When most people hear the word “robot,” they picture something with wheels, arms, and a power cable. But some of the most consequential robots being built today are invisible to the naked eye, composed of nothing but strands of DNA, and small enough to fit inside a single human cell.
A major 2026 review published in SmartBot from researchers at Peking University, Stanford, and King’s College London offers the most comprehensive picture yet of where DNA-based machines stand and where they are headed. The results are remarkable, and they deserve to be on the radar of anyone who cares about the future of medicine, technology, and science.
What Is a DNA Machine?
To understand why DNA machines are so exciting, it helps to appreciate what makes DNA such a special building material.
DNA, the same molecule that carries your genetic code, is built from four chemical letters: A, T, G, and C. These letters pair up predictably: A always bonds with T, and G always bonds with C. This simple rule, called Watson-Crick base pairing, allows scientists to design DNA sequences that self-assemble into precise, programmable shapes almost automatically.
Here’s the engineering insight at the heart of the field: double-stranded DNA (where two strands are paired up) behaves like a rigid rod, while single-stranded DNA (one strand on its own) is highly flexible. By combining these two properties in a single design, researchers can build molecular structures with stiff load-bearing “arms” connected by flexible “joints, exactly the architecture you need to build a functioning machine.
In 2006, a landmark technique called DNA origami transformed the field. Like folding a piece of paper into a crane, DNA origami uses hundreds of shorter “staple” strands to fold a long DNA strand into a precise, predetermined shape. The technique enabled researchers to construct remarkably complex 3D structures, gears, boxes with controllable lids, vases, humanoid robot shapes, and even a crank-slider mechanism that mimics the mechanics inside an engine, all at the nanometer scale.
Since then, the field has advanced rapidly. Today, DNA machines can walk along predefined tracks, grip objects, open and close in response to molecular signals, detect specific viruses in saliva, and deliver cancer drugs directly to tumor cells.
Biological Inspiration: Nature Already Built the First Nanomachines
The scientists designing DNA machines didn’t start from scratch. They looked to nature for blueprints, and what they found was extraordinary.
Living cells are already full of molecular machines operating at staggering precision. The protein ATP synthase acts as a rotary motor, spinning at the molecular level to produce the energy currency that powers nearly every process in your body. Kinesin and dynein are linear motors that literally walk along intracellular tracks, carrying molecular cargo. RNA polymerase is a molecular reader that moves along DNA strands and transcribes genetic information.
These biological machines operate with remarkable efficiency under chaotic conditions buffeted by heat, surrounded by competing molecules, and navigating environments that would destroy any conventional machine. They demonstrate that nanoscale robots capable of rotation, transport, and force generation are not just theoretically possible; they already exist in every cell of your body.
The challenge scientists have taken up is to build synthetic versions that can be programmed to do what we want rather than what evolution dictates.
What DNA Robots Can Do Right Now
The applications that have already been demonstrated span an impressive range of domains.
Fighting disease. One of the most striking demonstrations came from a DNA nanogripper, a machine composed of a central palm and four flexible, finger-like arms engineered to detect SARS-CoV-2 viral particles directly in human saliva. The device achieved sensitivity comparable to PCR testing, the gold standard for viral detection, and demonstrated the potential to block viral entry into cells physically. Meanwhile, DNA nanoswitches have been designed that change shape in response to specific RNA molecules, providing a platform for nucleic acid diagnostics. In another celebrated experiment, a DNA nanorobot was programmed to seek out tumor vasculature in living animals, deliver anticoagulant drugs specifically to the blood vessels feeding the tumor, and trigger localized clot formation that cut off the tumor’s blood supply, all with low immunogenicity and favorable biocompatibility.
Targeted drug delivery. DNA machines can be loaded with therapeutic cargo and designed to release it only in response to specific molecular signals within cells. Researchers have used this approach to deliver CRISPR-Cas9 gene-editing components to specific locations within living cells, with delivery regulated by temporal molecular cues.
Atomic-scale manufacturing. DNA nanostructures can serve as templates for positioning other materials, such as gold nanoparticles, semiconductor nanowires, and graphene, with precision approaching 0.34 nm, the distance between adjacent base pairs. This has enabled the construction of nanoscale photonic devices, electronic circuits, and nanoparticle arrays with properties that cannot be achieved through conventional manufacturing.
Molecular computing and data storage. In 1994, researchers first demonstrated that DNA could be used to solve computational problems. Today, DNA circuits can perform logic operations and even machine learning tasks. In 2025, researchers demonstrated DNA neural networks capable of supervised learning, a striking convergence of biology and computation. On the storage side, DNA’s theoretical information density reaches 10¹⁹ bits per cubic centimeter, roughly 1,000 times that of the best flash memory available today. Recent breakthroughs have enabled the writing of arbitrary digital data onto DNA molecules and its retrieval via high-throughput sequencing.
How Scientists Control DNA Machines
Building a machine is only half the challenge. The other is making it move on command.
Several strategies have been developed for controlling DNA machines, each with its own strengths and trade-offs. Electric fields can rotate or reposition entire DNA structures in milliseconds by exploiting the DNA backbone’s natural negative charge. Magnetic nanoparticles can be attached to DNA structures and then steered by external magnetic fields, which is particularly useful for in-body applications where non-invasive control is essential. Light and heat can trigger conformational changes by destabilizing or reforming DNA bonds.
The most programmable approach, however, is DNA strand displacement, a chemical process in which a new DNA strand is introduced that hybridizes with part of an existing structure, displacing the original strand and triggering a mechanical change. This technique allows for precise, step-by-step control of complex machines with minimal cross-talk between components. It is already being used to control machines with multiple independently movable joints, making it the closest DNA nanotechnology has come to the fine motor control seen in conventional robotics.
The Role of AI: Designing Tomorrow’s DNA Robots
One of the most exciting developments highlighted in the 2026 review is the convergence of artificial intelligence with DNA nanotechnology.
Designing a DNA machine is extraordinarily complex. Researchers must simultaneously consider the sequence of hundreds or thousands of nucleotides, their folding pathways, the mechanical properties of each structural element, and the kinematic behavior of the assembled machine. Currently, this requires deep expertise and significant trial and error.
AI is beginning to change that. Graph neural networks have been developed to predict the three-dimensional shape that a given DNA sequence will fold into, dramatically accelerating the design cycle. Large language models are being explored as design assistants that could allow researchers to specify a functional goal in plain language, “build a hinge with a specific rotational stiffness,” and receive an automatically optimized DNA sequence. Deep learning-based molecular dynamics tools are extending the timescales and spatial scales at which DNA machines can be simulated, opening new windows into how these machines actually move and behave.
The goal articulated by leading researchers is a closed-loop, AI-driven design pipeline: specify a function, receive a design, simulate its behavior, manufacture it autonomously, and iterate on all of this without the bottlenecks of traditional laboratory work.
The Challenges That Remain
The paper is candid about how far this field still has to go. DNA machines today are primitive compared to natural molecular motors. They operate stochastically rather than deterministically, meaning their behavior is governed by probabilities and thermal fluctuations rather than precise mechanical control. Scaling up from laboratory demonstrations to industrial production remains an unsolved engineering problem, though bacterial fermentation-based manufacturing platforms offer a promising path. And keeping DNA machines intact inside living systems is a persistent challenge, as biological enzymes are designed to break down DNA and often degrade these machines before they can complete their tasks.
There are also fundamental limits to how precisely DNA machines can be controlled. At the nanoscale, the random thermal motion of molecules, Brownian motion, is not a minor perturbation but a dominant force. Building machines that can operate reliably in this environment, rather than being jostled by it, requires rethinking many of the principles that apply to macroscopic robots.
Why This Matters
We are at the beginning of a new era in engineering, one where the boundary between the biological and the mechanical is dissolving. The same molecule that encodes the instructions for life is being repurposed as a construction material for machines that can navigate inside the body, assemble structures atom by atom, and process information in ways silicon cannot.
The 2026 review describes this convergence not as a distant possibility but as an active research frontier, with new demonstrations appearing monthly. DNA machines have already detected viruses with PCR-level accuracy, killed tumors in animal models, and stored digital images in molecular form. What they will be able to do in another decade, particularly as AI accelerates design and robotics automates fabrication, may be genuinely difficult to predict.
The molecule of life is becoming the material of the future.
Reference
An, Y., Wu, F., Xiong, Y., Zhang, C., Dai, J. S., & Zhou, L. (2026). Designer DNA-Based Machines. SmartBot, 2, e70029. https://doi.org/10.1002/smb2.70029