AI can now propose thousands of potential materials in the time conventional research methods might take to evaluate a fraction of them. The harder problem is what comes next: making those materials, testing whether they behave as predicted, and turning promising candidates into manufacturable devices. ATLANT 3D is targeting that gap with NANOFABRICATOR PRO, a physical manufacturing and experimentation platform designed to connect AI-driven materials discovery with atomic-scale fabrication, validation and device prototyping.
The next phase of AI for science may not be about generating better predictions. It may be about building the physical infrastructure needed to test them.
ATLANT 3D says its new NANOFABRICATOR PRO is designed to bridge that divide. The company describes the system as a physical platform for AI-driven materials discovery that combines materials design, programmable atomic-scale fabrication, experimental validation and device prototyping.
That positioning places the machine in an emerging category sometimes described as Physical AI for materials science: systems in which AI proposes or optimizes experiments while automated hardware produces and measures the physical results.
The distinction is important because AI-based materials discovery has advanced rapidly, but prediction alone does not create a usable material.
Microsoft Research’s MatterGen, for example, uses generative AI to propose novel inorganic materials based on desired properties. Google’s DeepMind has used its GNoME system to identify millions of potentially stable crystal structures. Yet those computational discoveries still have to survive synthesis and experimental testing before they can become useful technologies.
ATLANT 3D’s approach is to move that validation loop closer to the AI system itself.
The company’s Direct Atomic Layer Processing (DALP) technology is the underlying fabrication technology for NANOFABRICATOR PRO. Rather than treating fabrication as a separate downstream step, ATLANT 3D says its platform is designed to make material creation programmable at atomic scale.
For researchers working on semiconductors, advanced packaging, quantum technologies and other emerging applications, the potential advantage is a tighter design-to-experiment cycle.
A conventional materials workflow might involve computational researchers identifying candidates, fabrication specialists producing samples, characterization teams measuring them and device engineers determining whether the resulting material is useful. Each handoff can introduce delays and limit the number of experimental iterations.
An integrated platform could instead create a loop in which computational predictions are translated into physical samples, measured and fed back into subsequent experiments.
That concept is already gaining traction elsewhere.
Google DeepMind’s GNoME research demonstrated the scale of AI-assisted materials prediction, identifying 2.2 million crystal structures, including 380,000 predicted to be stable. Researchers working with Berkeley Lab subsequently demonstrated an autonomous laboratory capable of using AI-guided approaches to synthesize new materials.
A 2026 perspective published in Communications Materials similarly argues that the next generation of autonomous materials laboratories will need to orchestrate broader research campaigns rather than simply automate individual experiments. The authors describe self-driving laboratories as a way to accelerate learning, reduce resource use and automate increasingly complex experimental programs.
NANOFABRICATOR PRO fits into that broader movement, although ATLANT 3D is approaching the problem from the fabrication side.
The company says the platform is industrialized with Automated Industrial Robotics (AIR) and is SEMI-compliant for advanced semiconductor manufacturing. AIR is manufacturing the platform in the United States, according to ATLANT 3D, supporting the company’s expansion into the U.S. market.
That manufacturing relationship could prove important. Research platforms often demonstrate technical capabilities at laboratory scale but face a separate challenge when customers require repeatability, industrial controls, serviceability and manufacturing capacity.
ATLANT 3D says its next step is to add integrated metrology and additional processing capabilities, ultimately moving toward self-driving platforms built around NANOFABRICATOR PRO.
For enterprise and research organizations, integrated metrology is particularly significant. A self-driving materials system cannot simply fabricate samples; it needs reliable measurements to determine whether an experiment succeeded. Those measurements become the feedback signal for the next computational or manufacturing decision.
This is where Physical AI differs from a conventional generative AI application.
A chatbot can generate an answer and stop. A physical AI system must operate within a closed loop: predict, fabricate, measure, learn and repeat.
That loop could become increasingly important as semiconductor manufacturers, advanced-packaging companies and quantum technology developers search for materials with highly specific electrical, thermal, optical or mechanical properties.
The competitive landscape, however, is not limited to equipment vendors. Microsoft, Google and other major technology companies are developing AI models and scientific-computing systems that attack the discovery problem from the computational side. NVIDIA is also investing heavily in accelerated computing and AI platforms for scientific workloads.
The emerging opportunity lies in connecting these layers.
For researchers, that could mean using generative models such as MatterGen to propose candidates, high-performance computing to simulate them and platforms such as NANOFABRICATOR PRO to manufacture and test the most promising options. MatterGen itself is designed to generate materials under constraints including chemistry, symmetry and mechanical, electronic or magnetic properties.
The challenge will be proving that integrated platforms can consistently outperform conventional workflows on meaningful industrial metrics—not simply produce impressive demonstrations.
That means customers will need evidence around fabrication accuracy, repeatability, throughput, material compatibility, measurement quality and the ability to integrate the system into existing semiconductor and research environments.
ATLANT 3D’s launch therefore represents more than another piece of nanofabrication equipment. It is a bet that the future of AI-driven materials science will require physical infrastructure designed to close the loop between algorithms and atoms.
If that model works, the laboratory of the future may look less like a collection of disconnected instruments and more like an automated development system—one capable of moving continuously from an AI-generated hypothesis to a physical material, a measurement and the next experiment.
Market Landscape
AI materials discovery is developing along two interconnected tracks: computational generation and autonomous experimentation.
Microsoft’s MatterGen demonstrates how generative models can propose materials based on desired properties rather than simply screening existing databases. Google’s GNoME demonstrates how deep learning can expand the pool of predicted stable materials at unprecedented scale.
The remaining bottleneck is physical validation.
Self-driving laboratories address that problem by combining robotics, instrumentation, optimization algorithms and automated experimentation. Recent research suggests the field is moving toward systems capable of managing multi-step research campaigns rather than isolated experiments.
For semiconductor and advanced-materials organizations, this creates a new technology stack:
AI models → simulation → candidate selection → atomic-scale fabrication → metrology → experimental data → AI optimization.
ATLANT 3D is attempting to occupy the fabrication and validation portion of that stack.
The commercial opportunity extends beyond semiconductor manufacturing. Potential applications include quantum devices, advanced packaging, sensors, photonics and other technologies where material properties directly constrain device performance.
The key industry question will be whether Physical AI platforms can reduce the time and cost required to move from a computationally promising material to a reproducible, manufacturable technology.
Top Insights
- ATLANT 3D launched NANOFABRICATOR PRO, linking AI materials discovery with programmable atomic-scale fabrication, validation and device prototyping for research teams.
- The platform targets a major AI-for-science bottleneck: converting computationally predicted materials into physical samples that can be measured and manufactured.
- Its DALP technology and planned integrated metrology position NANOFABRICATOR PRO within the emerging self-driving laboratory and Physical AI ecosystem.
- Microsoft MatterGen and Google DeepMind’s GNoME demonstrate the discovery side, while autonomous laboratories increasingly address experimental synthesis and validation.
- Semiconductor, quantum and advanced-packaging teams could benefit if integrated physical-AI workflows improve experimentation speed, repeatability and materials development economics.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
