Microsoft researchers have pulled off a feat that sounds like science fiction: they’ve stored data in ordinary borosilicate glass—the same material used in Pyrex kitchenware—and they project it could remain intact for at least 10,000 years. In a paper published February 18, 2026 in the journal Nature, the Project Silica team detailed breakthroughs that move the technology from expensive fused silica to cheap, widely available glass, marking the end of the research phase and a decisive step toward real-world use.

No product, service, or pricing has emerged yet, and Microsoft warns that commercialization still lies ahead. But for anyone responsible for long-term data preservation, the advance is a milestone worth understanding now.

A Platter of Glass That Outlasts Empires

The core idea hasn’t changed: femtosecond lasers etch data into the physical structure of glass by creating tiny, polarization-altered voxels inside the material. What’s new is the glass itself. Earlier demonstrations required pure fused silica—hard to source and expensive. The latest work, according to the Nature paper, extends the technique to borosilicate glass, a commodity material that slashes media costs and eliminates supply constraints.

Just as important are the writing and reading improvements. The team invented a new storage method called “phase voxels,” which modifies the glass’s phase rather than its polarization and needs only a single laser pulse per voxel. A “pseudo-single-pulse” technique splits one beam to lay down two adjacent voxels simultaneously, enabling rapid scanning across the medium. The reader now uses a single camera instead of three or four, shrinking the hardware footprint. And parallel writing—using multiple beams at once—pushes write speeds far beyond early lab prototypes.

The headline longevity figure comes from accelerated aging tests: by stressing the glass and measuring optical changes, researchers estimate data would survive at least 10,000 years. That’s long enough to outlast any magnetic tape, hard drive, or civilization-level calendar.

What This Means for You

Before you cancel your LTO orders or start pricing glass platters, understand where Project Silica fits—and where it doesn’t.

For home users and everyday PC owners: This isn’t a OneDrive replacement or a Windows Backup upgrade. Project Silica is a write-once, archival medium. It’s designed for data that never changes, not the living files, settings, and applications you recover after a crash or ransomware attack. Your Windows 11 backup routine shouldn’t change one bit.

For IT administrators and backup architects: Project Silica isn’t a tape killer—yet. Microsoft hasn’t published a product specification, an Azure SKU, a cost-per-terabyte, or any availability timeline. You can’t buy it, benchmark it, or integrate it into Veeam or Commvault. Your current investment in LTO, cloud archive tiers, or immutable object storage remains sound. The technology addresses a narrow, specific need: records that must remain unaltered for decades or centuries, surviving multiple generations of media and hardware migrations. Think finalized CAD release packages, regulated case evidence exports, completed research datasets with full provenance, or authoritative media masters. If your dataset changes frequently or you need instant retrieval, glass storage won’t be the answer.

For archivists and long-term planners: This is the breakthrough you’ve been waiting for—a medium that could fundamentally reduce the cost and frequency of media refreshes. Tape libraries demand regular integrity checks, hardware replacement, and eventual migration. Hard drives spin down over five years. Glass, in theory, sits on a shelf inert. Microsoft’s move to borosilicate glass addresses the biggest commercial barrier, and the completed research phase signals that engineering effort will now pivot to productization. But until a service exists, your next audit should focus on identifying which datasets actually justify a century-scale immutable tier.

The Long Road from Lab to Data Center

Project Silica didn’t appear overnight. Microsoft Research has publicly explored glass-based storage for years, starting with a collaboration with Warner Bros. to store the 1978 “Superman” movie on a quartz glass coaster. It then partnered with Global Music Vault to preserve recordings under Arctic ice, and worked with students on a “Golden Record 2.0” project meant to represent humanity for millennia.

Those proofs-of-concept relied on fused silica, limiting scalability. The 2026 Nature paper represents the culmination of a multi-year effort to overcome that hurdle, along with parallel writing, simplified optics, and a machine-learning-assisted decoding pipeline robust enough to handle the noise inherent in phase voxels. The research phase is now complete, Microsoft says, and the team has published its findings so others can build on them.

Yet the leap from academic paper to purchasable service is vast. Even if Microsoft embeds Project Silica into Azure Archive Storage, customers will need to understand ingest throughput, retrieval latency, error rates at scale, and per-object durability terms. The 10,000-year claim is an extrapolation from accelerated aging, not a real-time observation, and it presumes that the reader hardware, software decoders, and metadata catalogs all survive alongside the glass. Physical immutability doesn’t protect a platter from being lost, stolen, or trapped in a flooded facility without geographic copies.

What to Do Right Now

You don’t need to budget for glass storage in 2026. You do need to get your archive house in order, because the arrival of a viable immutable medium will reward organizations that understand their long-term data—and punish those that treat backup and archive as the same thing.

Start by auditing your data for real archival candidates. Ask of each dataset: Is it finalized and immutable? Does its retention period exceed the lifespan of your current storage platform? Can you define it independently of any running application? If yes, document its metadata, format, checksums, and chain of custody as if you were handing it to a future civilization.

Don’t pause your current backup or disaster-recovery investments. According to Microsoft’s advisory, LTO tape, cloud backup services, and immutability features in Windows Server and Microsoft 365 remain the tools you rely on today. Project Silica’s existence doesn’t change the reality that ransomware attacks happen now, not in 10,000 years.

If you already manage long-term archives, calculate your refresh burden honestly. Tally the labor, hardware refresh cycles, migration failures, and integrity-verification overhead of your current tape or disk-based archive. That number will be your baseline when a future Project Silica service eventually announces pricing—and it will tell you whether a shift makes financial sense.

Keep an eye on Microsoft’s Azure roadmap. The research phase is over, and the company is exploring “the ongoing need for sustainable, long-term preservation.” An Azure Archive Storage tier built on glass would be a natural first product, but no timeline exists. Third-party archival services or specialized hardware appliances could also emerge if Microsoft licenses the technology.

The Outlook

Project Silica has cleared its biggest scientific hurdles. The move to borosilicate glass makes the medium cheap and the reader small enough to be practical. Parallel writing and single-camera readout bring speeds in sight of something a modern data center might accept. The remaining unknowns are economic and operational: cost per terabyte, ingest pipelines, retrieval service-level agreements, and how organizations will handle reader obsolescence over millennia.

For now, treat this as a signal that the decades-old treadmill of media migration may have an endpoint. The organizations that will benefit most are those that start designing technology-neutral immutable archives today—whether those archives eventually land on glass, tape, or something else entirely.