How does Microsoft store data in glass?

A laser-written, ultra-durable archival medium

Researchers have demonstrated a new approach to long-term data storage that encodes digital information inside ordinary glass using tightly focused laser pulses. The technique alters the glass’s internal structure at microscopic scales so that information can be written into, and later read from, a palm-sized piece of silica. Teams describe capacities equivalent to millions of books and claim the medium can remain stable for thousands to tens of thousands of years if stored correctly.

How the system works

• Writing: Femtosecond or tightly controlled laser pulses locally change the optical properties of the glass, creating microscopic marks or structures that represent bits of data.
• Reading: High-resolution optical imaging combined with machine-learning algorithms decodes the altered regions and reconstructs the original files.
• Durability: Glass is chemically and thermally stable relative to magnetic tape or hard drives, resisting corrosion, most solvents and environmental degradation over long timescales.

Why this matters

Long-term digital preservation is a growing challenge for libraries, governments and cultural institutions. Conventional media degrade or require active power and migration to newer formats. A passive, stable medium that needs no energy to sit on a shelf and can be read with appropriate optics promises a new option for archival backups that could span centuries or millennia.

Limitations and next steps

• Speed and scale: Current demonstrations focus on dense archival writes, but write and read rates lag conventional storage and require specialized hardware.
• Cost and standardization: Mass deployment will need affordable, reliable writing and reading devices and open standards for long-term accessibility.
• Migration and stewardship: Long-lived media still require institutional stewardship — good metadata, format documentation and multiple copies in geographically separated vaults.

The work points to a feasible route for ultra-long-term digital memory, but broad adoption will depend on engineering, cost reductions and agreements on how to preserve the ability to read the glass decades from now.