What New Data Storage Technologies Can You Actually Use?

What is the new storage technology people are talking about?

There is no single new storage technology: glass, ceramic, and DNA are distinct approaches designed mainly for data that can be stored for a long time without frequent access. Their shared pitch is durability and potential density; the trade is specialized equipment and slower, more involved data access.

Glass systems use ultrafast lasers to make physical modifications inside glass, then use optical microscopy to read them. Microsoft’s Project Silica describes quartz glass as write-once, read-many archival media. Ceramic Nano Memory, described in a Cerabyte white paper, uses a ceramic nanolayer on thin glass sheets, with laser ablation for writing and optical reading. DNA storage encodes digital information into DNA sequences, synthesizes and preserves the strands, then retrieves, sequences, and decodes them.

That difference matters when judging maturity. A working prototype demonstrates that a process can be done; it does not establish that a complete, affordable product is available. The practical comparison is not “which one holds the most?” but “which one can store the data, retrieve it when needed, and fit the buyer’s workflow?”

Can glass or quartz really store data for thousands or millions of years?

Glass can plausibly preserve its written structures over very long periods, but the reported lifetimes are estimates for the medium—not a guarantee that every reader, robot, or storage system lasts just as long.

Microsoft’s Project Silica page describes intended lifetimes of tens to hundreds of thousands of years. Its 2023 SOSP paper reports lifetimes over 1,000 years for the glass structures used to store data. These are claims made in different project materials; they should not be collapsed into a promise of a particular service life for a commercial product.

A 2025 study in Nature demonstrated glass archival storage and used accelerated-aging tests to estimate written data stability exceeding 10,000 years at room temperature. The paper also notes that its estimate does not account for outside factors such as mechanical stress or chemical corrosion. In other words, “long-lived medium” is not the same as “hands-off system that never needs care.”

For an archive, the full chain still matters: the media must survive, compatible equipment must remain available, and the data must be readable and interpretable. These technologies target long-term preservation, but the evidence does not support treating a lifetime estimate as a universal guarantee.

Is DNA data storage available to buy, and how does it work?

DNA storage is not established in these sources as a broadly available consumer product; it is a research and product-development effort aimed at archival use. The process is real, but the workflow is far from saving a file to a drive.

First, software encodes digital data into DNA sequences. Those sequences are synthesized into DNA strands and preserved. To retrieve information, a system selects the relevant strands, sequences them, corrects errors, and decodes the resulting data. Each step adds equipment, time, and cost.

A review in Biomedical Engineering Letters gives a theoretical density estimate of 2 × 10²⁴ bits per kilogram. That number describes the medium’s potential, not a practical storage product’s capacity or speed. The review identifies DNA synthesis as a bottleneck because of throughput limits, imperfect synthesis, and reagent requirements; sequencing and reconstruction also complicate reading. Its conclusion is direct: DNA is better suited as a complementary medium for cold archives than as a replacement for HDDs or SSDs.

A 2024 Library of Congress presentation by Twist Bioscience described a product-development effort and a planned terabyte-scale archive solution, while noting throughput, cost, and customer acceptance as challenges. A plan is not proof of general availability. Separately, Nature Computational Science identifies unresolved issues including error correction, retrieval from noisy sequencing reads, scalability, fault tolerance, random access, and data removal. For a buyer, the right question is not only how much information DNA could encode, but how quickly and reliably a provider can retrieve the specific file needed.

How much data can new glass-based storage hold compared with an SSD or hard drive?

Published glass demonstrations describe terabytes per platter, but those figures are not a like-for-like comparison with a consumer SSD or hard drive: the glass systems are designed for archival workloads, with different access equipment and performance priorities.

Microsoft reports raw capacity upwards of 7 TB on a square glass platter about the size of a DVD. The 2025 Nature study reports 4.84 TB of usable capacity on a 120 mm square, 2 mm thick fused-silica platter, with 301 data layers and a density of 1.59 Gbit/mm³. It also reports write throughput of 25.6 Mbit/s per beam in the demonstrated regime. In borosilicate glass, the study reports 2.02 TB usable per platter and 65.9 Mbit/s writing with four beams.

Those are research-system results, not a shopping specification for an external drive. Microsoft’s 2023 paper describes separate write and read technologies and a robotic library; it also says, “The production hardware is not yet fully operational.” That is the key maturity check: high capacity matters only if a usable system can write, locate, retrieve, and decode your data.

Cerabyte’s white paper reports a prototype write/read speed of 5 MB/s. Its higher speeds, rack capacities, and future media costs are roadmap targets, not verified commercial specifications. Its material also discusses accelerated-aging tests and potential millennia-scale life, which remains a company-reported estimate rather than evidence of consumer-ready media.

The practical call is simple: use conventional storage when quick, repeated access is the requirement. Treat glass, ceramic, and DNA as potential cold-archive layers where long retention matters more than instant retrieval. Before evaluating a provider, ask for demonstrated usable capacity, write and read throughput, retrieval workflow, and current product availability—not just a lifetime or density headline.