number that does not survive contact with the details: density. The spectacular figures you see quoted — tens of exabytes per gram — are almost always theoretical or derived upper bounds for a given encoding scheme, not measurements. A 2026 paper in Nature Communications on bijective coding, for instance, describes a scheme at 1.66 bits per nucleotide and extrapolates a ceiling near 41 exabytes per gram. Meanwhile actual prototypes remain in the gigabyte range. If you go looking for a single authoritative figure for the highest density actually demonstrated and verified, you will not find one, because it does not exist in a form anyone agrees on. Anyone who tells you that humanity has already achieved some specific number of gigabytes per gram is quoting an extrapolation as if it were a laboratory result.
The second number is worse, and it is the one that decides whether any of this matters. Synthesising DNA currently costs on the order of five thousand dollars per megabyte. Depending on which cost basis you use, that is roughly a million times the cost of an SSD and about two and a half million times the cost of tape. For DNA storage to be commercially serious, that figure needs to fall below one dollar per megabyte — a four-thousand-fold reduction. Nothing in the current trajectory suggests that happens soon, and no one has demonstrated it working at any meaningful scale.
And there is a third set of constraints that rarely makes the headlines. Writing is slow: throughput is measured in kilobytes per second, with a widely cited 2021 milestone reaching one megabit per second. Random reads are slower still — sequencing takes anywhere from twenty-four to seventy-two hours, so a read is an event you schedule, not an operation you perform. Errors are dominated by insertions and deletions, especially in homopolymer runs, which means the error-correcting codes inherited from magnetic storage do not fit and have to be rebuilt from scratch. Practical encoding densities land between 1.5 and 1.98 bits per nucleotide.
None of this makes DNA storage a fraud. It makes it something specific: a cold archive technology. If you need to write something once, store it for centuries, and read it rarely, or almost never, DNA is genuinely the best medium anyone has proposed. If you need a working filesystem, it is not a candidate at all.
Which leaves the interesting question, and it is not an engineering question. A medium that costs a million times more than tape and takes hours to read does not compete with disks. It competes with the decision to keep something at all. When writing is expensive and reading is slow, you stop asking how much can we store and start asking what would we regret losing. The cost becomes a filter, and the filter is the whole point.
A drive that holds everything holds nothing in particular. A medium you can only afford to use sparingly forces a choice — and a choice, made deliberately, is the only thing that has ever made an archive mean anything.
https://3026alive.com