Scientists Built a Memory Chip Out of DNA and Crystal — Here’s How It Actually Works
In short: A Penn State team built a “memristor” — a memory component that remembers an electrical state even after power is cut — by combining synthetic DNA with a crystalline semiconductor called perovskite. The device runs on under 0.1 volts, retains data for six-plus weeks at room temperature, and stays stable up to nearly 250°F. DNA’s job isn’t computing; it’s an ultra-dense, energy-efficient scaffold for storing the information. This is peer-reviewed, patent-filed lab research — not a shipping product, and read/write speed hasn’t been published yet.
Why anyone bothered fusing DNA with a computer chip
A single gram of DNA can theoretically hold around 215 million gigabytes of data — a storage density no conventional material comes close to matching, because DNA encodes information in molecules rather than in the electrical or magnetic states silicon chips rely on. Scientists have known about that potential for years. The hard part has never been DNA’s storage capacity; it’s been getting a biological molecule to actually function inside an electronic circuit built from entirely different materials.
A team led by Penn State materials scientists says it has closed that gap, publishing the work in Advanced Functional Materials with a patent application already filed. Postdoctoral researcher Kavya S. Keremane, one of the study’s co-corresponding authors, put the core challenge plainly: “Biology and electronics are different domains.” Bridging them, according to Penn State’s account of the research, required building an entirely new materials platform rather than adapting an existing one.
What a memristor actually is
An ordinary resistor — the kind found in nearly every electronic device, from a phone to a spacecraft — maintains a fixed resistance to current and forgets everything the instant power is removed. A memristor behaves differently: it can retain a record of which direction current last flowed through it, even with no power applied at all. That’s the property that makes memristors useful as a memory element in the first place, and it’s also what lets a single component both store a piece of information and, depending on its state, influence how current moves through it next — storage and processing happening in the same physical spot, rather than being handled by separate parts of a chip the way conventional computer memory works.
That combination is the reason memristor research keeps getting compared to how brains work. A biological neuron doesn’t have a separate “storage” region and “processing” region either — the same synaptic connection does both, which is part of why brains handle enormous amounts of parallel information on a small fraction of the power a comparable computer would need. Research professor Bed Poudel, another of the study’s co-corresponding authors, tied that directly to why the timing matters: as AI systems demand more from computing hardware, he said the field needs a fundamentally different strategy for combining low power draw with high storage capacity, rather than continuing to push conventional memory architecture further.
How DNA ended up doing the job
| The DNA | Not DNA extracted from a living organism. The team used commercially available synthetic DNA — short, chemically manufactured genetic sequences engineered to a specific composition and length, rather than the long, tangled strands natural DNA forms. |
| Why synthetic, not natural | Natural DNA behaves, in the researchers’ own description, “like wet spaghetti” once extracted — long, disordered, and structurally unpredictable at the nanoscale. Short synthetic fragments hold a precise, designed shape, which is what makes them usable as a building material rather than just a data archive. |
| The semiconductor | Crystalline perovskite — the same class of material already used commercially in solar cells, lasers, and some data storage devices. |
| How they’re joined | Silver nanoparticles are applied to the synthetic DNA layer — a process called “doping” — then integrated with thin films of perovskite. The doping step is what makes the DNA capable of conducting electricity in the first place, since DNA alone doesn’t conduct current well enough to be electrically useful. |
| What DNA is actually doing | Structural and storage scaffolding, not computation. The DNA’s job is to organize the material precisely enough, and densely enough, that the resulting device needs far less power to hold a given amount of information than conventional memory materials do. |
Research professor Neela H. Yennawar, who directs Penn State’s Biomolecular Interactions Core Facility and co-authored the study, described the design process as computational rather than trial-and-error: researchers can calculate exactly which DNA sequences and lengths a given application needs, then have those exact sequences synthesized to order — turning DNA, in her framing, from a biological molecule into what she called a programmable nanomaterials platform.
The actual numbers
On power consumption specifically, Penn State’s account of the research includes two different comparisons that are easy to mix up if you’re not reading closely. Poudel is quoted directly comparing the new device to conventional storage broadly — “flash drives” — saying it “consumes 100 times less power.” Separately, Penn State’s own narrative description compares the device to other existing perovskite-based memristors specifically, putting the saving at roughly one-tenth the power. Those are two different baselines, not two conflicting numbers for the same comparison, and it’s worth keeping them distinct rather than treating either one as the device’s single headline efficiency figure.
What “storage and processing in the same place” solves. In a conventional computer, moving data back and forth between a processor and separate memory chips burns real energy and time — a bottleneck engineers call the von Neumann bottleneck, named after the architecture nearly all modern computers still use. A memristor that stores and processes information in the same component sidesteps that back-and-forth entirely, which is the specific efficiency gain neuromorphic computing research is chasing, independent of any single material choice.
What hasn’t been demonstrated yet
Penn State’s own materials describe a working single device, tested for stability and power consumption — not a memory chip with multiple cells wired into an array, which is the step that would actually be needed to store real, addressable data the way a flash drive or RAM module does. Several practical questions aren’t addressed in the published research summary at all: how fast the device can read and write data, whether synthetic DNA can be manufactured at the volume and cost a commercial memory chip would require, and how performance holds up past the six-week window already tested. A patent application has been filed, and the authors describe plans to refine the approach and explore other bio-inspired electronic applications — language that describes an active, continuing research program, not an imminent product.
Lab result, not lab-to-shelf timeline. Peer review and a filed patent are meaningful markers of scientific rigor, but they say nothing about manufacturing cost, scale, or timeline. Historically, memristor technology in general has taken many years to move from academic demonstration to commercial products, and this specific bio-hybrid approach is earlier in that process than more established memristor materials. Treat this as a genuine materials-science advance worth tracking, not a preview of next year’s laptop specs.
Why this matters beyond one lab result
The framing Poudel gave Penn State connects this directly to a problem showing up across the AI industry right now: computing hardware’s power demands are growing faster than efficiency gains in conventional chip design can keep up with. Most of the current response to that pressure is happening at the infrastructure level — more power plants, bigger data centers, more efficient cooling. This result comes from a different direction entirely: changing the physical material memory is built from, so that the same amount of stored information simply costs less energy to hold and access in the first place. It’s a slower, more uncertain path to impact than a new data center coming online, but if bio-hybrid memristors like this one do eventually scale, the efficiency gain compounds at every single device that uses the resulting chips, rather than being confined to wherever the new power plant happens to be built.
Common questions
What did the Penn State researchers actually build?
A memristor — a memory device that retains information about prior electrical activity even without power — built by combining synthetic DNA with a crystalline perovskite semiconductor. It’s a single demonstrated device, not a multi-cell memory chip.
Is this using real human or animal DNA?
No. The device uses commercially available synthetic DNA — chemically manufactured genetic sequences designed to a specific length and composition, not DNA extracted from any living organism.
How much less power does it use than normal computer memory?
Penn State’s account gives two separate figures for two separate comparisons: roughly 100 times less power than conventional storage like flash drives, and roughly one-tenth the power of other existing perovskite-based memristors specifically. Neither figure has been independently verified outside the published study.
When could this be in an actual product?
No timeline has been published. The research demonstrates a working single device with a patent application filed, but doesn’t address manufacturing cost, scale-up to a full memory array, or read/write speed — all things that would need to be solved before commercial use.
Why does DNA make a good material for computer memory?
DNA can theoretically store around 215 million gigabytes of data per gram, far beyond what conventional storage materials achieve, because it encodes information in molecular structure rather than electrical or magnetic states. The challenge the Penn State team addressed was making DNA electrically functional and structurally precise enough to work inside a real electronic device, not DNA’s storage density itself.
Sources
- Penn State University — “Borrowing from biology to power next-gen data storage,” the original research announcement with full researcher quotes. psu.edu
- Keremane, K.S. et al. — “Molecularly Engineered Highly Stable Memristors with Ultra-Low Operational Voltage: Integrating Synthetic DNA with Quasi-2D Perovskites,” Advanced Functional Materials, 2026. doi.org
- ScienceDaily — August 2026 syndication of the Penn State release. sciencedaily.com
