Microsoft’s Project Silica can now store approximately 2 terabytes of data on a single piece of ordinary borosilicate glass—the same material as your kitchen cookware—using femtosecond laser pulses. A Nature paper published February 18, 2026, confirms that accelerated aging tests project this data will remain intact for at least 10,000 years. No power required. No magnetic decay. No migration cycles.
What Is Project Silica?
Project Silica is Microsoft Research’s multi-year effort to encode digital data directly inside glass substrates using ultrafast laser pulses. The concept is simple in principle and devilishly difficult in execution: fire a laser at sub-nanosecond intervals to modify the internal structure of glass at precise depths, creating voxels—three-dimensional pixels—that encode binary information. Read those voxels back later using polarization-sensitive microscopy or, in the latest system, a single camera with standard light.1
The project’s ambition stems from a genuine engineering problem. Every archival storage medium in use today has a meaningful shelf life—magnetic tape degrades and requires periodic migration, optical discs last decades at best, and hard drives are designed for active use, not centuries of dormancy. Glass, by contrast, is chemically inert, non-magnetic, and physically stable across an enormous temperature range. It doesn’t corrode. It doesn’t demagnetize. Properly sealed, it doesn’t absorb moisture.
The original Project Silica work encoded data in fused silica—a high-purity, expensive material used in precision optics. That approach worked. Microsoft stored the entire 1978 Warner Bros. “Superman” film on a piece of quartz glass roughly the size of a drink coaster as an early proof of concept.2 But fused silica’s cost and the complexity of writing to it made commercialization difficult. The February 2026 Nature paper changes that calculus.
How Does Phase Voxel Encoding Work?
The 2026 breakthrough introduces a new encoding method called phase voxels. Previous Project Silica work modified glass by changing how it interacts with polarized light—a measurable but complex property requiring multiple laser pulses per voxel and three to four cameras to read. Phase voxels work differently: instead of altering polarization, they alter the material’s refractive index, the rate at which light travels through it.
Critically, a single femtosecond laser pulse creates one phase voxel. That’s a significant simplification from the prior approach.3
The technical specifications of the February 2026 system illustrate what this achieves in practice:
- Platter size: 120mm × 120mm × 2mm (approximately the footprint of a standard optical disc)
- Layer count: 258 stacked layers of data within the glass volume
- Storage density: 0.678 Gbit/mm³
- Capacity per platter: approximately 2.02 TB at four modulation levels
- Write speed: 18.4 Megabits per second
- Reader: one camera (reduced from three to four in the prior system)
The move to borosilicate glass—the same material in Pyrex bakeware and laboratory equipment—matters beyond mere economics. Borosilicate is one of the most manufactured specialty glasses in the world. Supply chains exist. Fabrication tolerances are well understood. The material is available at commodity scale in ways that precision optical-grade fused silica is not.4
Why the Borosilicate Shift Changes Everything
To understand why the borosilicate breakthrough matters, consider what it replaces. The prior Project Silica demonstrations used fused silica, a material that requires exotic manufacturing conditions and is priced accordingly. The project was technically successful but commercially implausible for anything beyond the most high-value archival use cases—national archives, film studios with irreplaceable assets, or scientific datasets with no substitutes.
Borosilicate glass is not exotic. It is manufactured at industrial scale globally. Moving the technology to this substrate while retaining the 10,000-year longevity claim—backed by accelerated aging tests conducted by the research team—shifts Project Silica from a laboratory curiosity toward a potential commercial archival medium.5
Microsoft has not announced pricing or a commercial rollout date. The company’s research blog states the research phase is now complete and that Microsoft is “exploring the ongoing need for sustainable, long-term preservation of digital information.” That careful language suggests active development rather than an imminent product launch.
Real-World Proofs of Concept
Project Silica has moved beyond lab demonstrations. Two notable deployments illustrate where this technology fits today:
Warner Bros. “Superman” (1978): In an early proof of concept, Microsoft stored the complete film on a quartz glass platter roughly 75mm × 75mm × 2mm. The test validated that complex, high-volume data (a full feature film) could survive storage and retrieval. Warner Bros. has significant archival needs—decades of film negatives and master recordings require preservation beyond what digital tape can reliably guarantee.
Global Music Vault, Svalbard: Microsoft partnered with the Global Music Vault to archive irreplaceable musical recordings under the ice in Svalbard, Norway. The vault is engineered for millennium-scale preservation, and Project Silica glass platters were selected as the storage medium.6 This deployment is active, not theoretical.
Both cases share a profile: high-value, write-once content where retrieval may be infrequent but data loss is unacceptable. That’s exactly the niche where glass storage makes economic sense—even at current costs that remain higher than tape.
How Glass Storage Compares to Alternatives
The archival storage landscape has several competing approaches, each with genuine trade-offs:
| Technology | Estimated Lifespan | Capacity (per unit) | Requires Power? | Migration Needed? | Approximate Media Cost |
|---|---|---|---|---|---|
| Project Silica (glass) | 10,000+ years | ~2 TB/platter | No | No | Not disclosed (research phase) |
| LTO Magnetic Tape | 30–50 years (active) | 18 TB/cartridge (LTO-9) | No (storage) | Every 5 years | ~$5/TB |
| M-DISC Optical | Up to 1,000 years | 25–100 GB/disc | No | No | ~$2–5/disc |
| DNA Storage | Thousands of years (theoretical) | Extremely high density (theoretical) | No | No | Not commercially available |
| Standard Optical (BD-R) | 50–100 years | 25–100 GB/disc | No | Recommended | <$1/disc |
| HDD/SSD | 3–10 years | 4–20 TB/drive | No (storage) | Frequent | $15–30/TB |
Magnetic tape (LTO format) remains the dominant archival technology for enterprise use. At roughly $5 per terabyte for media costs and 18 TB per LTO-9 cartridge, it offers an unmatched cost-per-bit for cold storage.7 But tape requires periodic migration—every five years or so, data must be copied to fresh media to prevent bit rot. Over 10,000 years, that migration overhead becomes absurd. Over 50 years, it’s a real operational and budget consideration that organizations routinely underestimate.
M-DISC optical media, manufactured using an inorganic recording layer, has passed Department of Defense environmental testing and offers up to 1,000 years of estimated lifespan. It’s the current practical choice for organizations needing multi-century archival without the complexity of tape management. Glass storage’s claimed 10x longevity advantage over M-DISC becomes relevant only for genuinely multi-generational preservation needs.
DNA storage remains several years from practical scale deployment. The DNA Data Storage Alliance estimated in 2025 that archival DNA storage at scale remains three to five years out—and that assessment may be optimistic given the read/write complexity involved.8
The Data Crisis Project Silica Is Solving
The archival storage market is projected to reach between $8.7 billion and $25 billion by the early 2030s, depending on methodology.9 That range reflects real uncertainty about how organizations will handle the compounding problem of data volume growth and long-term preservation requirements.
The underlying math is stark. Global data creation grows roughly 23% annually. The fraction of that data qualifying as “archival”—regulatory records, scientific datasets, cultural heritage, legal evidence—is enormous. Magnetic tape handles the bulk of this today, but it creates a continuous migration burden. Every five years, every byte stored on tape must be re-read, validated, and re-written to fresh media. At scale, that’s not a one-time cost; it’s a permanent operational overhead that compounds as data volumes grow.
Glass storage, if it reaches commercial scale, would eliminate that cycle entirely for qualifying data. Write once, read whenever—across a timeframe that makes periodic migration meaningless.
What Isn’t Proven Yet
The February 2026 Nature paper represents a significant technical milestone, but the gap between research demonstration and commercial deployment remains wide. Several questions don’t have public answers:
Manufacturing at scale: A single research-grade glass platter is not a production process. Consistent quality at volume, with reproducible voxel density across millions of units, requires engineering that the research paper doesn’t address.
Read hardware availability: The simplified one-camera reader still requires specialized equipment that doesn’t exist commercially. Retrieval from glass platters in 2030 requires that the hardware to read them exists and is maintained.
Actual cost structure: “Cheaper than fused silica” doesn’t translate into a price point. Until Microsoft or a licensee publishes commercial pricing, comparisons to tape and M-DISC on cost remain speculative.
Ecosystem: LTO tape has a mature ecosystem—robotic libraries, software integrations, certified workflows. Glass storage starts from zero on all of that.
Microsoft’s own language about “exploring ongoing need” rather than announcing a product suggests the company is still assessing commercialization paths, possibly through Azure archival services rather than standalone hardware.
Frequently Asked Questions
Q: Can Project Silica glass actually last 10,000 years? A: The claim is backed by accelerated aging tests—a standard materials science methodology—not direct observation. Researchers subject written glass to stress conditions to model long-term degradation. The 10,000-year figure is a scientifically grounded projection, not a guarantee, and assumes controlled storage conditions.
Q: When will glass storage be commercially available? A: Microsoft has not announced a commercial timeline. The February 2026 Nature paper marks the completion of the research phase. Industry analysts suggest pilot deployments in specialized applications (government archives, film preservation) could appear within a few years, with broader enterprise availability potentially arriving in the late 2020s—but no product has been announced as of February 2026.
Q: How does glass storage compare to LTO tape for enterprise archives? A: LTO tape offers vastly lower media cost today (~$5/TB vs. undisclosed glass pricing) and higher per-unit capacity (18 TB vs. ~2 TB for current glass platters). Glass storage’s advantage is zero migration burden over centuries-long retention periods and write-once immutability. For most enterprise use cases with sub-50-year retention, LTO tape remains the rational choice.
Q: Is glass storage the same as what’s used in Blu-ray discs? A: No. Blu-ray uses laser pits and lands etched into a polymer layer within a glass-composite disc. Project Silica writes data volumetrically inside the glass itself, across hundreds of stacked layers, using femtosecond laser pulses. The physics are entirely different, as is the longevity—Blu-ray media is rated for 50–100 years under ideal conditions.
Q: What data belongs on glass storage versus conventional cloud or tape? A: Glass storage targets the “write once, never migrate” use case: cultural heritage (films, music, art), scientific raw data, legal records with extended statutory retention, and national archives. Active data, frequently accessed data, or data with uncertain long-term requirements belongs on cloud or tape infrastructure with lower per-terabyte costs and greater flexibility.
Sources:
- Project Silica’s advances in glass storage technology - Microsoft Research
- Microsoft team creates ‘revolutionary’ data-storage system that lasts for millennia - Nature
- Glass data storage solution could last millennia - C&EN
- Microsoft Project Silica finds cheaper glass, boosts write speeds - Blocks & Files
- Laser-etched glass can store data for 10,000 years, Microsoft says - TechXplore
- Microsoft’s Project Silica write-once storage - Tom’s Hardware
- Microsoft’s Glass Breakthrough: Coming to a Data Center Near You? - Data Center Knowledge
- Project Silica promises to store data in glass for millennia - TechRadar
- Tape, glass, and molecules – the future of archival storage - The Register
- Project Silica proof of concept stores Warner Bros. Superman - Microsoft Source
Footnotes
-
Microsoft Research. “Project Silica’s Advances in Glass Storage Technology.” Microsoft Research Blog, February 2026. https://www.microsoft.com/en-us/research/blog/project-silicas-advances-in-glass-storage-technology/ ↩
-
Microsoft News. “Project Silica proof of concept stores Warner Bros. ‘Superman’ movie on quartz glass.” Microsoft Source. https://news.microsoft.com/source/features/innovation/ignite-project-silica-superman/ ↩
-
Castelvecchi, Davide. “Microsoft team creates ‘revolutionary’ data-storage system that lasts for millennia.” Nature, February 2026. https://www.nature.com/articles/d41586-026-00502-2 ↩
-
ACS Chemical & Engineering News. “Glass data storage solution could last millennia.” C&EN, February 2026. https://cen.acs.org/materials/electronic-materials/Glass-data-storage-solution-last/104/web/2026/02 ↩
-
Blocks & Files. “Microsoft Project Silica finds cheaper glass, boosts write speeds but trims capacity.” February 2026. https://www.blocksandfiles.com/data-protection/2026/02/19/microsoft-project-silica-finds-cheaper-glass-boosts-write-speeds-but-trims-capacity/4091610 ↩
-
Microsoft Research. “Project Silica.” Project page. https://www.microsoft.com/en-us/research/project/project-silica/ ↩
-
The Register. “Tape, glass, and molecules – the future of archival storage.” June 2025. https://www.theregister.com/2025/06/12/archival_storage_feature/ ↩
-
Live Science. “This new DNA storage system can fit 10 billion songs in a liter of liquid — but challenges remain.” 2025. https://www.livescience.com/technology/computing/this-new-dna-storage-system-can-fit-10-billion-songs-in-a-liter-of-liquid-but-challenges-remain-for-the-unusual-storage-format ↩
-
Business Research Insights. “Archival Data Storage Market Size, Consumer Behavior & Forecast.” 2025. https://www.businessresearchinsights.com/market-reports/archival-data-storage-market-120392 ↩