The future of long-term data storage is clear and will last 14 billion years

After decades of research and development, humanity finally has a data storage medium that will outlast us. The 5D Memory Crystal stores data by using tiny voxels – 3D pixels – in fused silica glass, etched by femtosecond laser pulses. These voxels possess "birefringence," meaning that their light refraction characteristics vary depending upon the polarization and direction of incoming light. That difference in light orientation and strength can be read in conjunction with the voxel's location (x, y, z coordinates), allowing data to be encoded in five dimensional space. And because the medium is silica crystal, similar to optical cable, it's highly durable. It's also capacious: The technology can store up to 360 TB of data on a 5-inch glass platter. SPhotonix, a company formed last year to commercialize the technology, estimates that even at 190 degrees Celsius, voxels cut in silica crystal should last 13.8 billion years, the estimated age of the universe, barring some mishap – which seems inevitable at that time scale. While none of us will be around to verify that claim, silica crystal is vastly more stable than magnetic or electronic methods of data storage. Even alternative optical storage technologies like optical discs are only expected to last 5 to 100 years, though M-DISC advertises a lifespan of 1,000 years. SPhotonix was co-founded in 2024 by Peter Kazansky, professor in optoelectronics at the University of Southampton, and his son Ilya, an entrepreneur who was co-founder and CTO of Statiq, among other technology leadership roles. Last month, the Delaware-based company, which maintains research facilities in the UK and Switzerland, announced that it had raised $4.5 million to develop its technology for deployment in data centers. Ilya Kazansky told The Register in an interview that the funding received to date will take 5D Memory Crystal from Technology Readiness Level (TRL) 5 – the point of technical validation – to TRL 6 – prototype demonstration. "We're speaking to a lot of the hyperscalers about releasing some of our prototypes into their data centers over the course of the next couple of years," he said. Citing IDC predictions that by 2028 there will be 394 trillion zettabytes of data generated annually, Kazansky said the company is primarily focused on cold data storage applications, which refers to data that can wait ten seconds or more before delivery. Applications like financial transaction data or high frequency trading require hot data storage that can deliver information in less than five milliseconds, he explained. That's the realm of SSDs. There's also warm and cool data storage, when retrieval times can range from 20 milliseconds to one second – time periods suitable for streaming videos and document storage, for instance. "Statistics show that between 60 to 80 percent of all data which is currently stored globally is classed as cold data," said Kazansky. "However, because of the way that humanity is developing, because of all of the budgets and AI and so on and so forth, a lot of businesses historically have been like, 'look, we are just going to use hard disk drives or SSDs,' which are expensive, which are bad for the environment because they consume a lot of energy. They're non-recyclable. They fail often, but they're just easier to use. Through inertia, people have been using the incorrect type of tool for a use case that can be used with a different tool." Heart of glass: Human genome stored for 'eternity' in 5D memory crystal Meta and Google turn to NextEra to feed insatiable datacenter power hunger Micron ditches consumer memory brand Crucial to chase AI riches London grid crunch delays new housing amid datacenter boom Kazansky argued that 5D Memory Crystal would be a better choice because it's ultra-durable, sustainable, and scalable. "We believe this is the only way that the industry is going to be able to scale the data storage capacity given the growing demand," he said. Kazansky pointed to the impact that surging AI demand has had on the availability of hard drives with 30 terabyte capacity or greater. Over the next three to four years, Kazansky said, SPhotonix aims to improve the data transfer speed of its technology from a write time of 4 megabytes per second (MBps) and read time of 30 MBps to a read/write speed of 500 MBps, which would be competitive with archival tape backup systems. SPhotonix has already demonstrated that its 5D Memory Crystal can store the human genome; the Eon Ark Time Capsule, an archive of recorded conversations from 2024 and 2025; and the game Heroes of Might & Magic 3. This archival service is open for business. Data retrieval, however, still needs to be done through the company's lab. In the next 18 months, the company hopes to have a field-deployable read device that customers can use to read archived data. But SPhotonix isn't presently targeting the consumer market. Kazansky estimates that the initial cost of the read device will be about $6,000 and the initial cost of the write device will be about $30,000. Kazansky said that the company's main focus is the hyperscaler data center and business-to-business market, for commercial applications that have to store huge amounts of data for long periods of time. "We need another three or four years of R&D to get it to the production and marketing standpoint," Kazansky said. "But we now have a clear understanding based on the latest research and discoveries of how to get to production, of how to get the level of unity economics quite as well." He added that the company soon plans to announce the hiring of a former Microsoft researcher who worked on Project Silica – a parallel data storage effort stemming from Microsoft's participation in the initial University of Southampton research. "We are not aiming to become a manufacturing company," said Kazansky. "We are a technology licensing company. We love the model of Arm Holdings. And to a certain extent, we love the model of Nvidia. So we are developing the enablement technology, and then we're going to be forming some form of a consortium, some form of a group of companies that will help us to bring this technology to market." ®