DNA as a hard drive: 455 exabytes per gram and a lifespan of millions of years

The digital world is constantly growing, and although we talk about "the cloud," that cloud lives on solid ground, inside gigantic data centers. Enormous warehouses with endless corridors, rows of servers, and colossal electricity consumption now store our photos, videos, emails, and scientific data. In this scenario, one idea is gaining traction: using DNA as a data storage medium, a minuscule molecule with an overwhelming storage capacity.

The promise is powerful: to encode bits into sequences of A, C, G, and T to store everything from a historical document to a video file, and to do so stably for centuries or even more than a million years if properly preserved. The fact that sparks the imagination is well-known: in theory, a single gram of DNA could store up to 455 exabytes of data (455.000 billion GB), a figure that dwarfs current hard drives and silicon memories.

What is DNA storage and why it matters

DNA is the instruction manual for life, and its language is written with four “letters”: adenine (A), cytosine (C), guanine (G), and thymine (T). In computer terms, we can translate zeros and ones into combinations of these bases to create a synthetic sequence that, when read, retrieves the original file. This translation has been proven feasible since 2012, when the contents of a megabyte were successfully encoded and read, opening the door to a new paradigm of digital archiving.

The underlying reason is density: by packing information into molecules, the physical space required plummets. In numbers, we're talking about 455 exabytes per gram . To understand this without a calculator: a small test tube could hold all of Wikipedia along with Facebook , and if we scale it up, the knowledge of our civilization would occupy only a few cubic meters —nothing like the thousands of square meters of data centers.

This perspective is not merely aesthetic. On a practical level, DNA does not require electricity to remain readable: in cool, dry, and dark conditions, the data is preserved. We know this from molecular archaeology, which allows us to read genetic material from remains hundreds of thousands of years old. This behavior makes DNA an ideal candidate for very long-term archiving.

Capacity, comparisons, and the challenge of today's data centers

Data centers are the "cathedrals of the bit." There are more than 2.000 of them worldwide, and on average, each one occupies about 5 hectares (50.000 m²). If we compare that footprint to the compactness of DNA, the impact is clear. It's not just the land and the building: cooling and energy consumption are also massive, and the associated environmental footprint is significant.

Meanwhile, data is exploding. Google processes around 4.650 billion searches daily; YouTube sees nearly 4.700 billion video views per day; Facebook receives over 350 million photos every 24 hours; and Twitter sees approximately 600 million tweets. The global outlook suggests that by 2025, some 463 exabytes of data will be created every day , with a significant portion of humanity still not connected to the internet. That's no small feat.

This surge in demand is putting pressure on current technologies just as they are reaching their physical limits. Backblaze's experience—which monitors 25.000 hard drives in service—provides some clues: after four years, nearly 22% of the drives show wear or failure. Some last more than a decade, others fail very quickly. The conclusion is simple: conventional hardware is not eternal, and constant replacement has both economic and operational costs.

DNA as memory, on the other hand, shifts part of the problem to the chemical and preservation realms. With a stable medium and no need for energy to "keep the information alive," the deep archive—the one we only occasionally review but which must endure—could undergo a paradigm shift and alleviate the pressure on massive infrastructures.

How to code, preserve, and fix errors

Encoding data in DNA involves translating bits into bases. A simple scheme assigns A and C to "0" and G and T to "1". With this mapping, short fragments are written which, when combined, reconstruct any digital file. To reinforce robustness, Reed-Solomon error-correction codes are used , which add intelligent redundancy: if some fragments are damaged, the system can recover the original information.

The biggest leap forward in durability was made by the team at ETH Zurich, led by Robert Grass and Reinhard Heckel. Inspired by how DNA is preserved in fossils, they encapsulated DNA molecules in silica (glass) spheres . Why glass? Because it is a chemically inert material and protects against the elements that most degrade DNA: primarily water and oxygen.

To accelerate the "film of time," they subjected the encapsulated DNA to temperatures of 60, 65, and 70 °C, simulating decades or centuries of deterioration in a matter of weeks. The stability of the DNA within the glass was remarkable. Extrapolating, storage at -18 °C would allow the information to be preserved for more than a million years , a figure that changes our mental framework of what constitutes a long-term archive.

The contrast with a test tube exposed to the open air is striking: DNA survives for only two or three years before becoming unreadable. By encapsulating it in glass and keeping it in a cool, dry, and dark environment, its survival time is multiplied. Furthermore, sol-gel technology makes it easier to create this glass "shell" around the molecules, making the process technically feasible in the laboratory.

Experiments and results: from EBI and ETH Zurich to Microsoft and the UW

The proofs are piling up. In 2012, content was successfully encoded and read from DNA, and soon after, the European Bioinformatics Institute (EBI) in England took the idea further: it stored texts, images, and audio—including Shakespeare's sonnets, excerpts from Martin Luther King's "I Have a Dream" speech, an image of the institute itself, and the landmark article on the double helix—and then retrieved the information with 100% accuracy.

Their methodology combined overlapping fragments, positional indices, and redundancy to ensure reconstruction even if some copies were damaged. The total volume was around 760 KB, and the equivalent DNA was smaller than a speck of dust. From a biosafety perspective, they clarified that this synthetic DNA uses a different "code" and cannot be accidentally incorporated into the genome of a living organism; if it were to enter a body, it would degrade and be eliminated without any functional role.

At Harvard, the idea was also tested with a book of over 53.000 words and 11 images, synthesizing thousands of short fragments onto a glass chip, which were then read using standard sequencing techniques—the same ones we use to study ancient genomes or archaeological samples. This reinforced the idea that the “molecular library” can be accessed with widely available molecular biology equipment.

Back at ETH Zurich, Grass and his team subjected two historical documents (around 83 KB in total, including the Swiss Federal Pact of 1291 and passages from the Archimedes Palimpsest) to thermal stress. After a week at 60–70 °C, the texts remained legible. Their calculations place the durability at around 2.000 years at 10 °C, extending to scales of a million years if preserved at -18 °C. The cost at the time was high—around €1.350 to $2.000 for just 83 KB —but cost trends in genomics are favorable.

Proof of this is the dramatic drop in the price of sequencing a human genome: from several million dollars years ago to just a few hundred dollars today. Along these lines, researchers from Microsoft and the University of Washington built the first device that fully automates the process of storing and reading DNA. With it, they encoded the word "hello" and retrieved it, an engineering milestone that—although it took around 21 hours for 5 bytes —illustrates that automation is already underway.

Prototypes that look to the future: from DNA cassette tape to industry

A recent proposal, published in Science Advances by a team from the Southern University of Science and Technology, revives a very familiar format: the cassette tape. Their device integrates a nylon and polyester membrane with laser-printed barcode patterns . The white areas house compartments for depositing synthetic DNA with encoded files; the black stripes act as hydrophobic barriers to prevent mixing.

Each partition has a unique "address," enabling deposit-many retrieve (DMRM) operations. This means we can store multiple files, retrieve them, delete them, and rewrite them to the same area, emulating the behavior of a hard drive but on molecular media. In figures, a 1.000-meter cassette can hold more than 500.000 partitions and achieve up to 362 petabytes per kilometer —enough, according to the authors, to store several times the content of YouTube in a space smaller than a paperback novel.

This line of research coexists with other initiatives. Microsoft is also working on Project Silica, which explores quartz as an archiving medium: lasers permanently alter the crystal's structure, and then machine learning algorithms read those markings. It's not DNA, but it illustrates the search for ultra-stable and compact archiving media.

The biotechnology industry is also making strides. Catalog, a Boston startup, has developed a system to rearrange prefabricated DNA blocks and write data without synthesizing it from scratch, paving the way for what they call the first “machine” that uses DNA as if it were a physical operating system . In San Diego, Iridia combines DNA and nanotechnology to build units capable of operating in parallel, the beginnings of a “living hard drive.”

From the public sector, IARPA—the US intelligence agency for advanced research projects—is promoting the MIST program, which aims to write one terabyte per day to DNA and read it at a speed ten times faster. “We want to replace current hard drives with denser, more secure, and more durable molecular media,” they explain, in line with the idea that silicon is reaching its physical limits.

Density comparisons provide context: a hard drive has around 10^9 bits per cubic centimeter, while DNA reaches 10^18. It's no wonder that some reports—like the one from the Potomac Institute for Policy Studies—suggest that all the digital data on the planet could fit into approximately one kilogram of DNA. It may sound grandiose, but the physical (and biological) basis is solid, and cold, dry storage offers time windows that far surpass conventional magnetic and optical technologies.

The underlying question after achieving stability and density is: what do we preserve? For Robert Grass himself, the focus should be on selecting “truly important” information that deserves to be archived neutrally for the future. Just as our understanding of the Middle Ages depends on what was preserved, faithful photography of our time will require open criteria, curation, and standards that facilitate interpretation centuries from now.

Challenges remain: today, large-scale synthesis and sequencing are still expensive and relatively slow. However, the cost curve in genomics is stubbornly downward, and automation is already demonstrating technical viability. Correction algorithms, addressing formats, and architectures like the "molecular cassette tape" point toward systems that allow for more practical writing and erasing .

As if further proof were needed that DNA is a long-lasting medium, paleogenetics continues to produce records: DNA has been sequenced from a polar bear dating back approximately 110.000 years, from a horse around 700.000 years old, and 400.000-year-old human mitochondrial DNA has been recovered from the Sima de los Huesos cave in Spain. Although conditions matter—cold helps—cases in relatively temperate caves broaden the map of actual preservation.

The less glamorous, but crucial, aspect is that of everyday chemistry: water and oxygen are the biggest enemies. Therefore, encapsulating the material in glass and storing it in cold chambers minimizes reactions and chain breaks. On a laboratory scale, creating silica spheres with sol-gel techniques has simplified the process, and thermal stress experiments demonstrate that deterioration follows predictable patterns , comparable to those observed in fossils.

To further illustrate this point, it's worth remembering the contrast in size and cost: the test suite with 83 KB of documents cost around €1.350/$2.000 a few years ago. That's expensive when we think in terms of terabytes and petabytes, but not so long ago sequencing a human genome cost millions, and today it's around hundreds. If this trend continues, DNA will go from being an experimental medium to a competitive one for massive archiving and "cold" backups.

The union of biology and technology is no longer science fiction. From DNA “tapes” labeled with barcodes to molecular libraries that don't require electricity, and including alternatives like quartz crystal, the race for a durable and compact storage medium is underway. If one thing seems clear, it's that DNA—with its impossible density and its inherent drive for survival—is emerging as the leading candidate to store, on a grand scale, the digital memory of our species with a temporal robustness that no current hard drive can promise.