NSF plans up to $75 million for AI ready scientific instruments
The National Science Foundation plans to spend up to $75 million over five years on scientific instruments designed around data and AI. In its October 8 announcement, NSF said a funding opportunity for the effort, known as Super Intelligence Native Instruments, or SINI, will follow. The instruments would support its Programmable Cloud Laboratories Network. Awards and detailed technical requirements are still ahead.
Why the instrument layer matters
The earlier PCL solicitation helps explain the engineering problem. A programmable cloud laboratory lets a remote user specify an experimental workflow that physical equipment carries out. That workflow can include preparation, measurement and analysis, with results feeding the design of another experiment. The challenge is making those stages work together while keeping the resulting evidence trustworthy.
The 2025 call identifies standardized data collection, experimental design and interpretation as gaps. It also calls for common practices in instrument validation and metadata, the contextual information that accompanies a result. These are requirements of the earlier laboratory program. They explain the setting for SINI without establishing the specifications of its forthcoming call.
Consider two instruments returning the same numerical reading. A researcher may still need to know whether they measured comparable samples under comparable conditions. Making a number available to software is only part of the task. Making the measurement usable in the next scientific decision requires enough context to understand what that number represents.
A new layer for an existing network
NSF’s July 22 network announcement described $380 million for 20 teams over four years, alongside upwards of $20 million from the Astera Institute. That earlier commitment supports a national test bed spanning fields including biology, chemistry and materials science. Its goal is to let researchers run custom experimental workflows remotely across participating facilities.
Astera’s role includes helping standardize data and metadata and making scientific outputs easier to share and reuse. That matters for a network intended to connect experiments across institutions. NSF also describes researchers and students working alongside automated systems throughout the process. The July announcement sets out a network being established, rather than evidence that every proposed capability is already available.
What connected instruments look like in practice
Carnegie Mellon University offers one example of the infrastructure involved. In its July account of the AI Science Foundry, the university described more than 80 robotically controlled instruments across two cloud labs. It said the Foundry would receive up to $20 million over four years through the earlier NSF network effort.
The Foundry combines remote experiment design and execution with analysis and repeated cycles of learning from results. Initial research targets include functional polymers, microbial biomaterials, organoids and aluminum alloys intended for high temperature applications. CMU also plans shared standards, open software, data resources and training, so the project extends beyond acquiring equipment.
These details illustrate the range of instruments and workflows that a connected laboratory must coordinate. They come from a participating institution’s description of its program. They are not an independent measurement of scientific productivity, and CMU’s July award does not establish that it will receive SINI funding.
What researchers should watch for
The earlier solicitation calls for testing reliability and reproducibility, conducting failure analyses and establishing common approaches to experiment verification. Its proposal deadline was November 20, 2025. Researchers should wait for the separate SINI notice before treating any application rules or dates as relevant to the new effort.
A useful instrument demonstration would let another researcher inspect the experimental record and understand how an unusual result was handled. For a laboratory considering adoption, the practical questions include how much adaptation its workflow needs and how easily staff can diagnose a failed run. Those questions provide a way to assess future demonstrations without assuming that more automation guarantees better science.
The next concrete milestone is publication of the funding opportunity. NSF’s announcement leaves partner identities, eligibility, deadlines and award sizes unspecified. Its super intelligence terminology describes an ambition, with the instruments and their performance still to be established.
Related reporting covers DOE’s four autonomous science robotics testbeds.
