A DNA Computer Smaller Than 2 nm Semiconductors — and It Remembers

For decades, the semiconductor industry has raced to shrink the transistor — packing ever more switches onto a chip by pushing process nodes from 10 nanometers to 7, then 5, and now 2. But physics imposes a wall: electrons behave unpredictably at scales approaching the width of a few atoms, and silicon can go no further. Now, a team of South Korean researchers has vaulted past that wall altogether — not by reinventing silicon, but by abandoning it entirely.

On April 1, 2026, Professor Yeongjae Choi and colleagues at the Graduate School of Engineering Biology at the Korea Advanced Institute of Science and Technology (KAIST) published a study in Science Advances describing a DNA-based molecular computer that operates at approximately 0.34 nanometers — the natural spacing between adjacent nucleotide base pairs. That is roughly six times smaller than the current cutting-edge 2 nm semiconductor node, and the system does something no earlier DNA circuit could: it remembers.

The Reusability Problem DNA Computing Always Had

DNA is an attractive computing medium on paper. Its complementary base-pairing rules allow it to be precisely programmed, and its extraordinary information density — encoded in the 0.34 nm gaps between bases — is unrivalled by any human-engineered medium. Researchers have long known how to build DNA logic gates that respond to molecular signals: the presence of a cancer biomarker, for instance, could trigger a cascade of reactions that output a detectable result.

The fatal flaw was that these systems were fundamentally one-shot. Once a DNA circuit reacted to a signal, the molecules were consumed or their structure irrevocably altered. The circuit could not be reused, and it could not retain its output for participation in further computation. Every new calculation required a fresh preparation. For practical biomedical applications — where continuous monitoring, repeated diagnosis, or multi-step logical inference might be required — this was a crippling limitation.

A Self-Holding Circuit That Locks Its State

The KAIST team’s innovation was to design novel DNA molecules that respond to input signals not by consuming themselves, but by changing their spatial conformation — their three-dimensional shape — and then permanently locking into that new state. The resulting stable structure simultaneously encodes the output of the computation and serves as a persistent memory of it. Crucially, this locked state can participate in subsequent logical operations, enabling cascaded, multi-step computation without any external reset.

The team describes the design as a “reset-free” DNA logic circuit. Unlike all previous DNA computing architectures, it processes information in real time without requiring external intervention between steps, and it retains historical calculation results indefinitely. The paper’s title — Reset-free DNA logic circuits for real-time input processing and memory — captures the achievement precisely. This is, for the first time, genuine read-write storage at the molecular scale.

In engineering terms, the team has replicated the core logic function of a transistor at the DNA molecular level. A transistor is the fundamental semiconductor building block: it receives a signal, performs a switching operation, and outputs a result that can propagate to the next stage of a circuit. The DNA bio-transistor does exactly this, but at a scale and using mechanisms that silicon cannot approach.

Research Details at a Glance

Why This Matters for Medicine

The most immediate horizon for this technology is not a DNA laptop, but something far more impactful: molecular diagnostics that live inside the body. Conventional disease diagnosis requires a sample to be extracted, transported to a laboratory, processed by bulky instruments, and interpreted by specialists. The whole chain takes hours to days, and it is a snapshot in time.

A programmable, reset-free DNA circuit could in principle be introduced into a biological environment — perhaps circulating in the bloodstream — where it continuously monitors molecular signals associated with disease. When a threshold pattern is detected, the circuit’s self-locking memory records the event without external intervention. Subsequent stages of the circuit could then trigger a response, flag an alert, or release a therapeutic molecule. The entire diagnostic and therapeutic loop could unfold at the molecular scale, in real time, without ever leaving the body.

The KAIST team sees this as the foundational step toward “intelligent bio-systems” in which molecules themselves process and store information, going beyond simple chemical reactions to autonomous information-handling. That vision also has implications for environmental sensing, synthetic biology, and any domain where computation must occur in wet, biological conditions hostile to silicon electronics.

The Road Ahead

Important caveats remain. The current demonstration is a proof of concept: the team has shown that the core logical primitives — the transistor-equivalent building blocks — can be realized in DNA with persistent memory. Scaling from individual logic elements to complex, programmable circuits capable of real diagnostic tasks will require substantial further work. Questions of operational stability in physiological fluids, speed relative to electronic systems, and integration with biological readout mechanisms are all open.

Still, the significance of establishing the principle should not be understated. Silicon’s dominance in computing rests on exactly this foundation: the invention of the transistor, followed by decades of engineering refinement. The KAIST result plants a similar flag at the molecular level, and the base-pair spacing of 0.34 nm provides a physical lower bound on miniaturization that silicon will never touch.

The door to molecular computing, long theorized and long blocked by the reusability problem, has been opened.