Microsoft's Project Silica promises eternal storage. It can't get there from here
(2026/03/02)
- Reference: 1772446386
- News link: https://www.theregister.co.uk/2026/03/02/microsoft_project_silica_opinion/
- Source link:
Opinion There is more joy in heaven over a single report of genuinely new technology than in a thousand desperate AI marketing pitches. What the angels will make of Microsoft's Project Silica, a mixture of the two, is less clear.
The good bits first, and they are truly good. Project Silica is a [1]Microsoft research effort into data storage in glass, not in itself a new idea but one with plenty of innovation. It's gone through various iterations of how to write data into different sorts of glass matrices and how to read it out again. Project Silica can now store 2 TB in hundreds of layers within a 2 mm thick borosilicate glass plate. There are femtosecond laser pulses and machine learning readers. Proper science-led engineering.
The reason this matters is genuine. As a digital culture, we are extraordinarily good at growing and using data, and extraordinarily bad at keeping it for any length of time. The most economical form of bulk storage remains tape, a stretchy mix of thin plastics, binding gloops and metal oxides, as durable as it sounds. Witness the [2]arcane techniques behind the attempts to safely retrieve the Unix v4 source code from the only known tape, recently discovered at the University of Utah.
[3]
[4]Youtube Video
[5]
[6]
Glass, on the other hand, is silica, also known as silicon dioxide, and that can last basically forever. Flint tools – also silica – have been found, good as new, after 2 million years. Mesopotamian clay tablets – silica again – containing the first records of daily data from any civilization are crisply readable from 3,000 years ago. We haven't had borosilicate for a century yet, let alone tattooed it with femtosecond lasers, so we don't know how long it will last, but 10,000 years sounds like a conservative bet. It's also cheap enough to make kitchen bowls out of. If you don't love silica, you shouldn't be reading this website.
Microsoft says that deployed at cloud scale, commercialized optical storage in glass is the only plausible, scalable solution for ultra-reliable, long-term data storage. That's almost certainly right, and it will almost certainly never happen.
[7]
There are huge technical hurdles ahead, such as the writing speed of 20 Mbps per laser beam. That's USB 1.0, uncommon in datacenters. You can get more by multiple laser beams. Parallel femtosecond laser beams. Decoding the read signals needs to go through an AI model. Scaling this to gobble the 400 million terabytes – 0.4 zettabytes for the cool kids – thought to be generated daily is non-trivial.
However, it is probably not impossible. Our current storage technologies cope, and they contain a similarly insane level of technological innovation. They, however, have evolved over 70 years of intensely competitive demand, mixing bursts of magnificent technical innovation with a grinding march of continual improvements. There is no such path to market for Project Silica. There is no fundamental, universal, pressing economic need it can fulfill immediately and then evolve alongside. There's also the market flaw that once you've sold one 10,000-year storage solution, it'll be a while before anyone needs another. Even by utopian magic bean techbro business plan standards, that's a long way to the holy grail of return on investment.
[8]Every day in every way, passwords are getting worse and worse
[9]Digital sovereignty must define itself before it can succeed
[10]The Linux mid-life crisis that's an opportunity for Tux-led transformation
[11]Infrastructure cyberattacks are suddenly in fashion. We can buck the trend
Perhaps the worst sin of Project Silica is its utopianism. It's an easy feel-good parcel of hype to say that it could store all of our civilization's data, safe against decay and catastrophe, in a vast glassy lifeboat, ready for a factory restore and reboot of all we hold dear. Sure, things can get messy for a bit, but we can feel safe in the hope that once things cool off a bit, even a massive collapse can be Ctrl-Z'd.
Nope. The existence of an ultra-high-tech archive presupposes an ultra-high-tech culture to read it. Otherwise, it is as much use as Douglas Adams' intergalactic cruise ship, waiting for a delivery of lemon-soaked paper napkins before it can launch. When told that it was parked in the wreckage of a failed civilization, unable to provide this, it replied that future civilizations will rise and fall, and one will, in time, be able to provide lemon-soaked paper napkins. You cannot rebuild from scratch, even if you do have a record of how the last lot did it; they had resources, conditions, and millennia.
Pretending otherwise is a sinful distraction from facing up to the issues this pipe dream is supposed to ameliorate. It's as bad as Elon Musk banging on the absolute necessity to get to Mars to build the multi-planetary species, a similarly quasi-messianic promise and one that lasted as long as it took for him to find money on the Moon instead. Douglas Adams may have been one of Musk's favorite authors, but there's no sign he understood a word the man wrote.
[12]
There are accusations that the creators of technology read too much science fiction. They read too little. The preservation of data to save humanity is a trope that never works, from [13]Foundation to [14]Canticle for Leibowitz to [15]Hitchiker's Guide , it is deceptive, ineffective, or ephemeral. By all means, please, build robust storage that'll outlive Stonehenge. We need that, because knowledge and history matter more than anything. Just put it in the proper context of all the other things that need to be true for it to be useful, the discipline, vision, and honesty needed to make it viable. Otherwise, that hope will be as fragile as a heart of glass. ®
Get our [16]Tech Resources
[1] https://www.microsoft.com/en-us/research/blog/project-silicas-advances-in-glass-storage-technology/
[2] https://www.youtube.com/watch?si=X2-bLgoBc39SH5W3&v=-xlq_MPWNKk&feature=youtu.be
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/columnists&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aaVt0fGnuFrsJV9i3QjtsAAAA5g&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://www.youtube.com/watch?si=X2-bLgoBc39SH5W3&v=-xlq_MPWNKk&feature=youtu.be
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/columnists&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aaVt0fGnuFrsJV9i3QjtsAAAA5g&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/columnists&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aaVt0fGnuFrsJV9i3QjtsAAAA5g&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/columnists&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aaVt0fGnuFrsJV9i3QjtsAAAA5g&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[8] https://www.theregister.com/2026/02/23/password_opinion/
[9] https://www.theregister.com/2026/02/16/digital_sovereignty/
[10] https://www.theregister.com/2026/02/09/the_linux_midlife_crisis_thats/
[11] https://www.theregister.com/2026/02/02/energy_infrastructure_cyberattacks/
[12] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/columnists&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aaVt0fGnuFrsJV9i3QjtsAAAA5g&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[13] https://search.worldcat.org/title/1288403332
[14] https://search.worldcat.org/title/755017271
[15] https://search.worldcat.org/title/1015319233
[16] https://whitepapers.theregister.com/
The good bits first, and they are truly good. Project Silica is a [1]Microsoft research effort into data storage in glass, not in itself a new idea but one with plenty of innovation. It's gone through various iterations of how to write data into different sorts of glass matrices and how to read it out again. Project Silica can now store 2 TB in hundreds of layers within a 2 mm thick borosilicate glass plate. There are femtosecond laser pulses and machine learning readers. Proper science-led engineering.
The reason this matters is genuine. As a digital culture, we are extraordinarily good at growing and using data, and extraordinarily bad at keeping it for any length of time. The most economical form of bulk storage remains tape, a stretchy mix of thin plastics, binding gloops and metal oxides, as durable as it sounds. Witness the [2]arcane techniques behind the attempts to safely retrieve the Unix v4 source code from the only known tape, recently discovered at the University of Utah.
[3]
[4]Youtube Video
[5]
[6]
Glass, on the other hand, is silica, also known as silicon dioxide, and that can last basically forever. Flint tools – also silica – have been found, good as new, after 2 million years. Mesopotamian clay tablets – silica again – containing the first records of daily data from any civilization are crisply readable from 3,000 years ago. We haven't had borosilicate for a century yet, let alone tattooed it with femtosecond lasers, so we don't know how long it will last, but 10,000 years sounds like a conservative bet. It's also cheap enough to make kitchen bowls out of. If you don't love silica, you shouldn't be reading this website.
Microsoft says that deployed at cloud scale, commercialized optical storage in glass is the only plausible, scalable solution for ultra-reliable, long-term data storage. That's almost certainly right, and it will almost certainly never happen.
[7]
There are huge technical hurdles ahead, such as the writing speed of 20 Mbps per laser beam. That's USB 1.0, uncommon in datacenters. You can get more by multiple laser beams. Parallel femtosecond laser beams. Decoding the read signals needs to go through an AI model. Scaling this to gobble the 400 million terabytes – 0.4 zettabytes for the cool kids – thought to be generated daily is non-trivial.
However, it is probably not impossible. Our current storage technologies cope, and they contain a similarly insane level of technological innovation. They, however, have evolved over 70 years of intensely competitive demand, mixing bursts of magnificent technical innovation with a grinding march of continual improvements. There is no such path to market for Project Silica. There is no fundamental, universal, pressing economic need it can fulfill immediately and then evolve alongside. There's also the market flaw that once you've sold one 10,000-year storage solution, it'll be a while before anyone needs another. Even by utopian magic bean techbro business plan standards, that's a long way to the holy grail of return on investment.
[8]Every day in every way, passwords are getting worse and worse
[9]Digital sovereignty must define itself before it can succeed
[10]The Linux mid-life crisis that's an opportunity for Tux-led transformation
[11]Infrastructure cyberattacks are suddenly in fashion. We can buck the trend
Perhaps the worst sin of Project Silica is its utopianism. It's an easy feel-good parcel of hype to say that it could store all of our civilization's data, safe against decay and catastrophe, in a vast glassy lifeboat, ready for a factory restore and reboot of all we hold dear. Sure, things can get messy for a bit, but we can feel safe in the hope that once things cool off a bit, even a massive collapse can be Ctrl-Z'd.
Nope. The existence of an ultra-high-tech archive presupposes an ultra-high-tech culture to read it. Otherwise, it is as much use as Douglas Adams' intergalactic cruise ship, waiting for a delivery of lemon-soaked paper napkins before it can launch. When told that it was parked in the wreckage of a failed civilization, unable to provide this, it replied that future civilizations will rise and fall, and one will, in time, be able to provide lemon-soaked paper napkins. You cannot rebuild from scratch, even if you do have a record of how the last lot did it; they had resources, conditions, and millennia.
Pretending otherwise is a sinful distraction from facing up to the issues this pipe dream is supposed to ameliorate. It's as bad as Elon Musk banging on the absolute necessity to get to Mars to build the multi-planetary species, a similarly quasi-messianic promise and one that lasted as long as it took for him to find money on the Moon instead. Douglas Adams may have been one of Musk's favorite authors, but there's no sign he understood a word the man wrote.
[12]
There are accusations that the creators of technology read too much science fiction. They read too little. The preservation of data to save humanity is a trope that never works, from [13]Foundation to [14]Canticle for Leibowitz to [15]Hitchiker's Guide , it is deceptive, ineffective, or ephemeral. By all means, please, build robust storage that'll outlive Stonehenge. We need that, because knowledge and history matter more than anything. Just put it in the proper context of all the other things that need to be true for it to be useful, the discipline, vision, and honesty needed to make it viable. Otherwise, that hope will be as fragile as a heart of glass. ®
Get our [16]Tech Resources
[1] https://www.microsoft.com/en-us/research/blog/project-silicas-advances-in-glass-storage-technology/
[2] https://www.youtube.com/watch?si=X2-bLgoBc39SH5W3&v=-xlq_MPWNKk&feature=youtu.be
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/columnists&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aaVt0fGnuFrsJV9i3QjtsAAAA5g&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://www.youtube.com/watch?si=X2-bLgoBc39SH5W3&v=-xlq_MPWNKk&feature=youtu.be
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/columnists&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aaVt0fGnuFrsJV9i3QjtsAAAA5g&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/columnists&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aaVt0fGnuFrsJV9i3QjtsAAAA5g&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/columnists&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aaVt0fGnuFrsJV9i3QjtsAAAA5g&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[8] https://www.theregister.com/2026/02/23/password_opinion/
[9] https://www.theregister.com/2026/02/16/digital_sovereignty/
[10] https://www.theregister.com/2026/02/09/the_linux_midlife_crisis_thats/
[11] https://www.theregister.com/2026/02/02/energy_infrastructure_cyberattacks/
[12] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/columnists&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aaVt0fGnuFrsJV9i3QjtsAAAA5g&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[13] https://search.worldcat.org/title/1288403332
[14] https://search.worldcat.org/title/755017271
[15] https://search.worldcat.org/title/1015319233
[16] https://whitepapers.theregister.com/
Data integrity my behind!
Alumoi
Decoding the read signals needs to go through an AI model.
So it's a 50-50 chance of hallucination. I know, it will be called Schrödinger storage.
Re: Data integrity my behind!
Korev
Luckily there will be lots of cat videos to store
What happens in 1,000 years time ...
alain williams
when you try to access it and the reading software complains “no licensing servers available” ?
Doctor Syntax
"There's also the market flaw that once you've sold one 10,000-year storage solution, it'll be a while before anyone needs another."
In fact you do need another. You need one to replace the first that died of old age. You need one to interface to USB-D or whatever comes along in the next generation of computers to replace the interface you're currently using. And if that doesn't happen your borosilicate plate with all the world's knowledge on it, being only 2mm thick, isn't even useable as a dinner plate.
The great new BBC Domesday book of 40 years ago became unreadable in very short order without some scrambling about to find a surviving reader. The original from 940 years ago is still readable as in needs nothing more than the Mk 1 eyeball (backed up by a knowledge of palaeography and medieval Latin usages).
Too much
Anonymous Coward
There's too much data in the world to store on these (expensive and cumbersome) silicate discs. The problem isn't how to store it, but what to store. The internet currently generates many petabytes of data PER DAY.
I believe that, just like in antiquity, most data and knowledge will be lost over time. The only valid time capsules are books, printed on durable paper.
Will it work on the trinitite created by a nuclear apocalypse?
Asking for a friend.