Microsoft researchers have revealed that borosilicate glass, a material commonly found in laboratory apparatus and kitchenware, can be used with femtosecond lasers to store data at unmatched densities and durations compared to current archival methods.
<div class="media-with-label__label">
Credit: tomeqs / Shutterstock </div>
</figure>
</div>
</div>
</div>
</div>Businesses grappling with the expense and intricacy of archiving data for the long term might soon have a novel solution: a pane of glass.
Fresh findings published this Wednesday indicate that a borosilicate glass plate, measuring 120mm square and a mere 2mm thick, is capable of storing 4.8TB of data across 301 layers. Accelerated aging tests further suggest this data would remain preserved for at least 10,000 years.
Microsoft researchers articulated in an article featured in the science and technology journal, Nature, that “Glass serves as a permanent data storage medium, exhibiting resilience to water, heat, and dust.” They added, “We have pioneered the scientific principles for parallel high-speed writing and devised a method enabling accelerated aging tests on the written glass, which implies the data should endure for a minimum of 10,000 years.”
Earlier iterations of this technology necessitated fused silica, a high-purity glass sourced from only a select few manufacturers. The latest discoveries demonstrate the system’s equally effective performance with borosilicate—a widely produced and significantly more economical material—thereby moving the technology closer to market readiness, the paper detailed.
The timing of these developments holds considerable importance.
The global volume of data is experiencing a doubling approximately every three years, according to Seagate research mentioned in the paper. Despite this growth, the authors pointed out that “most digital archive systems depend on media that deteriorate” well before the multi-decade retention periods increasingly mandated by legal, financial, and regulatory obligations.
Magnetic tape, currently the most common archival medium, illustrates these limitations. An LTO-10 (Linear Tape-Open) cartridge, the industry benchmark for enterprises, offers 30TB to 40TB of native storage at 400MB/s but has a rated lifespan of only 30 years. It demands climate-controlled environments between 16°C and 25°C and data migration approximately every five to ten years.
Analysts assert that this operational burden, rather than the media itself, constitutes the true cost of tape. Sanchit Vir Gogia, chief analyst at Greyhound Research, commented, “Archival installations seldom fail due to scheduled chemical degradation of cartridges. Instead, failures occur because compatibility windows close, drive generations advance, firmware support ends, and robotics require upgrades.”
Vishesh Divya, principal analyst at Gartner, observed that tape-as-a-service models have shifted some of this responsibility, transferring hardware lifecycle management to providers under specific service-level agreements.
Divya stated, “LTO tape remains the standard for enterprise cold storage. The cost per terabyte for the media is low, the ecosystem is well-established, and enterprises have decades of practical experience managing refresh cycles.”
Sony’s Optical Disc Archive—the primary optical alternative, offering 5.5TB per cartridge and a 100-year rated lifespan—was discontinued in March 2025, leaving no comparable product currently available.
Mechanism for Writing and Reading Data from Glass
Project Silica, Microsoft’s initiative for glass-based storage, employs femtosecond laser pulses to embed data as three-dimensional structures known as voxels within the glass. This process achieves a rate of 25.6 megabits per second per beam, with an energy consumption of 10.1 nanojoules per bit.
The document details two approaches to encoding. The initial method, birefringent voxels, alters the polarization properties of the glass. The team optimized the laser pulses into a pseudo-single-pulse technique—where a single pulse is split to start one voxel and finish another simultaneously—thereby accelerating beam scanning.
The second method, phase voxels, represents a novel innovation that instead modifies the phase properties of the glass, requiring only one pulse per voxel. Critically, this method functions effectively in borosilicate glass, unlike the birefringent technique. The researchers stated, “Significantly higher levels of three-dimensional inter-symbol interference in phase voxels can be addressed using a machine learning classification model.”
Previous Project Silica readers needed three or four cameras for data retrieval; the updated system operates with just one, marking what the researchers termed the first fully demonstrated end-to-end glass archival system, encompassing writing, storage, and retrieval.
Longevity Confirmed via Accelerated Aging Methods
The claim of long-term durability is supported by a non-destructive optical technique developed by the team to assess voxel degradation in situ, coupled with accelerated aging procedures applied to samples of written borosilicate. The researchers highlighted, “Accelerated aging tests conducted on written voxels in borosilicate suggest data lifespans exceeding 10,000 years.”
For corporate purchasers, longevity alone will not be a sufficient selling point.
Gogia noted, “A realistic total cost of ownership (TCO) comparison must consider multi-decade lifecycle horizons, not merely procurement cycles. Glass storage fundamentally alters the economic landscape by potentially removing migration cycles, thereby reducing labor, reconciliation efforts, and operational disruptions.” However, write speeds remain considerably slower than tape, making glass more suitable for ultra-cold, low-ingestion storage environments.
Compliance introduces an additional factor. Data encoded as permanent optical modifications cannot be overwritten, thereby mitigating ransomware risks. Nevertheless, Gogia cautioned that “compliance is an attribute of the system, not merely the substrate. Businesses must still guarantee encryption key rotation, metadata indexing, and thorough audit trails. A medium with a 10,000-year lifespan does not negate the responsibility to demonstrate sound governance.”
No Current Commercial Offering
Microsoft indicated in a separate blog post that the research phase for Project Silica has concluded. The company stated, “We are continually incorporating insights from Project Silica as we investigate the ongoing necessity for sustainable, enduring digital information preservation,” without outlining a commercialization strategy. If brought to market, glass storage is unlikely to completely replace tape.
Gartner’s Divya opined, “It is more probable to emerge as a niche solution for ultra-long retention, rather than a direct substitute for tape-based cold storage. Any new medium would need to compete across the entire stack—including economics, hardware, software, and operational models—not solely on media longevity.”
