On September 29, 2026, researchers at the University of Shanghai for Science and Technology published a result in the journal Nature Photonics that sounds, at first, like a unit-conversion mistake. They reported a storage technique in which a single three-dimensional point on an optical disc can distinguish 1,024 states rather than the two states of ordinary binary storage, giving that one point an information capacity of 10 bits.
Based on their laboratory measurements, they estimate a disc the size of a standard DVD (Digital Video Disc, the familiar 12-centimeter optical platter) could theoretically hold about 0.4 petabytes. A petabyte (PB) is roughly one million gigabytes (GB) — for scale, that is vastly more than the 25 GB held by a common single-layer Blu-ray disc. They put the data-read reliability at more than 99.99 percent.
The word "theoretically" is doing real work there, and a serious reading of the paper keeps it front and center. But the underlying idea is genuinely clever, and it addresses a problem that is becoming urgent in the age of artificial intelligence (AI).
Why the World Needs a New Place to Put Data
Start with the demand side.
Training data for AI, scientific records, medical images, and high-resolution video all pile up fast — and much of it is "cold data": information that is no longer used every day but still has to be kept for years. Magnetic hard drives and solid-state chips are fast but cost power and money to run forever, and the densest options are concentrated in giant data centers. Optical discs, the descendants of the CD and DVD, are cheap to make and stable for long-term archival, but they have not stored enough to compete.
So the storage world has been hunting for a way to keep the advantages of light-based discs — low cost, long life, simple to stack and shelve — while closing the enormous capacity gap. This paper is a proposal for how to do that.
From a Bungalow to a Skyscraper
To understand the trick, recall how digital storage works.
A bit is the smallest unit of information, and in ordinary binary storage it has two possible states — 0 or 1. Every record point on a conventional disc is essentially a light switch: on or off, pit or no pit. To store more, engineers have historically done one of two things: make each point smaller, so more fit on a disc, or stack more layers on top of one another in three dimensions. Both run into hard physical limits around precision and material.
The Shanghai team opened a third route. Instead of asking a point to be only 0 or 1, they made one three-dimensional record point able to take 1,024 distinguishable states. Because two to the power of ten is 1,024, that single point encodes ten bits at once. The team's own analogy is architectural: a conventional point is a one-story bungalow, holding one unit of information on its footprint; the new point is a tall tower built on the same spot, packing far more into the same area.
That reframes the whole problem. Capacity no longer has to grow only by shrinking points or piling on layers; it can also grow by teaching each individual point to hold more. The states are read out as different levels of light intensity — closer in spirit to the many gray shades in an image than to a black-and-white switch. An 8K ultra-high-definition image contains roughly 33 million pixels, and an 8-bit grayscale pixel already has 256 possible tones; the new points go far further, to 1,024 levels.
The Physics Underneath
The scientific title is dense, but its key ingredients can be explained plainly.
The researchers work with a light-sensitive material — a photopolymer disc based on polyethylene glycol diacrylate, a polymer that changes when light strikes it. They use a laser to create tiny carbonized polymer dots inside the material and then exploit what the paper calls phonon-assisted upconversion luminescence. In everyday terms: vibrational energy in the material (phonons are the quantum units of that vibration) helps the tiny dots absorb lower-energy light and emit higher-energy light back, and the brightness of that emitted light can be controlled and read in many fine steps. Those brightness steps are the 1,024 states.
The dots are written and read in three dimensions inside the volume of the disc, which is how one platter can hold a stack of information rather than only a surface layer.
Why One Beam Instead of Two Matters
The second innovation is about the hardware needed to use such a disc — and it may be the more practical one.
Older high-density "super-resolution" optical-storage schemes generally needed two light beams working together: one to do the writing or reading and a second to assist, for example by switching the material's behavior on and off around the first beam. Aligning two beams precisely and keeping them coordinated is exactly the kind of finicky engineering that makes a device expensive, slow, and hard to shrink.
The team says its method performs both writing and reading with a single monochromatic beam. Removing the second beam removes the need for delicate dual-beam alignment and coordination, which should make the system faster and far easier to integrate into real equipment. They pair that with high-speed scanning for writing and "wide-field parallel" reading — grabbing information from many points across an area at the same time rather than strictly one at a time.
Where the 0.4 Petabyte Figure Comes From
It is worth tracing the headline number rather than treating it as a product specification.
The 0.4 PB figure is an estimate derived from the laboratory demonstrations: the 1,024-state points, the three-dimensional arrangement, and the spacing between the layers (the current layer separation is five micrometers, a micrometer being one-millionth of a meter). Multiplied out across a DVD-sized disc, the per-point capacity yields that theoretical total. It is a calculation about what the demonstrated physics could allow, not a measurement of a finished 0.4 PB disc sitting on a shelf.
The authors are explicit that petabyte capacity is not the only measure that matters. Turning the idea into a product would also require the right read-and-write speed, a manageable system design, and the ability to build and integrate real devices at scale.
The More Distant Claim: Store and Compute in the Same Spot
The team points to a farther-reaching possibility that explains some of the excitement.
Because a ten-bit point has room to spare after holding the usual eight bits of an image's grayscale value, those extra two bits could, in principle, provide added encoding space for in-place optical AI processing. If the same material point could both store information and help perform calculations on it, the two functions could merge into what the researchers describe as "zero-distance" integration of storage and computing — computing where the data already sits, rather than constantly moving it. They have not demonstrated a working computer built this way; it is a stated direction, not a result.
Their longer-term projections are equally forward-looking. If the layer spacing were compressed from five micrometers to one and the number of grayscale levels extended toward the tens of thousands, they estimate the theoretical capacity could reach roughly 20 PB, with read-and-write speeds potentially rising toward ten thousand times today's Blu-ray at a small fraction of the energy. Those figures are an extrapolation of an extrapolation and should not be read as near-term expectations.
What to Keep in Mind
A few qualifications keep the story honest.
First, this is peer-reviewed science published in a leading journal, which gives the core result real weight, but it is laboratory work on small samples — no 0.4 PB consumer or archival disc exists yet. Second, the most striking numbers (0.4 PB, and especially the future 20 PB and ten-thousand-times speed) are theoretical estimates and projections from the researchers, not measured products. Third, optical storage historically competes against magnetic and semiconductor systems that keep improving on their own, so capacity on paper is only one factor; cost, speed, durability, and whether equipment makers adopt the format all decide its fate. Finally, "store-and-compute integration" remains a concept here, not a demonstrated machine.
What to Take Away
The genuine advance is the change of strategy. For decades, denser optical storage mostly meant ever-smaller points and ever-more layers, pursued with complicated two-beam hardware. The Shanghai team instead made a single light spot encode far more — 1,024 states, ten bits — using a single beam and a light-sensitive polymer.
If that physics can be engineered into affordable, fast, durable discs and drives, it would give the exploding volume of cold AI-era data a new, compact, long-lived home, with the eventual possibility of computing right where the information is stored. The 0.4-petabyte DVD is not for sale and is not around the corner. But the result published in Nature Photonics shows a credible new path toward it — and makes clear how early that path still is.