Data in glass: Project Silica and the crystal orbiting the Sun

Two pieces of laser-written glass get quoted whenever someone asks how long data could last. One sits in a research lab at Microsoft; the other has been orbiting the Sun since 2018. They are often blurred into one story about data surviving millions of years in space. They are different projects making different claims, and the difference matters.

Project Silica: an archive for data centres

In February 2026 Microsoft Research published the main result of Project Silica in Nature: a complete system for writing, storing and reading data in glass. A femtosecond laser writes tiny three-dimensional marks, called voxels, in hundreds of layers inside a plate only 2 mm thick. The paper reports 4.8 TB on a 120 × 120 mm plate of fused silica, and — the part Microsoft emphasised — the same technique working in ordinary borosilicate glass, the kind used in oven doors, which is far cheaper and easier to source.

The headline figure is at least 10,000 years. It comes from accelerated ageing: heating written glass, measuring how fast the marks degrade, and extrapolating to room temperature. Independent coverage notes the estimate does not account for mechanical stress or chemical corrosion. Microsoft describes the research phase as complete; there is no product, price or release date.

Silica was always aimed at cloud archives on Earth — material written once and read rarely, which today sits on magnetic tape that has to be rewritten every few decades. Earlier demonstrations stored Warner Bros.’ Superman and music for the Global Music Vault.

The crystal in the Roadster: the “millions of years” claim

The space story belongs to a different glass. The University of Southampton’s 5D optical storage, developed by Peter Kazansky’s group, also writes with a femtosecond laser into nanostructured quartz. In February 2018 the Arch Mission Foundation put one such disc, holding Isaac Asimov’s Foundation trilogy, in the Tesla Roadster launched on the first Falcon Heavy. Southampton describes its lifetime as virtually unlimited at room temperature and the medium as stable up to 1,000 °C; the Foundation describes the library as orbiting the Sun for millions of years.

That disc will probably outlast every hard drive ever made. It will also almost certainly never be read, because nobody plans to go and fetch it.

What glass solves

Glass attacks the problem that dominates archive budgets today: media decay. Tape and drives fail within decades and must be copied forward on a schedule. A medium that simply does not degrade on human timescales removes that cycle for the bits themselves, and needs no power while it waits.

What glass does not solve

Readers. A plate is useless without a machine that can read it, and that machine has to exist when someone needs it. Silica’s reader is a microscope with a camera and a machine-learning model; whether one is available in 2200 is a question about institutions, not physics.

Formats. Perfectly preserved bytes in a format nothing can interpret are perfectly preserved noise. Durable media make format choice more important, not less.

Meaning. A plate can hold every photograph of a family and still say nothing about who is in them. Description is the part of an archive that no medium supplies.

Glass is very good news for the bottom layer of preservation, the bits on the medium. The other layers stay where they were: open formats, a self-describing index, and people with a reason to keep reading.