Building the base for laboratories that run themselves — NSF AI grant $4.96M (Stanford; ~$19.9M planned)
NSF awarded about $4.96 million to "PCL-Test Bed: Building America's AI Infrastructure for Self-Driving Laboratories" at Stanford University. The estimated total is about $19.9 million, running from August 2026 to July 2030.
Grant overview (primary data)
- Award amount$4,959,405 / Est. total $19,908,812
- RecipientStanford University (California)
- ProgramPCL-Programmable Cloud Labs
- Period2026-08-01 〜 2030-07-31
- FunderU.S. National Science Foundation (NSF) / NSF
Key points
- NSF Award 2607573, "Building America's AI Infrastructure for Self-Driving Laboratories," to Stanford University (CA).
- The theme is the base for self-driving laboratories — looping plan, run, measure, and decide with AI and robots.
- Program: PCL-Programmable Cloud Labs, oriented toward making instruments callable remotely.
- Obligated $4,959,405 against an estimated total of $19,908,812 — a roughly fourfold gap showing a plan partway through.
- Period: August 1, 2026 to July 31, 2030. No abstract is included in this dataset.
- No point in the loop of choosing, running, measuring and feeding back waits on a human judgement — that is what self-driving means here.
1What a self-driving laboratory is
A self-driving laboratory runs the loop of planning an experiment, performing it, measuring, and deciding what to try next without a person intervening at each step. Robots handle the samples, instruments take the measurements, and AI reads the results and sets the next conditions. Unlike a person doing the same work, the loop does not stop overnight and does not vary in judgment.
The gain is largest in fields such as materials and chemistry, where the number of times you can try and check shapes the result.
2Framing it as infrastructure
The award is named for building AI infrastructure for such laboratories, not for a laboratory itself. The direction is a shared mechanism rather than something each lab assembles on its own. The program name, Programmable Cloud Labs, points the same way: making instruments callable remotely.
Several awards under the same PCL program were recorded at other institutions around the same time, suggesting parallel work across multiple sites.
3The gap between obligated and estimated
Obligated funds are $4,959,405 against an estimated total of $19,908,812, a roughly fourfold difference reflecting a large multi-year plan partway through. The specific plan is not in this dataset, which carries no abstract, so consult the official NSF page (Award 2607573).
4A loop that does not stop when the person leaves
A self-driving laboratory earns the name not from fast instruments but from a closed loop. Deciding what to try next sits inside the mechanism too, so there is no point at which the process waits on a human judgement.
- 1Choose the conditionsAI reads the results so far and picks what to try next
- 2Run itRobots handle the samples and carry out the experiment as specified
- 3MeasureInstruments record the result and return it as data without passing through a written note
- 4Feed it backThe returned result enters the next choice of conditions, turning through the night without variation in judgement
Fields such as materials development and chemistry, where the number of times you can try and check decides the outcome, ask for this shape because how many turns you get is how fast you go. What this award funds is not an individual laboratory but the base that makes the loop commonly available.
Why it matters
A self-driving laboratory brings instruments, robotics, data platforms and AI into one moving system. If it is built as shared infrastructure, the standards for connecting to it become the question for instrument makers and software vendors. In fields where the number of trials shapes the result — materials, chemistry, drug discovery — it can change how R&D itself is run.
FAQ
What is a self-driving laboratory?
Why is it framed as infrastructure?
Why is the gap to the estimated total so large?
Sources (primary)
Source: NSF Award Search (U.S. National Science Foundation, public domain). Amounts are the obligated amount. For privacy, we do not handle principal investigator names.
- NSF Award (original, official)
- NSF Award ID: 2607573