Can Data Be Chemically Stored in DNA?

Inside the Science of DNA-Based Digital Data Storage

Budding Forensic Expert · Science & Technology Review · Forensic Biology / Molecular Information Science

Abstract

The volume of digital information produced worldwide is expanding far faster than the capacity of magnetic and optical storage media, and this gap has pushed researchers to look at molecular alternatives. DNA — the four-letter nucleotide alphabet that already carries biological information in every living cell — has been repurposed since the early 2010s as a medium for storing arbitrary digital files. Encoding maps binary or symbolic data onto sequences of adenine, thymine, cytosine and guanine while respecting biochemical constraints that make the resulting molecules synthesizable and sequenceable. Chemical or enzymatic synthesis then "writes" this sequence into physical DNA strands, which can be preserved in dried, encapsulated, or refrigerated form for extended periods. Reading the data back requires DNA sequencing followed by computational decoding, with redundancy and error-correcting codes compensating for synthesis, storage, and sequencing errors. Landmark demonstrations have stored text, images, audio, video, and even a full computer operating system in DNA, and random-access schemes now allow individual files to be retrieved from pools containing millions of molecules. DNA's theoretical information density vastly exceeds that of hard drives or tape, but practical systems carry substantial overhead from indexing, redundancy, synthesis cost, and slow, energy-intensive read/write cycles. This review traces the science of DNA data storage from its chemical basis through synthesis, preservation, sequencing, error correction, and random access, distinguishes demonstrated capability from experimental and speculative work, and considers what would need to change before DNA becomes a practical archival storage medium — including brief, carefully bounded implications for forensic and information-security contexts.

Key Takeaways

• Digital information can be chemically represented as a sequence of the four DNA nucleotides (A, T, C, G) using an encoding algorithm.
• DNA synthesis is the molecular "writing" step; DNA sequencing is the molecular "reading" step.
• No single universal binary-to-DNA mapping exists — different research groups use different encoding schemes, each balancing density against biochemical constraints.
• Error correction (redundancy, fountain codes, Reed–Solomon-style codes) is essential because synthesis, storage and sequencing all introduce errors.
• DNA is attractive chiefly for archival, "cold" storage — infrequently accessed, long-horizon data — rather than for active, frequently rewritten storage.
• Current barriers are overwhelmingly practical: synthesis cost and throughput, sequencing turnaround, random-access complexity, and automation — not the basic chemistry, which is well established.
• Every historical and technical claim in this article is drawn from a verified, peer-reviewed source listed in the References section.

---

Introduction

Conventional digital storage records information as physical or magnetic states — the orientation of magnetic domains on a hard-disk platter, the charge trapped in a flash-memory cell, or the reflectivity of a pit on an optical disc. These media are fast and mature, but they degrade, they consume standing power even when idle (in the case of spinning or refreshed memory), and periodically need to be copied onto newer hardware as formats and devices become obsolete. As the volume of data produced worldwide continues to grow, researchers have looked toward molecular alternatives that could, in principle, hold enormous amounts of information in a very small physical volume while sitting inert for long periods without power.

DNA is chemically unusual as an information-bearing molecule because it already performs this function in biology: every cell encodes the instructions for building and operating an organism in a linear sequence of four nucleotide bases. Researchers exploring DNA data storage are not reading or writing biological genetic information; instead, they use the same four-letter chemistry to encode arbitrary digital files — text, images, audio, software — that have no biological meaning at all. This is an important distinction: biological DNA in a cell encodes genes shaped by evolution, while "DNA data storage" uses synthetic, cell-free DNA strands purely as a chemical medium for artificial digital information, similar in spirit to writing on paper rather than the content of what is written.

It is worth stating plainly at the outset that DNA data storage should currently be regarded as an emerging, actively researched technology rather than a replacement for solid-state drives, hard-disk drives, or magnetic tape. As later sections describe, real demonstrations exist — some storing hundreds of megabytes of data — but the writing and reading processes remain slow and comparatively expensive next to electronic storage, and the technology is best understood today as a candidate for long-term archival storage rather than a general-purpose replacement.

What Does "Chemically Stored Data" Actually Mean?

DNA is a polymer built from a backbone of sugar and phosphate groups, with one of four nitrogenous bases attached at each position along the strand: adenine (A), thymine (T), cytosine (C), and guanine (G). The specific order of these bases along the strand is called the nucleotide sequence, and it is this order — not any electrical or magnetic property of the molecule — that carries information, in exactly the way that the order of letters in a sentence carries meaning rather than the ink itself.

Because there are four possible bases at each position, a strand of DNA is a natural information carrier with a four-symbol alphabet, similar in principle to how a hard drive uses a two-symbol (binary) alphabet of 0s and 1s. An encoding algorithm is needed to translate between the two: it takes the binary representation of a digital file and maps it onto a corresponding sequence of A, T, C and G bases, which are then physically synthesized.

Figure 2

Figure 2. DNA provides a native four-symbol alphabet (A / T / C / G), in contrast to the binary two-symbol alphabet of conventional digital storage. The exact rule used to map bits onto bases is algorithm-dependent — there is no single universal standard.

It is important to be precise here: there is no single, universal binary-to-DNA encoding scheme used across the field. Different research groups have designed different mapping algorithms, and the choice of scheme materially affects storage density, error resistance, and synthesis compatibility. Some early schemes used straightforward base-4 or base-2 mappings; later schemes, discussed below, use more sophisticated fountain-code-based approaches that approach the theoretical information limit per nucleotide while remaining robust to loss.

The Complete DNA Data Storage Pipeline

Figure 1

Figure 1. The complete DNA data-storage lifecycle. Teal-topped cards mark the "write" half (encoding through synthesis); gold-topped cards mark the "read" half (preservation through decoding).

Every DNA data storage system, regardless of the specific encoding or synthesis technology used, follows this same basic lifecycle. Each stage introduces its own scientific and engineering challenges, discussed in turn below.

How Information Is Encoded Into DNA

The simplest conceivable scheme maps each bit or pair of bits directly onto one of the four bases — for example, associating 00, 01, 10 and 11 with A, C, G and T. In practice, naive direct mappings of this kind create sequences that are difficult or impossible to synthesize and sequence reliably. Long runs of the same base (homopolymers, such as "AAAAAAA") are prone to synthesis and sequencing errors; sequences with extreme GC content (too much or too little cytosine and guanine relative to adenine and thymine) can be chemically unstable or difficult to amplify; and certain sequences fold into secondary structures such as hairpins that interfere with both synthesis and reading.

To avoid these problems, practical encoding schemes are "constrained codes" designed to translate binary data into DNA sequences that respect biochemical limits on homopolymer length and GC content while remaining as dense as possible. One influential example is DNA Fountain, introduced by Erlich and Zielinski, which applies a probabilistic "fountain code" — a class of erasure-correcting code originally developed for reliable data transmission — to DNA storage. The authors reported that this approach approached the theoretical maximum information capacity per nucleotide while remaining robust to a substantial fraction of molecule loss, and demonstrated it by encoding a full computer operating system, a movie, and other files into DNA oligonucleotides, achieving what they described as a density on the order of hundreds of petabytes per gram under laboratory conditions (Erlich & Zielinski, 2017).

Because there is no single standardized scheme, comparisons between different DNA storage systems must account for differences in redundancy, indexing overhead, and constraint strictness — a system that reports a very high "logical density" may be trading off resilience to errors, and vice versa.

DNA Synthesis — "Writing" the Data

Once a target nucleotide sequence has been computed, it must be physically synthesized as real DNA molecules. The dominant industrial method historically has been phosphoramidite chemical synthesis, in which nucleotides are added one at a time to a growing chain anchored to a solid support, using a well-established but chemically demanding cycle of coupling, capping, oxidation, and deprotection steps. This process is highly precise but is limited in the length of DNA it can reliably produce in one continuous run (typically on the order of a couple of hundred bases), generates hazardous chemical waste, and its cost and throughput have historically been major bottlenecks for storing large amounts of data (Meiser et al., 2020).

An alternative approach under active development is enzymatic DNA synthesis, most commonly built around the enzyme terminal deoxynucleotidyl transferase (TdT), a template-independent polymerase capable of adding nucleotides to a DNA strand in aqueous, comparatively mild conditions. Because natural TdT adds bases somewhat randomly, several groups have engineered controlled variants: Palluk et al. (2018) demonstrated an approach using TdT covalently tethered to individual nucleotides so that only one base is added per reaction cycle, while Lee et al. (2019) reported a terminator-free strategy that instead encodes information in the transitions between different bases within short homopolymer runs, later extending this into a multiplexed, light-directed enzymatic synthesis platform capable of writing many sequences in parallel on an array (Lee et al., 2020). More recent work continues to engineer faster and more accurate TdT variants for commercial-scale enzymatic synthesis (see, for example, the directed-evolution study reported in Nucleic Acids Research). Enzymatic synthesis is generally seen as more scalable and more environmentally favorable than phosphoramidite chemistry because it avoids many of the hazardous organic solvents and reagents involved, though as of the mid-2020s it has not yet displaced chemical synthesis as the dominant industrial method for large commercial DNA data-storage demonstrations.

Synthesis errors — insertions, deletions, and base substitutions introduced during the writing process — are unavoidable at some rate with both chemistry classes and are one of the principal sources of error that downstream error-correction coding must address.

DNA Storage — How the Molecules Are Preserved

Figure 3

Figure 3. The simplified five-stage DNA storage lifecycle. Teal marks the initial write step; gold marks the physical-storage half; navy marks the read-back half — matching the color logic used throughout this article.

Once synthesized, DNA can be stored in several physical forms. Dried or dehydrated DNA is generally more stable than DNA kept in aqueous solution, because water promotes hydrolytic degradation reactions. Low-temperature storage further slows chemical decay. A particularly well-studied preservation strategy, reported by Grass, Heckel, Puddu, Paunescu and Stark (2015), encapsulates DNA inside inert silica (glass) spheres, physically isolating it from oxygen and moisture. In accelerated aging experiments, the authors found that DNA encapsulated in silica and combined with error-correcting codes could be recovered with zero errors even after heat treatment at 70°C for one week, a condition the authors modeled as thermally equivalent to storing the DNA in a cool European climate for roughly two thousand years (Grass et al., 2015). The same study used this method to store a digitized version of the Archimedes Palimpsest as a demonstration file.

It is important to separate three distinct kinds of longevity claim that are often blurred in popular coverage of this field:

Three kinds of "how long does DNA last?"

• Theoretical estimates: extrapolations from known decay kinetics under idealized storage conditions (e.g., cold, dry, dark, encapsulated).
• Experimentally observed stability: results from accelerated-aging experiments, such as the heat-treatment study above, that model long-term decay by exposing DNA to elevated temperatures for shorter, measurable periods.
• Practical archival lifetime: what has actually been demonstrated by recovering real, intact digital files after real elapsed storage time — which, for synthetic DNA data-storage archives, is necessarily far shorter than either of the above, since the technology itself is barely more than a decade old.

Separately, forensic and ancient-DNA research on naturally occurring biological DNA — such as the widely cited kinetic study of DNA decay in bone by Allentoft et al. (2012) — has informed general models of nucleic-acid decay chemistry, though that work concerns biological specimens rather than synthetic archival DNA data-storage systems, and the two contexts should not be conflated. Claims that "DNA lasts thousands of years" as an unqualified, universal fact are not supported by the literature; DNA stability depends heavily on the storage environment, and even encapsulated, dried DNA is subject to slow chemical decay over sufficiently long time frames.

How Data Is Read Back

Reading stored information requires DNA sequencing: determining the order of bases along each strand. Early Sanger sequencing methods, still valued for accuracy on short, well-defined fragments, have largely given way in DNA data-storage research to high-throughput short-read platforms (such as Illumina sequencing-by-synthesis) and to long-read platforms, most notably nanopore sequencing, which reads a strand as it passes through a protein nanopore and can process much longer molecules in real time. Organick et al. (2018) used Illumina-based sequencing to retrieve individually addressed files from a pool of more than 13 million DNA oligonucleotides, while more recent enzymatic-synthesis demonstrations have paired their writing methods with streaming nanopore sequencing for readout (Lee et al., 2019).

Sequencing is not perfect: individual reads contain errors, and not every molecule in a pool is necessarily captured by a given sequencing run, a phenomenon known as molecular dropout. Because many identical or near-identical copies of each encoded strand are typically present in a synthesized pool, computational reconstruction usually combines multiple reads per underlying strand (a process called consensus or clustering) to reduce the effect of any one read's errors before decoding proceeds.

Error Correction

Figure 4

Figure 4. Errors can be introduced at four main points in the pipeline. Redundancy and error-correcting codes (Reed–Solomon–style codes, fountain codes) reconcile these errors computationally during decoding rather than preventing them outright.

Errors can be introduced at every stage of the pipeline: during synthesis (base insertions, deletions and substitutions), during storage (chemical degradation of strands or backbone breakage), during amplification by PCR (which can introduce or amplify bias between different sequences), and during sequencing (misreads, particularly around long homopolymer runs). Because of this, essentially every serious DNA data-storage demonstration incorporates redundancy and error-correcting coding rather than storing the raw encoded sequence alone.

Different systems have used different coding strategies. Goldman et al. (2013) used an encoding scheme with substantial built-in redundancy — encoding each byte of information across multiple overlapping short DNA segments — to protect against synthesis and sequencing errors, and successfully recovered a stored archive containing text, image and audio files, including a compressed copy of all 154 Shakespeare sonnets, an MP3 excerpt of Martin Luther King's "I Have a Dream" speech, and a photograph, after having the DNA synthesized in the United States, shipped by courier to Germany, and sequenced there. Grass et al. (2015) combined physical (silica) protection with Reed–Solomon-style error-correcting codes. Erlich and Zielinski's DNA Fountain (2017) built error resilience directly into its fountain-coding scheme, reporting perfect retrieval even when a meaningful fraction of the synthesized oligonucleotide pool was not sequenced at all.

There is an inherent trade-off in all of these approaches: adding more redundancy increases the reliability of retrieval but reduces the effective information density achieved, since a larger share of the synthesized nucleotides is devoted to error protection rather than payload data. Practical system designs balance this trade-off against the cost of synthesis and the acceptable retrieval error rate for the application.

Random Access

A file stored on a hard drive can be located and read without touching every other file on the disk. Early DNA data-storage demonstrations lacked this property: recovering any part of the archive typically required sequencing the entire pool of DNA, which becomes impractical as the number of stored files grows, since sequencing cost and time scale with the amount of DNA processed rather than with the size of the file actually needed.

Organick et al. (2018) addressed this directly, encoding 35 distinct files totaling more than 200 megabytes of data into more than 13 million DNA oligonucleotides and demonstrating that any individual file could be recovered on its own, without error, using a large validated library of PCR primers that selectively amplify only the oligonucleotides belonging to a chosen file — an approach generally described as PCR-based random access with sequence-specific primers. More recent research has continued to explore alternative random-access mechanisms, including CRISPR-Cas9-based selective retrieval and similarity search strategies for DNA archives (see, for example, the Cas9-based random access and semantic search work published in Nature Communications). Random access of this kind is considered essential for any DNA storage system intended to function as a practical, queryable archive rather than a single monolithic, all-or-nothing backup.

How Much Data Can DNA Store?

Discussions of DNA storage density often cite theoretical estimates on the order of hundreds of exabytes to roughly an exabyte-per-gram scale, derived from the physical size of a nucleotide and the four-symbol information capacity of each base position. Erlich and Zielinski (2017) reported an experimentally demonstrated density of 215 petabytes per gram of DNA using their DNA Fountain architecture — itself far short of the absolute theoretical maximum, but, as the authors noted, orders of magnitude higher than densities achieved by earlier DNA storage demonstrations.

These figures require several caveats. First, theoretical maximum density assumes every nucleotide carries payload information with no redundancy, indexing, or primer-binding sequence overhead — conditions no practical system meets, since real systems must reserve part of every strand for addressing information and error correction. Second, published density figures often describe the density of the synthesized DNA payload itself and do not necessarily include the container, buffer, or encapsulation matrix (such as the silica spheres used by Grass et al., 2015) in which the DNA is physically stored and handled. Third, achieving high density in a demonstration experiment is different from achieving it economically and reliably at commercial archival scale, where synthesis cost per nucleotide remains the dominant practical constraint, not the physical information limit of the DNA molecule itself.

DNA vs. Conventional Storage

Figure 5

Figure 5. Qualitative positioning of DNA against HDD, SSD and magnetic tape across five properties. Dot count reflects the authors' qualitative synthesis of the cited literature (see the table below and References), not a standardized numeric benchmark.

Property	DNA	HDD	SSD	Magnetic Tape
Density	Very high in principle; up to ~215 PB/g demonstrated experimentally (Erlich & Zielinski, 2017)	Moderate; areal-density limited	Moderate–high; cell-density limited	Moderate; among the densest conventional archival media
Longevity (idle, unpowered)	Potentially very long under controlled/encapsulated conditions (Grass et al., 2015); requires further real-world validation	Years; mechanical/magnetic decay	Years; charge leakage over time	Multiple decades under proper storage; periodic migration still recommended
Energy for passive preservation	Very low once written and stored appropriately	Ongoing power if spinning/kept online	Low when powered off, but periodic refresh often needed	Very low; no power needed while shelved
Writing speed	Slow; synthesis-limited (chemical or enzymatic)	Fast	Very fast	Fast (sequential)
Reading speed	Slow; sequencing-limited	Fast	Very fast	Fast (sequential)
Random access	Emerging; demonstrated via PCR-based primer addressing (Organick et al., 2018)	Native, mature	Native, very fast	Poor; largely sequential
Rewritability	Limited; mostly write-once in current demonstrations	Fully rewritable	Fully rewritable (with wear limits)	Rewritable (sequentially)
Cost (current)	High; synthesis and sequencing costs dominate	Low per terabyte	Moderate–low per terabyte	Very low per terabyte
Technological maturity	Early-stage / research and pilot demonstrations	Mature, commercial	Mature, commercial	Mature, commercial
Best current use case	Long-horizon, infrequently accessed archival research demonstrations	General-purpose, active storage	High-performance active storage	Cold/archival backup at scale

Note: Quantitative comparisons for DNA reflect figures reported in the cited primary studies rather than standardized, audited benchmarks; HDD/SSD/tape figures are described qualitatively here rather than with invented numbers, consistent with the source-integrity requirements of this review.

Real-World Demonstrations

Year	Researchers	What Was Stored	Method	Significance
2012	Church, Gao & Kosuri	A 5.27-megabit book (text and images) encoded on DNA microchips	Direct binary-to-base mapping; DNA microarray synthesis; next-generation sequencing readout	Among the first demonstrations of storing a substantial arbitrary digital file (not just a short message) in DNA (Church et al., 2012)
2013	Goldman, Bertone, Chen, Dessimoz, LeProust, Sipos & Birney	Text (Shakespeare's sonnets), audio (an MLK speech excerpt), and a photograph	Redundant, overlapping-segment encoding with robust error correction; DNA synthesized in the US and sequenced after shipping to the UK/Germany	Demonstrated multi-format storage and error-free recovery after physical transport, with an early scalability analysis (Goldman et al., 2013)
2015	Grass, Heckel, Puddu, Paunescu & Stark	The digitized Archimedes Palimpsest text	Silica encapsulation of DNA plus Reed–Solomon-style error-correcting codes; accelerated thermal aging tests	Directly addressed long-term chemical preservation, not just encoding density (Grass et al., 2015)
2017	Erlich & Zielinski	A full computer operating system, a movie, and other files (~2.14 million bytes total)	"DNA Fountain" probabilistic fountain coding, approaching the theoretical information capacity per nucleotide	Demonstrated near-capacity information density with error-free retrieval and resilience to partial data loss (Erlich & Zielinski, 2017)
2018	Organick, Ang, Chen, Lopez, Yekhanin, Makarychev et al.	35 distinct files, over 200 MB of data, in more than 13 million oligonucleotides	PCR primer-based random access enabling individual file recovery without sequencing the entire pool	Largest scale (by data volume and file count) random-access DNA storage demonstration at the time (Organick et al., 2018)
2020	Koch, Gantenbein, Masania, Stark, Erlich & Grass	Digital files embedded directly into a physical, 3D-printed object ("DNA-of-things")	DNA encapsulated in silica nanoparticles mixed into a printable polymer matrix	Extended DNA data storage from standalone archives toward information embedded within everyday materials (Koch et al., 2020)

Current Research Directions

Active research areas reported in the recent literature include faster and cheaper enzymatic DNA synthesis (Palluk et al., 2018; Lee et al., 2019, 2020), improved constrained and fountain-style coding schemes that push closer to the theoretical channel capacity of the DNA storage medium, alternative and more efficient random-access mechanisms including CRISPR-Cas9-based retrieval, molecular indexing and similarity search across DNA archives, and automated, integrated end-to-end write/store/read platforms (Takahashi, Nguyen, Strauss & Ceze, described in the DNA data storage automation literature). Researchers have also explored non-synthesis-based writing strategies, such as enzymatic methylation "movable-type" printing onto pre-made DNA templates, which the authors argue could reduce the cost bottleneck associated with de novo synthesis of every archived file (Nature, 2024/2025 movable-type printing study). Alternative substrate engineering, such as adsorbing DNA onto highly porous cellulose-acetate particles to improve handling density and durability, has also been reported (Lv et al., 2024, Nature Nanotechnology). Broader reviews continue to track the field's overall trajectory, security considerations, and application space, including the comprehensive review by Ceze, Nivala and Strauss (2019), the more recent chemistry-focused review by Yang, Bögels, Wang and colleagues (2024), and application-oriented surveys such as Meiser, Nguyen, Chen, Nivala, Strauss, Ceze and Grass (2022).

Could DNA Become an Information-Carrying Forensic Material?

Because this review is written for a forensic-science audience, it is worth separating what is established from what is speculative in this specific area. There is no evidence in the peer-reviewed literature reviewed for this article that forensic laboratories currently use synthetic, information-bearing DNA as a routine digital storage medium in casework. The connections between DNA data storage and forensic science that are supported by the literature are conceptual and largely forward-looking rather than operational:

• Molecular tagging and authentication: synthetic DNA sequences can, in principle, be used as covert molecular tags or barcodes embedded in physical materials — an idea closely related to the "DNA-of-things" demonstration by Koch et al. (2020), in which information-bearing DNA was embedded directly into a manufactured object.
• Provenance and anti-counterfeiting: because synthetic DNA sequences can be made to encode arbitrary identifiers, researchers have discussed molecular barcoding as a conceptual tool for tracking the origin or authenticity of products or documents, though widescale forensic deployment of this idea is, at the time of writing, still experimental.
• Distinguishing biological from synthetic DNA: any forensic application of information-bearing synthetic DNA would need robust methods to distinguish it from an organism's own biological DNA at a crime scene, and to establish clear chain-of-custody practices for molecular evidence that may itself contain embedded digital information — an area that, as far as the sources reviewed here indicate, remains largely unaddressed in published forensic protocols.

In short: the underlying chemistry that makes DNA data storage possible is the same chemistry forensic scientists already work with every day, and researchers have begun to sketch conceptual links between molecular information storage and questions of provenance, tagging and authentication. However, readers should not infer that forensic laboratories currently use DNA as a digital storage medium in routine practice; the literature at this time supports discussion of the concept, not a claim of established forensic use.

Security and Privacy

As with any storage medium, researchers have begun to consider security implications specific to DNA data storage. Because DNA sequencing technology has become widely available outside specialized laboratories, unauthorized sequencing of a physical DNA sample is a plausible route by which stored information could be read without the intended recipient's involvement — a concern conceptually similar to physical media theft, but with the added wrinkle that a very small physical sample can, in principle, carry a large quantity of information. Researchers have discussed molecular-level access-control concepts, including sequence-level encryption approaches applied before encoding and steganographic techniques that hide payload data within larger, biologically plausible-looking sequences. As with the forensic applications discussed above, most published security-oriented work in this space remains at the research or proof-of-concept stage rather than reflecting deployed, audited security infrastructure, and readers should treat specific security claims in this area with appropriate caution pending further peer-reviewed validation.

Environmental and Economic Questions

DNA data storage should not be assumed to be "green" or "eco-friendly" without qualification. Traditional phosphoramidite chemical synthesis consumes organic solvents and reagents and generates chemical waste that must be handled and disposed of appropriately; this is one of the stated motivations behind interest in enzymatic synthesis alternatives, which generally use milder, aqueous reaction conditions (Meiser et al., 2020; Ashley et al., 2023). Sequencing likewise consumes reagents and instrument energy, and both synthesis and sequencing currently require specialized laboratory equipment and skilled operation, which carries its own economic and infrastructure cost. Set against this is the fact that once written and properly preserved, DNA in a dried or encapsulated state requires very little ongoing energy to simply sit in storage — unlike, for example, actively spinning hard drives or memory that must be periodically refreshed. A fair environmental comparison, then, needs to weigh a currently high energy and chemical cost for writing and reading DNA against a potentially very low energy cost for passive, long-duration preservation — and the literature reviewed here does not support a blanket claim that DNA storage is currently more environmentally favorable than conventional media once the full write/read lifecycle is included, particularly at today's synthesis costs and throughput.

Major Limitations: Why Aren't We Storing Everything in DNA Already?

Candid barriers to adoption

• Cost: synthesizing and sequencing DNA remains far more expensive per byte than conventional storage media.
• Slow writing and reading: both chemical and enzymatic synthesis, and both short- and long-read sequencing, are orders of magnitude slower than electronic read/write operations.
• Synthesis throughput: scaling synthesis to archive-relevant volumes of data requires substantial parallelization that is still maturing.
• Sequencing requirements and error rates: synthesis, storage, amplification and sequencing all introduce errors that must be corrected computationally, adding overhead.
• Random access is still an emerging capability rather than a mature, off-the-shelf feature of DNA storage systems.
• Limited rewritability: most demonstrated systems are effectively write-once; efficiently updating or deleting stored data in DNA remains an active research problem.
• Automation and infrastructure: end-to-end automated write/store/retrieve/read systems exist as research prototypes but are not yet commercially deployed at scale.
• Chemical waste and specialized handling add operational complexity relative to electronic media.
• Retrieval latency: even successful random-access retrieval currently takes hours, not the milliseconds typical of electronic storage.

What Is Realistic in 2026?

Category	Status
Already demonstrated	Encoding and error-free recovery of text, images, audio, video and software files from synthetic DNA (Church et al., 2012; Goldman et al., 2013; Erlich & Zielinski, 2017); PCR-based random access at multi-file, multi-megabyte scale (Organick et al., 2018); long-term chemical preservation strategies validated by accelerated aging (Grass et al., 2015); enzymatic synthesis of DNA for data storage in laboratory settings (Palluk et al., 2018; Lee et al., 2019, 2020)
Experimental / research-stage	Automated end-to-end write-to-read pipelines; CRISPR-based and other advanced random-access and similarity-search mechanisms; DNA embedded in manufactured physical objects (Koch et al., 2020); non-synthesis "movable-type" writing strategies
Technically plausible but not yet practical	Commercial-scale archival deployment competitive with tape on cost per byte; efficient in-place rewriting of stored DNA archives; routine forensic or supply-chain deployment of information-bearing synthetic DNA tags
Speculative	DNA storage as a general-purpose replacement for consumer or enterprise electronic storage; DNA-based information systems as a mainstream forensic evidentiary category

Future Outlook

Figure 6

Figure 6. A conceptual, fully automated future DNA archival system. Dashed borders and the routing note in the caption above deliberately mark this as an integrated research direction rather than a system that exists as shown today.

The published literature converges on a consistent set of priorities for the field: cheaper and higher-throughput synthesis (very likely centered on continued enzymatic-synthesis engineering); faster, cheaper and more portable sequencing; error-correction and coding schemes that make more efficient use of the underlying information capacity of DNA; standardized file formats and indexing conventions so that archives built by different groups can interoperate; robust, well-validated random-access mechanisms; and automated, robotic end-to-end systems that reduce the amount of specialized manual laboratory work required to write and read a DNA archive. Progress on any one of these fronts tends to make the others more valuable, which is part of why the field has diversified quickly since Church et al.'s (2012) and Goldman et al.'s (2013) early demonstrations.

Myth vs. Reality

Myth	Reality
DNA storage is already replacing hard drives	DNA storage remains a research- and pilot-stage technology; conventional media remain dominant for essentially all active storage use cases (see "What Is Realistic in 2026" above).
DNA can store unlimited information	DNA has very high theoretical information density, but any real archive is bounded by the amount of DNA synthesized, sequencing capacity, and indexing/error-correction overhead (Erlich & Zielinski, 2017).
DNA storage requires living cells	Most DNA data-storage demonstrations use cell-free, synthetic DNA synthesized and handled entirely outside living organisms.
DNA lasts forever	DNA decays chemically over time; longevity depends heavily on storage conditions, and even encapsulated DNA has finite, condition-dependent stability (Grass et al., 2015).
DNA storage is automatically environmentally friendly	Current synthesis and sequencing processes consume chemicals, reagents and energy; passive storage is low-energy, but the full write/read lifecycle carries real environmental costs.
DNA data can be retrieved instantly	Retrieval requires sequencing and computational decoding, which currently take substantially longer than reading electronic storage (Organick et al., 2018).

Glossary

DNA data storage

The use of synthetic DNA molecules as a physical medium for encoding, storing and retrieving arbitrary digital information.

Nucleotide

The basic building block of DNA, consisting of a sugar, a phosphate group, and one of four nitrogenous bases (A, T, C or G).

Nucleotide sequence

The specific linear order of bases along a DNA strand, which carries the encoded information.

DNA synthesis

The chemical or enzymatic process of building a DNA strand with a specified nucleotide sequence — the "writing" step.

Sequencing

The process of determining the order of nucleotides in a DNA sample — the "reading" step.

Encoding

The algorithmic translation of digital (typically binary) data into a corresponding DNA nucleotide sequence.

Decoding

The algorithmic translation of a sequenced DNA readout back into the original digital file.

Random access

The ability to retrieve a specific file or subset of data from a DNA archive without processing the entire archive.

Error correction

Coding strategies (e.g., redundancy, fountain codes, Reed–Solomon-style codes) that allow the original data to be recovered despite errors introduced during synthesis, storage, amplification or sequencing.

Homopolymer

A run of identical, consecutive nucleotide bases (e.g., "AAAA"), which is prone to synthesis and sequencing errors.

GC content

The proportion of guanine and cytosine bases in a DNA sequence, which affects chemical stability and is a key constraint in encoding-scheme design.

Molecular indexing

The use of short DNA "address" sequences (e.g., primer-binding sites) to label and later selectively retrieve specific stored files.

Archival storage

Storage intended for long-term, infrequently accessed data, as opposed to actively used, frequently rewritten storage.

Enzymatic DNA synthesis

DNA synthesis carried out using enzymes (most commonly terminal deoxynucleotidyl transferase) rather than solely conventional phosphoramidite chemistry.

Final Conclusion

Can data be chemically stored in DNA? Yes — this is not a hypothetical. Digital information can be, and repeatedly has been, chemically represented as a nucleotide sequence, physically synthesized into real DNA molecules, preserved, sequenced, and computationally decoded back into the original file, with essentially perfect fidelity when appropriate error-correction coding is used. This has been demonstrated across text, images, audio, video, and software, at scales reaching hundreds of megabytes with individually retrievable files.

What the evidence does not support is the idea that DNA storage is currently a practical replacement for conventional digital storage. Writing and reading DNA remain slow and expensive relative to electronic media; random access, rewritability and automation are still maturing; and the theoretical density figures often quoted in popular coverage substantially overstate what current systems achieve once real-world overhead is included. DNA storage is, at present, best understood as a promising candidate specifically for archival, "cold" storage of very large volumes of data that are written once and read rarely — precisely the use case where its combination of extreme density and low passive-storage energy needs matters most, and where its current weaknesses in speed and cost matter least. Whether it eventually expands beyond that niche will depend on the kind of incremental, well-documented progress in synthesis, sequencing, coding and automation that the research summarized in this review continues to report.

---

References

Allentoft, M. E., Collins, M., Harker, D., Haile, J., Oskam, C. L., Hale, M. L., Campos, P. F., Samaniego, J. A., Gilbert, M. T. P., Willerslev, E., et al. (2012). The half-life of DNA in bone: Measuring decay kinetics in 158 dated fossils. Proceedings of the Royal Society B: Biological Sciences, 279(1748), 4724–4733. https://doi.org/10.1098/rspb.2012.1745

Ceze, L., Nivala, J., & Strauss, K. (2019). Molecular digital data storage using DNA. Nature Reviews Genetics, 20, 456–466. https://doi.org/10.1038/s41576-019-0125-3

Church, G. M., Gao, Y., & Kosuri, S. (2012). Next-generation digital information storage in DNA. Science, 337(6102), 1628. https://doi.org/10.1126/science.1226355

Erlich, Y., & Zielinski, D. (2017). DNA Fountain enables a robust and efficient storage architecture. Science, 355(6328), 950–954. https://doi.org/10.1126/science.aaj2038

Goldman, N., Bertone, P., Chen, S., Dessimoz, C., LeProust, E. M., Sipos, B., & Birney, E. (2013). Towards practical, high-capacity, low-maintenance information storage in synthesized DNA. Nature, 494(7435), 77–80. https://doi.org/10.1038/nature11875

Grass, R. N., Heckel, R., Puddu, M., Paunescu, D., & Stark, W. J. (2015). Robust chemical preservation of digital information on DNA in silica with error-correcting codes. Angewandte Chemie International Edition, 54(8), 2552–2555. https://doi.org/10.1002/anie.201411378

Koch, J., Gantenbein, S., Masania, K., Stark, W. J., Erlich, Y., & Grass, R. N. (2020). A DNA-of-things storage architecture to create materials with embedded memory. Nature Biotechnology, 38, 39–43. https://doi.org/10.1038/s41587-019-0356-z

Lee, H. H., Kalhor, R., Goela, N., Bolot, J., & Church, G. M. (2019). Terminator-free template-independent enzymatic DNA synthesis for digital information storage. Nature Communications, 10, 2383. https://doi.org/10.1038/s41467-019-10258-1

Lee, H., et al. (2020). Photon-directed multiplexed enzymatic DNA synthesis for molecular digital data storage. Nature Communications, 11, 5246. https://doi.org/10.1038/s41467-020-18681-5

Meiser, L. C., Antkowiak, P. L., Koch, J., Chen, W. D., Kohll, A. X., Stark, W. J., Heckel, R., & Grass, R. N. (2020). Reading and writing digital data in DNA. Nature Protocols, 15(1), 86–101. https://doi.org/10.1038/s41596-019-0244-5

Meiser, L. C., Nguyen, B. H., Chen, Y.-J., Nivala, J., Strauss, K., Ceze, L., & Grass, R. N. (2022). Synthetic DNA applications in information technology. Nature Communications, 13, 352. https://doi.org/10.1038/s41467-021-27846-9

Organick, L., Ang, S. D., Chen, Y.-J., Lopez, R., Yekhanin, S., Makarychev, K., Racz, M. Z., Kamath, G., Gopalan, P., Nguyen, B., Takahashi, C. N., Newman, S., Parker, H.-Y., Rashtchian, C., Stewart, K., Gupta, G., Carlson, R., Mulligan, J., Carmean, D., Seelig, G., Ceze, L., & Strauss, K. (2018). Random access in large-scale DNA data storage. Nature Biotechnology, 36(3), 242–248. https://doi.org/10.1038/nbt.4079

Palluk, S., Arlow, D. H., de Rond, T., Barthel, S., Kang, J. S., Bector, R., Baghdassarian, H. M., Truong, A. N., Kim, P. W., Singh, A. K., et al. (2018). De novo DNA synthesis using polymerase-nucleotide conjugates. Nature Biotechnology, 36, 645–650. https://doi.org/10.1038/nbt.4173

Yang, S., Bögels, B. W. A., Wang, F., et al. (2024). DNA as a universal chemical substrate for computing and data storage. Nature Reviews Chemistry, 8, 179–194. https://doi.org/10.1038/s41570-024-00576-4