Microsoft Research Details 10,000-Year Data Storage Breakthrough in Nature

Microsoft Research has announced a breakthrough in long-term data storage, detailing a method to preserve information for 10,000 years. Published February 18, 2026 in Nature, the research extends Project Silica’s glass-based storage technology from costly fused silica to readily available borosilicate glass—the same material found in everyday kitchen cookware. This innovation addresses critical barriers to commercialization by lowering costs and increasing media availability. “Storing data inside glass with femtosecond laser pulses is one of the few technologies on the horizon with the potential for durable, immutable, and long-lived storage,” notes Microsoft Research. The new technique also simplifies the reading process, requiring only one camera instead of three, and enables faster data encoding.

Borosilicate Glass Enables Cost-Effective, High-Volume Data Storage The potential for archiving data for 10,000 years is now significantly closer to reality thanks to a shift in materials for long-term digital storage. Researchers have moved beyond expensive fused silica, successfully demonstrating data encoding within ordinary borosilicate glass – the same substance commonly found in kitchen cookware and oven doors. This breakthrough, detailed in a recent Nature publication, tackles a major impediment to widespread adoption of glass-based archival systems: prohibitive costs and limited media availability. Prior methods relied on a glass type “relatively difficult to manufacture and available from only a few sources,” but this new approach promises scalability. The innovation doesn’t stop at material science.

The team has also streamlined the reading process, reducing the number of cameras required from three or four to just one, thereby lowering both cost and the physical size of the reader. Manufacturing is also simplified, with writing devices now requiring fewer components for calibration and faster data encoding. A key element of this advancement lies in the development of “phase voxels,” a new storage method where only a single laser pulse is needed to create a data point. “We invented a new type of data storage in glass called phase voxels, in which the phase change of the glass is modified instead of its polarization, showing that only a single pulse is necessary to make a phase voxel,” explained the researchers. Furthermore, the team has unlocked methods for parallel, high-speed writing, utilizing a multi-beam delivery system to encode multiple data points simultaneously. Accelerated aging tests, combined with a new nondestructive optical method for identifying data degradation, support the claim that data stored in this manner should remain intact for at least a millennium. “We extended the industry standard Gray codes to apply to nonpower-of-two numbers of symbols,” adding another layer of robustness to the system. This progress builds on years of innovation in the field, positioning glass as a viable solution for the enduring preservation of digital information. Birefringent & Phase Voxel Methods Advance Laser Data Encoding Recent progress in long-term digital data preservation centers on novel approaches to encoding information within glass, moving beyond traditional storage media susceptible to rapid degradation. This shift addresses critical limitations in cost and material sourcing that previously hindered widespread adoption of glass-based archival storage. Researchers detailed these advancements in a Nature publication, outlining breakthroughs in both writing and reading techniques. A key innovation lies in the refinement of birefringent voxel writing, where the team discovered that the polarization of the initial laser pulse is inconsequential to the final voxel’s polarization. This allowed them to develop a “pseudo-single-pulse writing” method, splitting a single pulse to form voxels more efficiently and accelerate the writing process through beam scanning.

The team successfully mitigated the increased interference inherent in phase voxels using a machine learning classification model. Parallel writing capabilities were also significantly enhanced by combining a mathematical model of thermal effects within the glass with a multi-beam delivery system, enabling the simultaneous encoding of numerous data voxels. This dramatically increases writing speed. Moreover, the new reader design requires only one camera, streamlining the process and reducing both cost and size. Glass is a permanent data storage material that is resistant to water, heat, and dust.

Parallel Writing Achieved via Multi-Beam Delivery Systems Researchers at Microsoft Research are dramatically accelerating data storage within glass media, moving beyond limitations of earlier techniques with a new focus on parallel writing. This innovation builds upon their existing work with Project Silica, an effort to create archival storage capable of preserving information for up to 10,000 years, but addresses a critical bottleneck in the writing process. Previously, the process relied on sequentially addressing each storage location, limiting write speeds.

The team also explained a method for using light emissions—a byproduct of voxel formation—for both static calibration and dynamic control, enabling fully automatic writing operations. This advancement isn’t merely about speed; it’s about practicality. This shift tackles key barriers to commercialization, specifically cost and material availability. They’ve demonstrated that these phase voxels can be formed in borosilicate glass and have devised a method to read the encoded phase information, even mitigating the resulting three-dimensional interference with a machine learning classification model. This advance addresses key barriers to commercialization: cost and availability of storage media.

Machine Learning Optimizes Symbol Encoding & Longevity Testing The longevity of digital archives is being radically extended through the application of machine learning, according to findings from Microsoft Research’s Project Silica. Beyond simply identifying a durable storage medium – ordinary borosilicate glass – the team has focused on optimizing how data is written and verified for millennia. A key innovation lies in a new approach to symbol encoding, where machine learning algorithms refine the process to minimize errors and maximize data integrity over extended periods. This isn’t just about storing bits; it’s about ensuring those bits remain reliably readable after thousands of years. Researchers developed a new nondestructive optical method to identify the aging of data storage voxels within the glass, combining this with standard accelerated aging techniques to support data lasting 10,000 years. This allows for proactive assessment of data degradation, a crucial step in guaranteeing long-term preservation. “We developed a new way to optimize symbol encodings using machine learning and a better way to understand the tradeoff between error rates, error protection, and error recovery when evaluating new digital storage systems,” explained the team. The advances don’t stop at error correction. Project Silica also refined the writing process itself, moving towards “pseudo-single-pulse writing” where a single laser pulse is split to form multiple voxels simultaneously, dramatically increasing speed. This combination of optimized encoding, accelerated aging tests, and faster writing promises a robust and scalable solution for truly long-term data archiving, moving beyond the limitations of current magnetic and solid-state technologies. Source: https://www.microsoft.com/en-us/research/blog/project-silicas-advances-in-glass-storage-technology/ Tags: