Imagine a storage medium so durable you could literally bake it in a kitchen oven at five hundred degrees, toss it into boiling water, or scrub it with steel wool, and the data inside would remain perfectly intact. We are talking about a physical format that doesn't just outlast a human lifetime, but potentially outlasts the Roman Empire several times over. We have this near-perfect archival solution sitting in research labs, yet here we are in April twenty twenty-six, and the world is still largely backing up its most critical data on thin strips of plastic coated in magnetic dust.
It is the great paradox of the digital age, Corn. I am Herman Poppleberry, and today we’re diving into a prompt from Daniel about Microsoft’s Project Silica and the quest for the "eternal" storage medium. Daniel’s asking a really pointed question: can quartz glass storage actually scale up before the sheer volume of global data makes it obsolete? We are looking at a world that is projected to hit over one hundred and eighty zettabytes of data produced annually, and our current methods of keeping that data "cold" are starting to look incredibly fragile.
One hundred and eighty zettabytes. I can barely conceptualize a terabyte when it’s filled with my blurry vacation photos, let alone a zettabyte. But before we get into the existential dread of data deluges, we should probably mention that today’s exploration is actually being powered by Google Gemini Three Flash. It’s the model writing our script today, which is fitting since we’re talking about the literal foundations of where models like this might be stored in the distant future.
It’s a perfect match. And to set the stage for Daniel’s question, we have to understand what Project Silica actually is. Most people hear "glass storage" and they think of a CD or a Blu-ray, but this is fundamentally different. A Blu-ray stores data on the surface, or in a couple of very thin layers just beneath it. Project Silica uses a femtosecond laser—that’s a laser that pulses at one quadrillionth of a second—to reach inside a thick slab of fused silica glass. It creates these tiny 3D structures called voxels.
Voxels. Like pixels, but with volume.
Precisely. And because the glass is thick, you can stack hundreds of these layers on top of each other. You’re not just writing on the glass; you’re writing inside the glass. It’s like the difference between writing a note on a piece of paper versus engraving a message inside a solid crystal. Once those voxels are there, they are a permanent physical change in the molecular structure of the quartz. They don't "fade" like the dyes in a recordable DVD, and they aren't susceptible to the magnetic interference that can wipe a hard drive or a tape.
So, it’s the ultimate "set it and forget it" technology. But Daniel’s point about the timeline is what really gets me. We’ve been hearing about this for a while. I remember the big splash back in twenty nineteen when Microsoft stored the original Superman movie on a piece of glass the size of a drink coaster. That was seven years ago. If it’s so indestructible and amazing, why am I not buying a glass drive for my server rack yet? Is this just one of those "forever five years away" technologies?
That’s the tension Daniel is tapping into. In twenty twenty-three, we saw another big milestone when they partnered with Warner Brothers to archive The Wizard of Oz. That was a huge proof of concept for the film industry because, believe it or not, original film stock is terrifyingly fragile. But as of right now, in early twenty twenty-six, it still hasn't hit wide commercial availability. The bottleneck isn't the glass itself—sand is everywhere, and quartz is cheap—it’s the physics of the write speed and the complexity of the reading system.
Right, because if I remember correctly, the Superman demo was only about seventy-five gigabytes. In an era where a single high-end game install is a hundred gigs and enterprise data is measured in exabytes, seventy-five gigs on a coaster feels... well, it feels like a floppy disk.
It does, but the density is improving. The latest specs we’re seeing for twenty twenty-six show a standard twelve-centimeter glass platter holding about four point eight terabytes. Now, that’s a massive jump from seventy-five gigs, but compare it to an LTO-Ten tape cartridge which holds eighteen terabytes natively. Tape is still the king of capacity, and it’s a mature, moving target. Every time glass storage gets better, tape gets better too.
And tape is cheap. But tape has that "bit rot" problem you’re always complaining about, doesn't it? You have to move it to a new tape every ten years or so just to make sure the plastic hasn't degraded or the drive hardware hasn't become a museum piece.
That is the "migration cycle," and it is the hidden tax on all modern data. If you’re a company like Google or Microsoft, you are spending millions of dollars and a massive amount of energy just moving data from old tapes to new tapes. It’s like having a library where you have to photocopy every book every decade or the ink will disappear. Project Silica’s pitch isn’t just that it holds a lot; it’s that once you write it, you never have to move it again. For ten thousand years. That is a zero-marginal-cost archive.
Okay, let’s dig into the "how" for a second, because I think the mechanics explain why this is taking so long to commercialize. You mentioned the femtosecond laser. That sounds expensive. Is it a situation where the "paper" is cheap but the "pen" costs a million dollars?
That is exactly the issue. A femtosecond laser is a highly specialized, sensitive piece of equipment. You can't just stick one in a consumer-grade external drive. These lasers work by creating "nanostructures" in the glass that change how light passes through it. To read the data back, you don't use a standard laser like a CD player. You use a computer-controlled microscope with polarized light. The microscope looks at the glass, sees how the light is being polarized by those 3D voxels, and uses machine learning to decode those patterns back into bits.
So the "player" is basically a high-tech lab microscope? That doesn't exactly sound like something that scales to a massive data center library with robots flying around grabbing platters.
Well, Microsoft is working on that part too. They are building these "jukeboxes" that use AI-driven computer vision to find and read the platters. But you hit on the write speed, and that’s the real kicker. For a long time, writing to glass was incredibly slow—like, one megabyte per second slow. If you’re trying to backup a petabyte of data at one megabyte per second, you’re going to be waiting until the next century just to finish the upload.
I’ve had internet connections like that. It’s not a fun way to live. But I heard there was a breakthrough recently, right? Something about the type of glass they're using?
Yes, this just came out in February of twenty twenty-six. Microsoft Research announced they’ve shifted some of their focus from pure quartz to a specific type of borosilicate glass, which is basically a high-tech version of Pyrex.
So we’re storing the world’s secrets in baking dishes now?
In a way, yes! The reason is fascinating. Pure quartz is very hard and requires a lot of laser energy to modify. Borosilicate is slightly more "malleable" at a molecular level when hit by the laser. This change has allowed them to significantly increase the write speeds. They’re still not quite at the speed of writing to a modern hard drive, but they are closing the gap with tape. They did sacrifice a little bit of the theoretical maximum density, but in exchange, they made the technology much more "manufacturable."
It’s the classic engineering trade-off. You can have the perfect theoretical material that takes forever to use, or the "good enough" material that you can actually ship. But even with the borosilicate shift, we’re still looking at twenty twenty-seven for actual commercial integration into Azure data centers. Daniel’s question remains: is twenty twenty-seven too late? If data volumes are doubling every few years, will a five-terabyte platter be a joke by the time it’s ready?
It’s a valid concern. If you look at the IDC forecasts, the sheer volume of data is outstripping our ability to manufacture storage media—period. We aren't making enough hard drives or tapes to keep up with the world’s cameras, sensors, and AI training logs. But Microsoft’s counter-argument is about "volumetric density." Think about a tape library. It needs climate control because tape hates humidity and heat. It needs a lot of physical space for the robotic arms and the cartridges, which are relatively bulky.
And those glass platters are tiny. They’re what, two millimeters thick?
You can pack them like thin slides in a rack. Because they don't need power to maintain the data and they don't care about the temperature, you can turn off the air conditioning in the archive room. You can pack ten times more data per square foot of data center floor space compared to tape. So even if an individual platter holds less than a tape cartridge, the entire "library" might actually be more dense and significantly cheaper to operate over a fifty-year span.
I like the idea of a data center that’s just a dark, dusty room full of glass slides that nobody has to touch for a century. It feels very "Indiana Jones." But let's talk about the competition. Daniel mentioned optical archiving struggling to keep pace. We’ve had things like M-DISC for a while—those "thousand-year" DVDs. Why didn't those win?
M-DISC was a great step, but it was fundamentally limited by the DVD and Blu-ray form factor. You’re still stuck with about a hundred gigabytes on a triple-layer Blu-ray. For a consumer, that’s fine for wedding photos. For a company like Warner Brothers or a national archive, a hundred gigs is nothing. You’d need millions of discs. Project Silica is aiming for the petabyte-to-exabyte scale. It’s also about the "read" side. Blu-ray players have moving parts, lasers that burn out, and very specific hardware requirements. The Silica reading system is basically just a camera and a light. As long as we have the ability to build a magnifying lens and a digital sensor, we can read that glass.
That’s a really important point for the "Digital Dark Age" problem. If I find a floppy disk from forty years ago, I don't just need the disk to be intact; I need a working floppy drive and a computer that can talk to that drive’s specific controller. With glass, the data is physically there as a 3D structure. Even if the "software" is lost, you could theoretically reconstruct the data just by looking at the glass with a powerful enough microscope and working out the patterns from scratch. It’s much more "future-proof" in a fundamental way.
It really is. It’s the closest thing we have to the Rosetta Stone for the digital era. But let's look at the practical barriers for a second, because this is where Daniel’s "will it be too small" question gets really pointed. Right now, the Global Music Vault in Svalbard is using this tech. They are storing master recordings of "culturally significant" music. That’s a perfect niche use case: low volume, high importance, extreme longevity. But that’s not the same as being the backbone of the cloud.
Right, the "Global Music Vault" sounds like a prestige project. It’s a great headline. But the real money—and the real data problem—is in things like medical records, financial logs, and legal archives. Stuff that has to be kept for seventy years by law, but is almost never accessed. That’s the "cold storage" market. Is Microsoft actually going to convince a major hospital system to trust their records to glass platters instead of the tried-and-true IBM tape libraries?
That is the multi-billion dollar question. And it comes down to the ecosystem. To replace tape, you don't just need the medium; you need the reliability. Tape has been the standard for decades. We know exactly how it fails. We know how to fix it. Project Silica is still "the new kid" in a world that is incredibly risk-averse. If a library robot drops a glass platter, does it shatter? Microsoft says no, these are chemically strengthened, but you can bet an IT director is going to worry about it.
I can see the "oops" moment now. "Sorry, we lost the last ten years of insurance claims because the robot had a glitch and turned the archive into confetti." Though, to be fair, I’ve seen what happens to a tape when it gets "chewed" by a bad drive. It’s not pretty either.
Fair point. And Microsoft has actually done "abuse testing" where they dropped them, baked them, and even put them in microwaves. The glass is incredibly tough. But the real hurdle is the "Write-Once, Read-Many" or WORM nature of it. You cannot erase a glass platter. You cannot "update" a file. If you make a mistake, that mistake is etched in quartz for ten thousand years.
That sounds like a nightmare for my typos, but for a legal archive, that’s actually a feature, right? Immutability is a huge deal for compliance.
It’s a feature for cold storage, but it means your data management software has to be perfect. You have to be absolutely sure that what you are sending to the "glass printer" is exactly what you want to keep forever. There is no "undo" button.
Let's talk about the cost, because Daniel mentioned the economics. Sand is cheap, but the process isn't. If I’m a CFO, I’m looking at the cost per gigabyte. Tape is incredibly cheap—sometimes less than a cent per gigabyte. Where does glass sit on that spectrum?
Right now, it’s much higher because the lasers are expensive and the throughput is low. But Microsoft is betting on the "TCO"—the Total Cost of Ownership. If you factor in the cost of electricity for cooling a tape library, the cost of the "migration" every ten years, and the cost of the physical space, they argue that glass eventually becomes cheaper. But that "eventually" might be twenty years down the line. Most corporate budgets don't look twenty years ahead; they look at the next quarter.
That’s the "quarterly earnings" trap. It’s hard to justify a massive capital expenditure today to save money in twenty forty-six. But I wonder if the sustainability angle might be the tipping point. We’re seeing a lot of pressure on big tech to reduce the energy footprint of data centers. If you can move thirty percent of your data to "passive" glass storage that uses zero watts to maintain, that’s a massive win for your carbon footprint.
You nailed it. That is a huge part of the marketing for Project Silica. They are calling it "sustainable by design." Traditional storage is a constant energy drain. Even a hard drive sitting idle in a rack is drawing a tiny bit of power, and it’s generating heat that needs to be pumped out by an air conditioner. Glass is inert. It’s just a shelf of rocks. From a "Green IT" perspective, it’s the holy grail.
So, to Daniel’s question about whether it will be "too small" by the time it arrives—I’m looking at these capacity numbers again. Four point eight terabytes per platter. If the growth of data is exponential, and we’re hitting zettabytes, four terabytes really does feel like a drop in the bucket. But I guess if you can fit ten thousand of those drops in a single rack, it starts to add up.
It does. And remember, we’re talking about "cold" data. Most of the zettabytes being created are "hot" or "warm" data—TikTok videos that are watched a million times today and forgotten tomorrow, or real-time sensor data from self-driving cars. That stuff stays on SSDs or hard drives. Project Silica is for the "frozen" data. The stuff that we might need in fifty years, but we definitely don't need today. When you look at it that way, the capacity requirements are slightly different. You’re not trying to store the whole internet; you’re trying to store the "heritage" of the internet.
That’s a good distinction. But even "heritage" data is growing. Think about high-resolution medical imaging. Every year, the resolution of an MRI or a CT scan gets higher. The files get bigger. If you want to keep a patient’s lifetime medical history on glass, you’re going to need more than a few gigabytes.
True. And that’s where the competition from things like DNA storage comes in. We’ve talked about that before—the idea of encoding data into synthetic DNA strands. DNA density is light-years beyond glass. You could theoretically store the entire world’s data in a couple of buckets of DNA.
Yeah, but then you have to deal with "sequencing" the DNA to read it back, which is its own slow, expensive nightmare. And DNA is biological—it can degrade if it’s not kept in very specific conditions. Glass feels much more "industrial strength" compared to a bucket of synthetic goo.
I agree. Glass is the "bridge" technology. It’s more dense than tape, more durable than anything we’ve ever made, and it’s based on physics we understand. DNA is still very much in the "science fiction" phase for enterprise use.
So let’s look at the "Takeaways" for the people listening who are actually managing this stuff. If you’re an enterprise IT person today, and you’re staring at a growing mountain of data, what do you actually do with this information about Project Silica?
The first takeaway is: don't throw away your tape drives yet. For the next three to five years, tape is still your only pragmatic choice for high-volume cold storage. LTO-Ten is here, LTO-Eleven is on the roadmap, and the economics are still unbeatable for most companies. If you need to store petabytes today, you’re buying tape.
Right. Glass is currently a "watch and wait" technology. But you should be watching it, especially those partnerships like the Warner Brothers one or the Global Music Vault. Those are the "canaries in the coal mine." If those projects start scaling up, it means the "write speed" problem is being solved.
And the second takeaway is to re-evaluate your data retention policy. A lot of companies are "hoarding" data on expensive "warm" storage—like spinning hard drives—because they’re afraid of the friction of moving it to tape. If Project Silica becomes a reality in Azure, it might change the math on what you decide to keep. You might move from "keep for seven years" to "keep forever" because the cost of "forever" on glass becomes so low.
It changes the philosophy of archiving. Right now, archiving is a "burden." It’s a chore we have to do. If glass works, archiving becomes an "asset." You’re building a permanent, indestructible library of your company’s history. That’s a huge shift in mindset.
It really is. And for the individual, while we won't have "glass burners" in our laptops anytime soon, we might see "permanent backup" services as a tier in consumer cloud storage. Imagine paying a one-time fee to have your family photos etched into a glass platter in a Microsoft vault, knowing that even if the company goes bankrupt or the internet changes forever, that physical object exists and can be read in a hundred years.
I’d pay for that. I mean, I’ve already lost photos from ten years ago because a cloud service changed its terms or I forgot a password to an old account. The idea of a physical, permanent "master copy" that isn't dependent on a subscription model is really appealing.
It’s the return of the physical artifact in a digital world. We spent thirty years trying to get away from physical media—no more CDs, no more DVDs, everything in the "cloud." And now we’re realizing that the "cloud" is just someone else’s hard drive, and those hard drives are surprisingly fragile. Project Silica is the cloud finally getting a "hardcover" version.
I like that. "The Cloud: Now in Hardcover." We should pitch that to Microsoft’s marketing team. But seriously, Daniel’s question about the "Data Deluge" is the one that stays with me. If we are producing data at a rate that outstrips our ability to build the machines to store it, we’re effectively "deleting" our own history in real-time. We’re only keeping what we can afford to store.
That is the "Digital Dark Age" in a nutshell. We are the most documented generation in history, but we might leave the least amount of evidence behind because our "paper" is so temporary. If Project Silica can bridge that gap—if it can provide a way to store the "essentials" of our civilization in a way that doesn't require constant maintenance—it might be the most important infrastructure project of the twenty-first century. Even if the platters are "small" by tomorrow’s standards, the fact that they are "permanent" is what matters.
It’s the difference between a high-speed stream and a stone carving. The stream carries more water, but the carving is what tells the story a thousand years later.
Well said, Corn. I think we’ve thoroughly explored the "quartz" of the matter here. Daniel, thanks for the prompt—it really highlights how the most "futuristic" solutions often involve going back to the most basic materials, like sand and light.
And thanks to our producer, Hilbert Flumingtop, for keeping the wheels turning. Also, a big thanks to Modal for providing the GPU credits that power the generation of this show.
This has been My Weird Prompts. If you find yourself wanting to dive deeper into the technical specs of femtosecond laser etching—or if you just want to see those Superman glass coasters—check out the links on our website at myweirdprompts dot com.
And if you're enjoying the show, leave us a review on Apple Podcasts or Spotify. It helps other curious humans—and sloths and donkeys—find us in the vast, ever-expanding zettabyte-sized ocean of the internet.
We're also on Telegram if you want to get notified the second a new episode drops. Just search for My Weird Prompts.
Catch you in the next one. Hopefully, by then, Herman will have figured out how to etch his grocery list into a Pyrex dish.
It’s for the sake of history, Corn! Future archaeologists need to know I was out of oat milk. Goodbye, everyone.
See ya.
