A patent from Seagate Technology, granted in May 2024, describes a method for storing data on a flat sheet of metal foil. The sheet is treated to carry a positive surface charge. A layer of synthetic DNA is sprayed onto it, then encapsulated in a thin coating of silica or gold. To read back a specific file, the user pipes a buffer through a microfluidic channel to dissolve the coating at a chosen coordinate, releases the DNA with a polyanion, and sequences it.

If you squint, this is a hard drive. Addressable region, protective overcoat, head that selectively accesses one location, the rest of the platter undisturbed. The active medium just happens to be a polymer of adenine, thymine, guanine, and cytosine instead of cobalt-chromium-platinum.

Seagate is a hard-drive company. It owns no genome facility, runs no clinical lab, sells no reagents. And yet a search of US patent grants from 2018 onward turns up four DNA-data-storage patents assigned to Seagate Technology LLC, three to Western Digital Technologies, three to EMC IP Holding (the patent vehicle for Dell’s storage business), and five to Microsoft Technology Licensing. The four largest patent-holders in DNA data storage are not biotech companies. They are the storage industry.

That fact, by itself, is not a story. The patents themselves are.

What the storage companies are actually patenting

Read the four Seagate filings together and a thesis appears. US 11,066,661 (2021) and the continuation US 11,643,647 (2023) describe a “DNA symbol library” — a stockroom of pre-fabricated short oligonucleotides, each with two non-complementary “overhang” ends that act as connectors. To write a long DNA strand encoding a payload, you don’t synthesize it base by base. You pick symbols off the shelf and ligate them like Lego. US 11,990,184 (2024) takes the next step: Cas9 enzymes and guide RNAs are shuttled around a “hydrophobic fluidic platform comprising multiple cells each configured to independently receive a voltage.” Voltage chops the backbone. More voltage moves the next symbol into place. A repair enzyme stitches.

What Seagate has described is a chip — addressable cells, electronic actuation, parallel write — whose write head happens to be a programmable CRISPR complex and whose recording medium is a polymer.

Western Digital, meanwhile, has spent its three grants on something different and more revealing: error-correction codes. US 12,355,468 (2025) splits a long DNA strand into shorter sub-strands, each with its own parity bits, “separately decodable from the other short DNA strands.” US 12,373,283 introduces “syndrome weight” thresholds and indel-aware decoders. US 12,430,202 nests LDPC-style codes inside one another. These are not biology patents. These are the same family of techniques that keeps your laptop SSD from corrupting under cosmic-ray bit flips, ported wholesale to a substrate whose dominant error modes are insertion and deletion errors during synthesis — not bit flips.

Microsoft’s filings split the difference. US 10,793,852 covers drying DNA in calcium- or lanthanum-salt films to push density past 30% DNA by weight. US 12,430,567 (2025) describes “multiplex similarity search,” where a query strand and a result strand are physically linked inside a tube so that a single sequencing run can return matches for many queries at once — a content-addressable memory architecture, encoded in chemistry. EMC’s three grants are even more on-the-nose, including US 11,106,633, “DNA-based data center with deduplication capability.” The storage-industry obsession with dedup, applied to a fluid archive.

This is the coherence test that data-driven patent stories usually fail. Strip the phrase “DNA storage” from these filings and you are still looking at the storage engineer’s toolkit: error-correction codes, addressable substrates, deduplication, content-addressable lookup, voltage-actuated write heads. The patents share an engineering DNA, in the metaphorical sense, that has nothing to do with the literal kind.

Where the money is moving

The literature has been quietly catching up to the patents. A search of OpenAlex returns 17 papers on DNA data storage published in 2018. In 2025 the count was 186. The first five months of 2026 are already at 66, on track to nearly double again.

The capital is following. On May 5, 2025, Twist Bioscience — the silicon-photolithography-derived oligo synthesis company that anchors the DNA Data Storage Alliance — spun out its data-storage program into a standalone company called Atlas Data Storage, seeded with \$155 million from ARCH Venture Partners, Deerfield, Bezos Expeditions, In-Q-Tel, and the venture firm Tao Capital, according to a Twist investor release covered by Blocks & Files. Atlas’s chief executive is Varun Mehta, who previously co-founded Nimble Storage, an array vendor acquired by Hewlett Packard Enterprise for \$1.2 billion. Atlas’s chief technology officer is Bill Banyai, the co-founder of Twist whose background is in photolithography for semiconductor mask-writing. A storage-array entrepreneur is running a DNA company; a semiconductor-process engineer is its CTO.

The same pattern shows up in the smaller players. Iridia, a Carlsbad startup building a CMOS chip that writes DNA polymers via electronic switches one molecule at a time, raised its 2021 Series B from a syndicate that included Western Digital’s venture arm, per HPCwire. Catalog Technologies — founded in 2016 by two MIT researchers and the most public-facing DNA storage startup — announced in September 2022 that it was integrating Seagate’s lab-on-a-chip microfluidics into its Shannon platform, aiming for what Catalog described as a 1,000-fold volume reduction. Ed Gage, then a VP at Seagate Research, told Blocks & Files in 2022 that DNA “promises to have the ability to store terabytes of data in very small amounts of fluid.” That is a statement an areal-density engineer makes when he is staring down the end of his curve.

Why now

The conventional reading is that hyperscalers need cold storage and silicon-based archive media are starting to embarrass themselves. IDC pegs the global datasphere at 181 zettabytes in 2025 and 527.5 ZB by 2029, and JLL forecasts hyperscale IT load expanding from 24.4 GW to 147 GW over the next decade. SSDs cost roughly five to ten times more per terabyte than HDDs. HDDs cost roughly five to ten times more per terabyte than tape. And the entire stack carries a recurring power and refresh bill — magnetic media stored cold still needs to be rewritten every five to seven years to stop bit rot.

DNA, by contrast, sits in a vial. A 2023 Nature Nanotechnology study from ETH Zürich’s Robert Grass demonstrated thermoresponsive microcapsules that let researchers retrieve specific files from a DNA archive without burning the whole pool — random access, in a fluid. Synthetic DNA stored in silica is stable on geological timescales. The write economics are still bad — Advanced Science in 2025 reported a record-low \$122 per megabyte using a “DNA movable type” assembly approach, roughly eight orders of magnitude off the IARPA Molecular Information Storage program’s archival target — but the storage industry is not in the habit of waiting until the cost curve bends. It is in the habit of bending the cost curve, and you do that by buying optionality early in the form of patents.

What to watch

The single most useful number to keep an eye on is not the cost-per-megabyte. It is whether any of the three big storage hardware vendors makes a public product announcement before the end of 2027. Patents are a leading indicator; an SKU is a trailing one. The patents are now four to seven years deep. Atlas Data Storage exists. Catalog has Seagate’s microfluidics. Iridia has Western Digital’s money. The Twist–Microsoft–WD–Illumina alliance has been publishing interoperability specs for almost five years.

What is striking about Seagate’s foil-substrate patent — the one I led with — is how thoroughly it has been translated out of biology and into the dialect of disk drive engineering. Flat surface. Functionalization. Protective overcoat. Selective release at an addressable coordinate. The chemistry is unfamiliar; the conceptual model is the conceptual model of a 1985 ST-506. The next archive tier in the cloud is going to look, in patent form, a lot more like the last one than the press coverage of synthetic biology would suggest.

Method note. Patent counts come from a snapshot of US utility grants drawn from USPTO bulk grant XML through May 2026, restricted to records whose title or abstract matches “DNA data storage,” “DNA-based data storage,” or “DNA storage.” Each company’s total combines variant spellings and assignment vehicles (EMC IP Holding entries count toward Dell). Literature counts are full-text matches against OpenAlex’s English-language publications. Funding figures, the Atlas spinout, the Catalog–Seagate partnership, and the Gage quote are sourced from Blocks & Files, TechCrunch, HPCwire, Nature Nanotechnology, and Advanced Science as cited inline. The IDC datasphere and JLL hyperscale-load figures are from IDC’s 2025 Global DataSphere Forecast and JLL’s 2025 Data Center Outlook.