[HN Gopher] Microsoft team creates data-storage system that last...
___________________________________________________________________
Microsoft team creates data-storage system that lasts for millennia
Author : gnabgib
Score : 92 points
Date : 2026-02-18 20:12 UTC (3 days ago)
(HTM) web link (www.nature.com)
(TXT) w3m dump (www.nature.com)
| gnabgib wrote:
| Paper [Laser writing in glass for dense, fast and efficient
| archival data
| storage](https://www.nature.com/articles/s41586-025-10042-w)
| ortusdux wrote:
| Any idea why they are reporting the estimated lifespan at
| 290degC? Testing seems to have been done at 440degC and above.
| casey2 wrote:
| Coz the paper gives a function for extrapolating from these
| tests. This is purely testing thermal decay.
|
| 10,000 years sounds like a good benchmark and isn't as
| obviously ridiculous as saying a million years at 260degC
| idiotsecant wrote:
| It's common to perform longevity testing at higher temperatures
| to simulate longer lifetimes, in account of nobody has decades
| of time to actually perform a 1x time test.
| HPsquared wrote:
| I wonder if "damp" modes of decay could still damage them
| though, which isn't captured in this style of testing. Like
| some wet chemical or biological process.
| jurgenburgen wrote:
| Mechanical decay would also damage them. I think it's
| assumed that the media will be stored in a place protected
| from humidity, chemicals and hammers.
| HPsquared wrote:
| Yes I suppose a strong casing can protect against all
| that, but not against temperature so that's the one thing
| they still need to test for.
| adrian_b wrote:
| Borosilicate glass is much more resistant to chemicals and
| shocks than ordinary glass, which is why it is used in
| glassware for laboratories and in also in the better
| cooking vessels.
|
| Except for hydrogen fluoride (a.k.a. hydrofluoric acid) and
| for hot and concentrated strong alkalies, biological or
| chemical agents will not have effects.
|
| The main danger is either breaking the glass or its
| crystallization at high temperatures. A HDD is also
| unlikely to survive a drop on a hard floor and high
| temperatures would demagnetize it much more easily than
| affecting a glass stab.
| homarp wrote:
| also at https://news.ycombinator.com/item?id=47065175
| wumms wrote:
| Current write speed (No read speed given): Blu-
| ray (1x) ~36 Mbit/s MS-Glass (single beam)
| ~25.6 Mbit/s MS-Glass (multi-beam) ~65.9 Mbit/s
|
| That's ~7-18 days per 120mm x 120mm medium (4.8TB). Glass prices
| stable for now. Also, the authors make no statement about
| horizontal vs. vertical storage.
| NitpickLawyer wrote:
| Thanks for digging this up. Every "scientists create new
| storage medium" is always a disappointment when you get to see
| the write speeds. This seems decent? At least in "raw" numbers
| there's nothing obviously making this useless. Let's hope they
| have a path to quick commercialisation and make it available.
| If there's any DC adoption will be the real test, I think.
| stackghost wrote:
| >This seems decent?
|
| Definitely. If it actually achieves those speeds it's
| perfectly reasonable for long-term/cold storage.
| Someone wrote:
| Depends somewhat on the read speed, too. Extreme example:
| if that is one bit per year, it doesn't matter that you can
| write stuff on it.
| dyauspitr wrote:
| I imagine if you can use lasers to etch at that speed,
| you can use them to read at similar speeds as well.
| po1nt wrote:
| First CDs would take hour and a half to write with a laser.
| Once engineers take over the tech, it will might get faster.
| wumms wrote:
| If they get the read speed up to a couple of GBit/s (~100x
| current max write speed), 4.8TB might be a good fit for 32k
| movies.
| bdbdbdb wrote:
| Of course there are people out there watching 32k movies.
|
| Was 4k not enough?
|
| Am I the only one who's still content with 720p?
| thegrim33 wrote:
| Write speed is probably the least important metric for people
| that are considering something like this. After everything
| with storage and longevity is taken care of, improving write
| speeds is a nice to have, but not the important part.
| Am4TIfIsER0ppos wrote:
| > No read speed given
|
| Write only medium!
| npodbielski wrote:
| At least it is safe for 10k years! And from everybody ever
| basically.
| adrian_b wrote:
| The reading is the done with a high-resolution video camera
| and the image is processed to extract the data.
|
| This can be easily done many times faster than the writing,
| which is why the article is focused on the progress that
| Microsoft has achieved in increasing the writing speed, in
| comparison with their prototypes from a few years ago. It is
| also easy to make separate readers that are much cheaper and
| smaller than the writers.
|
| The most important limitation of this device is the current
| very high cost of the lasers used for writing. Had they been
| cheaper, the writing speed could be increased by adding more
| lasers.
|
| Microsoft argues that if this kind of short-pulse lasers
| would be mass produced, they could become much cheaper, like
| it has happened with the many lasers that are used now
| everywhere in optical fiber communication and with optical
| discs.
|
| For now. this is a chicken-and-egg problem. This kind of
| optical storage cannot be converted into a commercial product
| because the lasers are too expensive and the lasers are too
| expensive because there is no high-volume market for them.
|
| Even the current level of performance would be enough for
| myself. If I could afford such a device, I would buy it
| instantly, to stop worrying about having to buy periodically
| new HDDs, to migrate my data from old HDDs and to buy
| periodically new tape drives, to migrate my data from tape
| formats that become obsolete.
| jmclnx wrote:
| The big question, is it patented to the point were no one can buy
| the burners and media ?
|
| Will it run on Linux ?
| misswaterfairy wrote:
| They're definitely pursuing patents...
|
| > The authors of the paper have filed several patents relating
| to the subject matter contained in this paper in the name of
| Microsoft Corporation.
|
| Page 12 of the paper:
| https://www.nature.com/articles/s41586-025-10042-w.pdf
|
| It's whether Microsoft will be fair and flexible licensing
| their patents to third-parties.
|
| Otherwise I'd suggest that if they keep it all to themselves
| and charge like a wounded bull, uptake would be quite limited.
|
| At least until the original patents expires, which might be the
| better strategic move for third-parties in light of a hostile
| Microsoft given how long this archival format is expected to
| last.
| HPsquared wrote:
| 20 year patent lifespan seems like nothing to the overall
| lifespan of this invention though.
| nandomrumber wrote:
| Or, next year I can buy ten for five dollars off AliExpress.
| vasco wrote:
| Yeah but then 1000 years from now nobody will have the right USB
| cable to read it.
|
| I think we should stick to proven solutions for millennia-robust
| information storage and paint it on walls inside pyramids.
| stackghost wrote:
| If Nanni could have engraved his shitpost about Ea-nasir's
| copper into multiple glass tablets, easy to distribute, that
| would last for 10000 years, he probably would have.
| userbinator wrote:
| You could "bootstrap" all the information required to produce
| the hardware to read this, by starting with human-readable
| instructions for the next step.
| dguest wrote:
| What language will humans be reading in 10,000 years?
| cute_boi wrote:
| With all these climate change and AI, will there be humans
| in 10k years?
| lofaszvanitt wrote:
| Yep, on top of trees, fighting for real estate.
| nubinetwork wrote:
| 20 years ago, I thought Esperanto might be a thing, but
| nope. :)
| adrian_b wrote:
| Esperanto has succeeded to become the best known of the
| artificially-constructed languages.
|
| Unfortunately, its creator had only modest knowledge of
| linguistics, so there are many features of Esperanto that
| can be considered as mistakes, and they have contributed
| to its little success.
|
| Some of the artificial languages that have been designed
| later are much better than Esperanto, but they have
| achieved even less notoriety than it.
|
| What has doomed all artificial languages, despite the
| fact that some of them would have been much better than
| English for international relations and for the
| publication of scientific and technical literature, has
| been the absolute dominance of USA over the entire world
| after the end of WWII.
|
| This unbalanced relationship between USA and everyone
| else has forced the use of English both in commercial
| relations and in the scientific and technical
| publications, excluding all alternatives and replacing
| not only any artificial languages, but also the European
| languages that previously had been more important than
| English, i.e. German and French.
| userbinator wrote:
| Pictures.
| Aardwolf wrote:
| You don't necessarily need the same hardware to read it, just
| like you can read a vinyl record optically without a needle
| dyauspitr wrote:
| Depends on what you etch on there. If it's binary
| representation of actual alphabets then sure. If it's a video
| file then without the software to decipher and manipulate the
| data, it would be pretty indecipherable. How to read an mp4
| is not part of the data itself.
| simicd wrote:
| At 4.8TB one could add a header section with the full code,
| instructions how to compile it etc. That would certainly
| help to reproduce it, assuming civilizations in 10k years
| still can decypher todays language.
| ksec wrote:
| >4.84TB in a single slab of glass, (the slabs are 12 cm x 12 cm
| and 0.2 cm thick).
|
| So a rough estimate, at the size of UMD, used in Playstation
| Portable, slightly smaller than the size of Mini Disc, it could
| store 1TB.
|
| I assume we could do double layer in the future for 2TB.
|
| For comparison that is roughly 1000x times the capacity of UMD. I
| would love to have this. Burn a few of these as backup and call
| it a day.
| Karliss wrote:
| It is already way beyond double layer. The 4.8TB is achieved
| using 301 layers.
| ksec wrote:
| There goes my hope of non-cloud backup. I was thinking 1TB
| doesn't quite make it. Or at least I need a dozens of these.
| adrian_b wrote:
| The capacity per device is irrelevant.
|
| What matters is the capacity per volume, per mass and per
| dollar.
|
| The capacities per volume and per mass for these glass
| slabs are already very competitive. They are about the same
| as for the best tape cartridges currently available. The
| capacity per volume is about twice better than for the best
| HDDs, and the capacity per mass is much better than that,
| because HDDs are very heavy.
|
| If such optical storage had not been so expensive as it is
| for now, it would have already been much better than any
| cloud storage. The slow writing speed is similar to that of
| file downloading or uploading over the Internet. Reading
| can be done much faster than writing, because it uses
| ordinary lasers and a video camera, not the very expensive
| femtosecond-pulse lasers used for writing.
| rarisma wrote:
| I swear this happens at least once a year.
|
| Wheres my futuristic storage guys?
| winrid wrote:
| in your hands :)
| lofaszvanitt wrote:
| If you watch the movie johnny mnemonic, they throw around data
| cubes the size of a stamp. Modern nvme ssds weigh around 4
| grams per TB. So we've already achieved scifi movie parity :D.
| The only problem is the price.
| quinndexter wrote:
| I'm a little bit old. When I ordered my first M.2 drive, I had
| never seen one IRL. I'd assumed about RAM-stick size. Nope!
| Thumb-sized! The future is amazing! So... give it enough time
| and eventually the mundane will scratch that itch, I guess?
| idiotsecant wrote:
| Glass is one of the more stable things we can make. This seems
| pretty good! I don't have an application that requires ten
| thousand years of storage but I'm sure someone out there does!
| tbrownaw wrote:
| > _I don 't have an application that requires ten thousand
| years of storage but I'm sure someone out there does!_
|
| A) record (a representative cross-section of) "everything" and
| leave multiple copies where future archeologists might find it.
| To avoid things like how present-day archaeologists apparently
| have holes in the kinds of things they can find, due to
| different social classes not leaving equally-robust trails.
|
| B) this is "at least as long as I could possibly care about"
| storage. If I need to retain say financial records for seven
| years, and then later the government retroactively increases
| that to 20 years, there's no need to re-archive it all onto new
| media.
| TacticalCoder wrote:
| > B) this is "at least as long as I could possibly care
| about" storage. If I need to retain say financial records for
| seven years, and then later the government retroactively
| increases that to 20 years, there's no need to re-archive it
| all onto new media.
|
| In many countries this "maximum (6 or) 7 years" for financial
| records is only if the local IRS decides that you're not
| potentially committing fraud. If they decide you've
| _potentially_ committed fraud at any time in the past, there
| 's no limit as to how far they can go. Even in the US stuff
| like (some of the) funds stolen by the Enron scam have been
| successfully clawed back more than two decades after the
| fact.
|
| At least that's the case in several EU countries: there's
| literally no limit if the country's IRS equivalent decides
| you're potentially committing fraud (or if you did in the
| past).
|
| Which is insane and totally arbitrary but that's how it is.
|
| In addition to that under a great many KYC/AML excuses, there
| are banks out there that shall have zero issue asking you to
| justify the "source of funds" and at times I've had to
| provide info dating from _way_ more than seven years in the
| past. I 've heard --and I'm not shitting you-- from someone
| proving he bought for about 5 K EUR of something that went up
| more than 100x (think Bitcoin or some exceptionally
| successful stock), that his bank answered something like: _"
| OK, but now that you've proven you actually made 100x, prove
| us the source of the 5 K EUR in 2013!"_.
|
| That's what happens to a society when you give too much power
| to petty people.
|
| There are literally collaborationists out there that are
| going to fill SARs (Suspicious Activity Reports) when someone
| can prove he turned 5 K into 500 K not on the 500 K (which
| are impossible to dispute) but on the 5 K that were used in
| the first place. That's how jealous and incompetent some
| people are in this world.
|
| Things became so bad that I now have a Git versioned repo
| (and backups everywhere) where I keep track of, among other,
| every single wire transfer above 10 K EUR. I've got stuff
| dating back to 2001 when I bought my first apartment etc.
|
| Don't underestimate how pathetic and bitter some of the
| people you'll have to deal with (be it from your local IRS or
| a bank) are going to be.
| lencastre wrote:
| given a long enough period, glass is a fluid, i.e. viscosity
| sandbach wrote:
| Not at low temperatures.
|
| https://www.scientificamerican.com/article/fact-fiction-
| glas...
| dyauspitr wrote:
| You can get a 6mmx6mmx1.2mm pure industrial diamond sheet for
| about $1000 [1]. That should be able to hold around 300 GB
| with this method and would last practically forever.
|
| [1] https://e6cvd.com/us/material/single-
| crystalline.html?utm_so...
| userbinator wrote:
| ?utm_source=chatgpt.com
|
| Seriously?
| idiotsecant wrote:
| This is mostly a myth based on some medieval glass panels
| that had structurally wider bases. This material is going to
| take until the heat death of the universe to deform 1mm at
| room temperature. I'm sure it'll be fine for way longer than
| it will take for the data to fail in a different way.
| gigel82 wrote:
| I have read a variation of this headline once every 2 years since
| the early 2000s, yet never seen it turn into something real (that
| a consumer / enterprise can buy).
| pyrex2026 wrote:
| is pyrex a public stock
| wowczarek wrote:
| Corning is. As to pyrex, that depends:
| https://news.ycombinator.com/item?id=45310995
| adrianN wrote:
| How much cheaper is it compared to those orbs you can get from
| the Long Now Foundation?
| canterburry wrote:
| Very impressive new format. 10,000 years...wow. That's great.
|
| Now, can someone please help me get some data of this Iomega
| ZIPdrive disc?
| micw wrote:
| 10k years ... Or until it's dropped...
| avadodin wrote:
| You inserted it at the wrong angle, lol.
| Alifatisk wrote:
| I swear I've read similar headline multiple times for the past
| decade. This can't be new.
|
| I thought I was experiencing some Mandela affect, had to Bing it.
| This is from 2022 https://www.microsoft.com/en-
| us/research/video/project-silic...
| wumms wrote:
| They mention it in the article:
|
| > Microsoft began to build on their work in 2017. Although
| Kazansky's approach maximizes durability and the density of
| data, in the latest work, Microsoft has gone for practicality.
| They explore a method that enables data to be written faster
| and decoded more reliably than did Project Silica's previous
| iterations, says Black, and it uses cheaper borosilicate glass,
| rather than harder-to-make fused silica.
|
| Following your link, I found a prototype of the media storage
| system (2023) with just 2828 views:
| https://www.youtube.com/watch?v=xnK-uB4OsgU
| adrian_b wrote:
| They report now significant progresses in writing speed, data
| density, and also in cost, in comparison with their prototypes
| from a few years ago.
|
| Unlike before, at the current performance this kind of optical
| storage could have actually been used in practice, had the
| writing lasers not been so expensive.
|
| For a given amount of data, such glass slabs, which have the
| size of CD cases, would occupy a volume of about half of that
| required for the highest-capacity HDDs and about the same as
| tape cartriges.
|
| The writing speed is similar with the file downloading speed
| over the Internet from most sources that throttle their
| connections, instead of allowing full speed.
| Razengan wrote:
| DNA?
|
| Something I've always wondered.. could things like tree rings, or
| birdsong and other mating rituals (the species can survive only
| by replicating the sequences) be used for "data storage" by some
| hypothetical intelligences that think on scales of thousands of
| years?
| alfiedotwtf wrote:
| I don't trust my Kodak Gold CDs that advertised 40 years, let
| alone humans to not self annihilate within the next 400 years.
|
| In fact, look what we're doing right now with all our past's
| relics!
| theAdminWave wrote:
| LOL I've done holographic data storage in borosilicate glass
| using fs laser pulses for my masters thesis in physics more than
| a decade ago and guess what, this is not going anywhere. The
| claims are all wildly exaggerated also. Lots of buzzwords micro
| nano plasma explosions but the truth is hidden in the details:
| needs specialist hardware... Yeah like a 50.000 USD femto second
| laser setup that needs an entire basement and you wearing ski
| googles at all times to not get blind type of specialist
| hardware. Guess we're all gonna put that in our living rooms,
| won't we?
|
| And the storage density is limited by all kinds of effects that I
| won't even get into it but you can roughly assume its at best
| half or even less of that and then it starts becoming much less
| impressive.
|
| Yes you can microwave a slab of glass or go diving with it and it
| will still be intact but unless we make machines that read and
| store data much more easily, like significantly absurdly more
| easily, this is the biggest pipe dream of them all.
|
| Cool tech though :)
| baxtr wrote:
| Agreed. On the other hand: didn't any cool tech start as a
| overpriced, oversized version of its later breakthrough
| product?
| stanac wrote:
| It doesn't have to be consumer hardware to be economically
| viable. I can imagine something like this replacing or
| complementing tape storage at data centers. We already have
| hard drives filled with gas for dust-proofing. For archival
| storage it does not have to be fast (in terms of latency) it
| just needs to be reliable with high data density.
|
| Hard drives where the size of a car decades ago, we could now
| have archival storage of the same physical size that can hold
| petabytes (just guessing, didn't do the actual math).
| fragmede wrote:
| The question is how much are readers. If I have to take my
| data to my local strip mall to write my datacube, but this
| data cube can be read with a much cheaper reader that I can
| reasonably believe will be available in 40 years, I could see
| that as being viable.
| jmpman wrote:
| I believe the helium in the hard drives allows the heads to
| fly closer to the platter due to the different Reynolds
| number, not dust.
| zem wrote:
| i could swear i've read about prototypes of glass storage all
| the way back in the 80s. been waiting patiently for it to
| become an actual product.
| inemesitaffia wrote:
| NSA Utah Data Centre
| zkmon wrote:
| Data loses its usefulness and relevance with time. Unless it's
| updatable, just archiving is not going to be useful. Just like
| how we don't find some 2000 years old writings any useful now,
| except for museum storage.
| b0rtb0rt wrote:
| pretty sure there are a few 2000 year old texts that people are
| still reading today
| bubblewand wrote:
| Pretty sure the oldest stuff I've read was around 4,500 years
| old (in translation, but still). Volume I of Lichtheim's
| Ancient Egyptian Literature.
|
| There are a couple incomplete tales in that 3-volume work and
| I really wish they'd had a more stable storage medium,
| because now I'm stuck with accidental millennia-old
| cliffhangers.
| bossyTeacher wrote:
| The Bible doesn't count
| adrian_b wrote:
| _You_ do not find 2000 years old writings useful, presumably
| because you have not read any, or at most some bad
| translations.
|
| _I_ find old writings extremely useful, including many of
| those that are 2500 years old or even older. The really useful
| old writings are those in bilingual editions, which allow you
| to skim quickly in English through passages in which you are
| not interested at the moment, but then allow you to read the
| original language to see what the author really meant. Most
| English translations are very bad when you have a scientific
| interest in the meaning of the text, because the translators
| are usually completely ignorant about mathematics, physics,
| chemistry, mineralogy, biology, technologies and so on, so they
| fail to translate correctly anything that includes words with
| special meanings in those domains.
|
| Unfortunately, nowadays fewer people read old writings, which
| is obvious in the deluge of publications that claim that
| various new things have been discovered or invented, but those
| supposedly "new" things are old things that were well known
| decades or even centuries earlier. Worse, the "new" things
| usually were not only known previously, but previously better
| alternatives were already known which are ignored by the
| "inventors" of the "new" things.
|
| It is well known, or it should be well known, that the space of
| solutions for many technical problems contains a finite number
| of alternatives, among which at a given time one is the best.
| However, which is the best changes in time, depending on the
| evolution of technologies in other domains. Over a long enough
| time interval, it is frequent that the optimum solution cycles
| through all alternatives, going back to a solution that had
| been preferred a long time before, but which had then been
| abandoned for some time.
|
| 2500 years old writings do not provide detailed solutions for
| present problems, e.g. how to design the schematics for the
| power supply of a laser. Nevertheless, they may still provide
| logical thinking frameworks that are as valid today as they
| were millennia ago, and they also show the original form of
| various concepts that have evolved through time until their
| current form. Understanding the reasons why certain things have
| evolved through history is essential for understanding in which
| direction they should be changed today.
|
| Those who believe that we are today better in all domains in
| comparison with our ancestors from 2000 years ago, are
| delusional. I consider a few of the laws that were valid in
| certain ancient societies as greatly superior to the laws that
| have replaced them in most modern countries.
|
| (An example is that there are now a lot of countries that claim
| to be "republics", not to mention a certain US party. All such
| claims are lies, because today there exists no republic on
| Earth. The principle on which the Roman Republic was grounded,
| and which distinguished it from other forms of government, was
| that, with temporary exceptions for emergencies like wars,
| plagues or natural calamities, no important public executive
| function should be occupied by a single human, but only by 2 or
| more humans with equal power, so that the abuse attempts of 1
| of them would be stopped by the others. There are also examples
| in more rational and more efficient penal systems, which did
| not include stupid punishment methods, like prison, but where
| any non-irreversible kind of damage (e.g. theft, fraud, non-
| permanent bodily harm) was punished by a fine, but not by a
| fine equal to the damage, but by a fine that was a big multiple
| of the damage from double to 10 times, to discourage such
| acts.)
| thaumasiotes wrote:
| We've had data storage that lasts for millennia for several
| thousand years already. The invention of millennia-long storage
| more or less coincided with the invention of writing.
|
| There isn't really a benefit. Our durably-stored several-
| thousand-year-old records suffer from various problems:
|
| - They're hard to understand.
|
| - They tend not to be relevant to much.
|
| - Most of them have gotten lost. They're not gone, but it would
| be extremely expensive to find them.
|
| Interestingly, these are the same problems that occur with stored
| data of much more recent vintage. But they get worse and worse
| over time, and the fact that the storage medium itself doesn't
| degrade does nothing to help. It tends to make those usability
| problems _worse_ by giving people a false sense of security that
| the data is still there, until the cost of recovering it becomes
| too great and for practical purposes it isn 't there anymore.
|
| If something matters, it will be stored on ephemeral media and
| recopied over time onto more ephemeral media.
| hilbert42 wrote:
| It's good to know people are finally working on long-term
| nonvolatile storage. To date, just about every storage syatem
| we've developed has had a storter lifespan than the previous one,
| NAND versus magnetic storage for instance.
|
| The idea isn't new of course, just think back to Dave removing
| HAL's glass/crystal memory modules in 2001. Clarke/Kubrick were
| thinking along those lines in the 1960s.
| jrflowers wrote:
| I like that they call it a "deployable archival system" that kind
| of sounds like they've invented a laser that can read/write this
| stuff that can last for 10,000 years
___________________________________________________________________
(page generated 2026-02-22 16:01 UTC)