How does synthetic DNA store exabytes of data per gram?

Synthetic DNA stores archival binary data by translating digital ones and zeros into sequences of the four nucleotide bases: adenine, cytosine, guanine, and thymine. Because each molecule is only nanometers wide and can be packed in three dimensions rather than spread across a two-dimensional surface, DNA achieves a theoretical information density of hundreds of exabytes per gram. This makes it a promising medium for ultra-dense cold storage in classical computing, capable of preserving massive datasets for centuries without electrical power.

Encoding Binary Data into Nucleotides

Classical computers represent information using binary digits (\(0\) and \(1\)). In DNA data storage, an encoding algorithm maps these bits onto the four chemical bases: adenine (\(\text{A}\)), cytosine (\(\text{C}\)), guanine (\(\text{G}\)), and thymine (\(\text{T}\)). A straightforward mapping assigns two bits per base, such as:

\[00 \rightarrow \text{A}, \quad 01 \rightarrow \text{C}, \quad 10 \rightarrow \text{G}, \quad 11 \rightarrow \text{T}\]

In practice, raw two-bit mapping can produce long repeating runs of the same base (homopolymers) or unbalanced ratios of \(\text{G}\) and \(\text{C}\) bases, which cause errors during chemical synthesis and sequencing. To prevent this, modern encoders use constrained coding schemes—such as DNA Fountain or rotating base codes—that avoid repeats and maintain a balanced \(\text{GC}\) content. Error-correcting codes, such as Reed-Solomon or low-density parity-check (LDPC) codes, are also added so that lost or mutated strands can be mathematically reconstructed.

Why DNA Achieves Exabyte-per-Gram Density

Conventional magnetic tape and hard disk drives store bits on a flat, two-dimensional surface, where each bit occupies tens of nanometers in width and length. By contrast, a DNA double helix has a diameter of roughly \(2\text{ nm}\), and adjacent base pairs are spaced only \(0.34\text{ nm}\) apart along the strand.

When accounting for the molecular weight of a base pair (approximately \(650\text{ Da}\), or \(1.08 \times 10^{-21}\text{ g}\)) and a theoretical maximum of \(2\text{ bits}\) per base, the raw physical limit approaches \(455\text{ EB/g}\) (\(4.55 \times 10^{20}\text{ bytes/g}\)). Even after factoring in error-correction overhead, indexing primers, and encapsulation, practical demonstrations and models still reach hundreds of petabytes to several exabytes per gram—orders of magnitude denser than any silicon or magnetic medium.

Writing, Storing, and Reading the Data

The lifecycle of DNA archival storage follows three main stages:

  1. Synthesis (Writing): Short DNA strands, typically \(100\) to \(300\) nucleotides long, are chemically synthesized base by base using phosphoramidite chemistry or enzymatic synthesis. Each strand includes a short address sequence (primer) so specific files can be identified later.
  2. Preservation (Storage): The synthesized DNA pool is dried or encapsulated in silica nanoparticles and kept in a cool, dark, and dry environment. Under these conditions, DNA remains stable for hundreds to thousands of years with zero active power consumption.
  3. Sequencing (Reading): To retrieve a file, polymerase chain reaction (PCR) selectively amplifies only the strands carrying the target file's primer. A DNA sequencer then reads the base sequences, and a decoder translates the \(\text{A}\), \(\text{C}\), \(\text{G}\), and \(\text{T}\) characters back into binary files.

Role in Classical Computing

Synthetic DNA is not intended to replace fast primary memory like DRAM or solid-state drives. Chemical synthesis and sequencing currently have high latency and high cost per megabyte, making DNA unsuitable for frequently accessed data. Instead, it targets the deepest tier of classical archival storage: cold data such as scientific datasets, national archives, and long-term backups that are written once, stored for decades, and rarely read.