Is it possible to store movies, photos, and documents by writing them in the A, C, G, and T letters of a DNA molecule? The idea circulates with force every time a headline about “DNA memory” appears. Yet the work published in January 2026 by the team led by Kavya S. Keremane in Advanced Functional Materials doesn’t store a single byte inside a genetic sequence: what has been built is a memristor with synthetic DNA, and understanding the difference completely changes what one thinks they read.
When DNA Acts as a Scaffold and Not as a Hard Drive
A new device combines synthetic DNA, perovskite, and silver to reduce the power draw of experimental resistive memories, but its operation is not biological; it is purely electronic: DNA stores no data, it merely helps build the circuit that does. These are the key takeaways to keep in mind:
- The device does not encode data in genetic letters: bits are stored as high or low electrical resistance states inside a memristor.
- Synthetic DNA acts as a matrix and scaffold: its 22-base pairs regulate the charge and the silver filaments.
- The 17 µW consumption measures the switching power: it is compared against other laboratory memristors, not against commercial SSDs.
- The 1.1-hour retention and the use of lead set the current limits: it works in the lab, but is far from industrial scaling.
But, what is a memristor?
Before proceeding, it is helpful to clarify the term that gives the device its name. A memristor, short for “memory resistor,” is a two-terminal electronic component whose electrical resistance changes when a voltage or current pulse is applied, and retains that resistance value even when power is removed, like a switch that remembers which way it was left. It is different from a conventional resistor, which stores no state, and also from a computer’s RAM, which loses its data when powered down.
The concept was formulated in 1971 by engineer Leon Chua as the fourth fundamental element of electronics, alongside resistance, capacitance, and inductance, although a real device was not built until 2008, at HP laboratories. Since then, memristors have been studied mainly for two uses: low-power nonvolatile memories, like the one in this work, and neuromorphic computing, where their gradual resistance states mimic how the strength of a neuronal connection changes.
From the Headline to the Scalpel: “DNA Memory” Is Not a Biological USB Drive
It is worth starting with the key correction. Classical molecular data storage, the approach that does encode information in the sequence of nucleotides to archive it in vitro, is a real and fascinating line of research, but it is not what these Pennsylvania State University researchers did. Here synthetic DNA does not function as an archive or documentary medium: it is an additional structural brick in nanoelectronics.
The device is a memristor, a resistive memory that combines three materials: 22-base synthetic DNA fragments, silver nanoparticles, and a quasi-2D perovskite based on lead and iodine. The information does not live in the genetic letters, but in different electrical resistance states.
The media language “DNA as memory” thus blends a biological metaphor with an electrical measurement. They are completely different planes. And that initial confusion also contaminates the headline’s star figure, that “100 times less consumption,” which does not mean what it seems at first glance.
Memristor with Synthetic DNA vs. Storage in A-C-G-T Sequences: How Do They Differ?
To understand the technical boundary between both worlds, it helps to contrast their operating principles:
Does this device store information in the DNA sequence, like the “DNA hard drive” discussed in other projects?
No. In molecular storage where data is encoded in A, C, G, and T, the order of the bases themselves is the data: to read it, one must sequence the molecule through synthesis and enzymatic reading. In Penn State’s memristor, the 22-base sequence is fixed and merely provides an insulating and porous molecular scaffold; the data is stored as an electrical resistance state and read by applying a voltage in milliseconds, not by sequencing anything.
Can this chip be sequenced to recover the information?
There would be little point: the information is not in the nucleotide chain, but in whether the silver conductor path between the electrodes is open or closed at that moment.
Why use DNA in the fabrication then?
Because DNA is an extraordinarily regular, dense, and predictable polymer at the nanometer scale. Researchers use it as an ultrathin dielectric layer that helps confine the electric field and stabilizes the formation of conductive metal filaments at very low voltages.
The information does not reside in the sequence of nitrogenous bases, but in the electrical resistance that traverses the material.
Anatomy of the Chip: What’s Really Inside the Device
The prototype is organized as a vertically layered structure, a nanoscale sandwich where each element performs a specific physical function:
- Bottom electrode: an indium tin oxide (ITO) layer deposited on glass.
- Hybrid active layer: a thin quasi-2D perovskite film intercalated with 22-base synthetic DNA oligonucleotides.
- Top electrode: silver contacts that supply the metallic ions for switching.
The 22-mer synthetic DNA does not encode anything in its sequence. It acts as a dense, adjustable molecular scaffold that helps control charge transport and switching. It is solid-state material and physics, not a support for a genetic file. This is the key correction that almost no headline respects when talking about “DNA memory”.
Meanwhile, the perovskite provides the semiconductor behavior: it belongs to the halide perovskite family and contains lead and iodine, a detail far from trivial when assessing viability and environmental impact. The silver nanoparticles actively participate in the formation and rupture of the conductive filaments that reconfigure the device’s logical state.
How the Information Is Stored: Electrical States, Not Letters
Let us reiterate the point that dispels the misunderstanding: here information is not written in A/C/G/T sequences. The data are tied to different resistance states of the memristor, exactly as in other experimental resistive memories (RRAM). The device’s language is electrical pulses. A write operation, technically a SET or RESET, reconfigures the silver conducting paths and fixes a new resistance state:
- SET pulse: Connects the silver filaments and lowers the electrical resistance (low-resistance state, equivalent to a logical “1”).
- RESET pulse: Partially breaks those metal bridges and resistance spikes (high-resistance state, equivalent to a logical “0”).
To read, it is enough to apply a very small voltage that does not alter those paths and measure how much current flows: high current indicates one state, low indicates the other. With an ON/OFF ratio greater than 10^5, the system has ample margin to clearly separate the two logic states.
Electrical pulses connect or disconnect atomic silver bridges, enabling ultraprecise reads with an ON/OFF ratio above 100,000.
The Star Data “100× Less” or What Exactly Was Measured
We reach the figure that dominates international headlines. The claim of “100× lower consumption” carries a common confusion among three physical magnitudes that are not interchangeable: the instantaneous switching power, the energy per operation, and the total energy consumption of the computing system. Confusing them is like mixing the peak speed of an engine with the fuel it uses during a single acceleration or during an entire trip.
Let us look at the verifiable numbers of the prototype under low-current conditions. The memristor operates at 0.17 V and 100 µA, with a switching power of about 17 µW (P = V × I). The article also describes a second operating regime (with 1 mA and voltage below 0.1 V) where the power is around 100 µW. These are distinct operating points within the lab.
Where, then, does the “100×” factor come from? The comparison is made strictly against other experimental memristive devices described in the scientific literature, using the switching power during SET/RESET at low-current operation as the reference.
It does not demonstrate, in any case, that a commercial SSD, a flash memory card, or a conventional RAM module consumes a hundred times more electricity in your computer. In fact, the Penn State statement itself sometimes uses more cautious formulations, talking about “approximately one-tenth of the power” of comparable technologies, which counsels treating multiplicative figures with critical caution.
There is an additional nuance: the 17 µW corresponds to a calculated programming power, not to static continuous consumption. And to compute the real energy per operation ($E = P times t$), it is essential to know the exact duration of each pulse.
Laboratory Performance: Cycles, Retention, and Quick Comparisons
The Keremane team’s prototype logged around 1,000 switching cycles and a retention of 4,000 seconds, equivalent to 1.1 hours. These are notable records for a materials science proof of concept, but they remain well short of the commercial demands of consumer electronics.
It is wise not to confuse “seconds of retention” with a “years-long lifetime.” Extrapolating its real durability would require accelerated degradation tests under different thermal conditions, ambient humidity, and matrices of thousands of interconnected cells.
Endurance and retention are two independent properties: a device can withstand millions of writes and lose charge in minutes, or keep data for decades but degrade by the tenth cycle. Although the team reports stability for six weeks at room temperature and thermal tolerance near 121 °C, these are basic characterization milestones, not industrial homologation.
The Big “But”: Stability, Manufacturing, and Scaling
An academic demonstration with a single isolated cell is far from a technology ready to be integrated onto silicon wafers. The leap to industry requires solving three major challenges:
- Large-scale reproducibility: manufacturing matrices of millions of identical memristors with no performance variation from cell to cell.
- CMOS integration: making the deposition of DNA biomolecules and perovskites compatible with the standard thermal processes of semiconductor foundries.
- Toxicity and sustainability: The presence of lead and iodine-containing halides in the perovskite raises significant regulatory constraints (such as the European RoHS directive) for eventual commercialization.
The memristor with synthetic DNA and perovskite is an elegant advance that expands the boundaries of hybrid materials for neuromorphic computing and in-memory processing. But, for now, it remains a promising lab demonstration: it won’t replace your phone’s memory in the near term, nor does it constitute a genetic storage of data in the biological sense of the term.
