SWITCHED ON
The daily technology series nobody asked for but everyone needed
The DNA Hard Drive: Storing the World's Data in Molecules
One gram of DNA can theoretically store roughly 215 million gigabytes of data, stable for thousands of years without power. The world is generating data faster than any storage medium can archive it long-term. DNA might be the only medium dense and durable enough to actually keep up.
DNA has been storing the complete operating instructions for every living thing on Earth for roughly four billion years, in a format so dense and so durable that we can still read genetic information out of woolly mammoth remains frozen in permafrost for tens of thousands of years. The proposal to use that same molecule to store our photos, documents, and institutional archives is not science fiction. It is an active engineering effort with real, if narrow, commercial deployment already underway.
Yesterday we examined workplace surveillance technology — bossware's activity metrics, the measurement validity problem, and the evidence that trust and outcome-based goal setting outperform granular monitoring for actual productivity. Today we are addressing a problem that receives far less public attention than it deserves given its scale: the world is generating digital data at a pace that existing long-term storage technology genuinely cannot sustainably archive, and DNA data storage — encoding digital information directly into synthetic DNA molecules — has moved from a laboratory curiosity to a technology with real, if still niche, commercial deployment as one of the more promising answers.
01 — The Archive Crisis Nobody Is Talking About
Global data creation is projected to continue growing at a substantial compound rate through the remainder of this decade, driven by the AI training data volumes, video content, sensor data from IoT devices, and scientific and medical data covered across dozens of episodes in this series. The storage media used for long-term digital archiving — magnetic tape remains the dominant medium for the coldest, longest-term archival storage tiers, having proven more durable and cost-effective than hard disk drives for data that is rarely accessed — has genuine physical limitations that are becoming increasingly apparent as the volume of data requiring truly long-term (decades to centuries) preservation grows.
Magnetic tape has a practical usable lifespan of roughly thirty years under proper storage conditions before signal degradation becomes a genuine data integrity risk, requiring active migration to fresh media — a substantial and recurring operational cost for archives holding petabytes to exabytes of data, and a real risk of data loss for archives that fail to maintain rigorous migration schedules. Hard disk and flash storage have even shorter reliable lifespans and require continuous power for the storage arrays and cooling infrastructure. The fundamental problem is that every mainstream digital storage medium requires active maintenance, periodic media refresh, and continuous energy input to remain reliable over genuinely long timescales — none of which describes what long-term archival storage of humanity's most important records (scientific data, cultural heritage, institutional and legal records) ideally requires.
DNA, by contrast, requires no power, no active maintenance, and has demonstrated genuine multi-thousand-year stability under the right storage conditions — the sequencing of DNA extracted from ancient specimens, including the woolly mammoth genome sequenced from permafrost-preserved remains over a million years old, is direct empirical evidence of the medium's durability under favourable conditions, a timescale that dwarfs anything conventional digital storage media can approach.
02 — How DNA Storage Actually Works
Encoding digital data into DNA requires translating binary data (the ones and zeros of conventional digital storage) into the four-letter genetic alphabet (A, C, G, T) that DNA uses to encode biological information, using an encoding scheme that maps binary sequences to nucleotide sequences while accounting for the specific chemical and biological constraints of DNA synthesis and sequencing — avoiding sequences that are difficult to synthesise reliably or prone to sequencing errors, and incorporating error-correction redundancy similar in principle to the error correction used in conventional digital storage and transmission.
The encoded sequence is then physically created through DNA synthesis — the same underlying chemical process covered in our Season Two synthetic biology episode, used here not to create functional biological genes but simply as an extraordinarily dense and stable data storage medium. Reading the data back out requires DNA sequencing, using the same next-generation sequencing technology that has become dramatically cheaper over the past two decades (the cost of sequencing a human genome has fallen from approximately $3 billion for the original Human Genome Project to under $200 today, a cost trajectory we noted in Season Two's synthetic biology episode), which is precisely what has made DNA data storage newly practical rather than merely theoretically interesting.
03 — Who Is Actually Building This
Microsoft Research, in partnership with the University of Washington, has conducted some of the most extensive DNA data storage research to date, having successfully encoded and retrieved a range of test datasets including digital video, images, and text documents, and has demonstrated random access — the ability to retrieve a specific piece of data from within a much larger DNA archive without needing to sequence the entire archive, an essential capability for the technology to be practically useful rather than merely a proof of concept for sequential archival storage.
Catalog Technologies and Twist Bioscience, among a small number of specialised DNA data storage companies, have moved toward commercial services, with Twist Bioscience notably partnering with the Global Music Vault and other cultural preservation initiatives to encode selected data into DNA for genuinely long-term preservation purposes, positioning the technology specifically for the archival use case where its cost and speed disadvantages relative to conventional storage matter less because the data is being written once and preserved for an extremely long, low-access-frequency horizon, rather than for active, frequently accessed storage where conventional media remain vastly more practical and economical.
04 — Why It Is Not Replacing Your Hard Drive Anytime Soon
The honest constraints on DNA data storage's near-term practicality are significant and worth stating clearly, because the "one gram stores 215 million gigabytes" statistic that circulates widely in popular coverage, while technically accurate as a theoretical density calculation, obscures the practical cost and speed realities that currently confine the technology to a narrow archival niche rather than general-purpose storage.
DNA synthesis remains, despite the dramatic cost reductions of the past two decades, orders of magnitude more expensive per byte of stored data than conventional storage media — writing data into DNA currently costs roughly thousands of dollars per megabyte, compared to fractions of a cent per megabyte for magnetic tape, a cost differential that must fall dramatically for DNA storage to become viable beyond the highest-value, longest-timescale archival applications where the alternative is not "cheap storage" but "eventual guaranteed data loss without active, costly migration." Write and read speeds are similarly far slower than conventional storage — encoding and synthesising even modest datasets currently takes hours to days, and sequencing to retrieve the data takes further time, making the technology entirely unsuitable for any application requiring fast or frequent data access, and appropriate only for true "cold storage" archival use cases where data is written once and rarely, if ever, retrieved.
05 — The Realistic Niche and the Bigger Storage Picture
The realistic near-to-medium-term application for DNA data storage, consistent with where the actual commercial deployment has concentrated, is the highest-value, longest-timescale archival category: national and cultural heritage archives, critical scientific datasets, and institutional records where the value of genuinely multi-century preservation without active maintenance outweighs the currently substantial cost premium — a genuinely narrow but genuinely important niche, analogous in some respects to how this series has found many technologies (quantum sensing, gene drives, advanced nuclear designs) occupy valuable but bounded niches rather than delivering the sweeping general-purpose transformation their most enthusiastic coverage sometimes implies.
The broader digital storage crisis this technology partially addresses connects to a theme this series has returned to repeatedly: the energy and physical infrastructure demands of the AI and data economy, covered extensively in our discussions of AI electricity demand and data centre concentration, extend beyond active computing infrastructure into the less-discussed but genuinely significant energy and resource cost of maintaining humanity's growing digital archive over genuinely long timescales. DNA storage's zero-maintenance, zero-power stability profile is not merely a scientific curiosity — it represents a genuinely different approach to a resource and sustainability problem that conventional storage technology roadmaps, focused overwhelmingly on active-use speed and cost rather than passive long-term stability, have not adequately solved.
Tomorrow we are addressing a topic that has been implicit across this entire season's opening run — the economics and technology of the semiconductor industry's next frontier beyond the chips themselves: advanced packaging and chiplet architecture, and why how chips are assembled is becoming as strategically important as how they are manufactured. See you then.
Switched On is a daily technology series covering the ideas, systems, and arguments shaping the digital world. Opinionated. Witty. Occasionally wrong. Always worth the argument.
