A $3M award for Athena, the national AI institute pursuing big AI for small devices — a five-year plan with an estimated total of $15M (Duke University)
A national AI institute creating and deploying advanced AI on devices used in everyday applications while ensuring availability, efficiency, functionality and trustworthiness. It addresses the move from standalone smart devices to integrated intelligent systems embedded in and interacting directly with the physical world. Estimated total $15 million, obligated $3 million.
Grant overview (primary data)
- Award amount$3,000,000 / Est. total $15,000,000
- RecipientDuke University (North Carolina)
- ProgramAI Research Institutes
- Period2026-11-01 〜 2031-10-31
- FunderU.S. National Science Foundation (NSF) / NSF
Key points
- A national AI institute creating and deploying advanced AI on everyday devices while ensuring availability, efficiency, functionality and trustworthiness.
- Its motto is big AI for small devices, premised on running on the device rather than sending data to the cloud.
- It addresses the move from standalone smart devices to integrated intelligent systems embedded in and interacting directly with the physical world.
- The mission is an AI-native full stack design interacting tightly with both classical and quantum computation.
- Estimated total is $15 million with $3 million obligated; the five-year period starts 2026-11-01, the same date as another institute under the same program.
- Processing moves from the cloud onto the device, and from single devices to integrated intelligent systems acting in concert.
1Big AI onto small devices
The motto recorded for the institute is big AI for small devices. Large models were built on the premise of running on large computing resources, while the devices people actually touch are small and limited in power and computation.
Placing advanced AI there while holding availability, efficiency, functionality and trustworthiness means an entirely different set of design constraints from sending data to the cloud for processing.
2From standalone devices to connected systems
What the record describes is a move from standalone smart devices to integrated intelligent systems embedded in and interacting directly with the physical world. Not each unit becoming cleverer but several acting in coordination.
Alongside it sits the phrasing of transitioning from assistive and analytical use toward coordinated autonomous systems — the difference between a tool that helps a person and a system that judges and acts, stated expressly as the subject of research. That the mission names a full stack design interacting with both classical and quantum computation is distinctive as well.
3A fifth of the estimated total is obligated
The estimated total is $15 million with $3 million obligated at present. Across the 120 NSF awards this site holds as of 2026-08-31, most have estimate and obligation matching, while a certain number carry a gap as this one does.
The period runs five years from 1 November 2026 to 31 October 2031, and the records this site holds as of 2026-08-31 show another institute under the same AI Research Institutes program beginning on that same date. Establishing institutes evidently moves together at the turn of a fiscal period.
4One device getting smarter, and a system getting smarter
The record describes a move in two stages. First, taking AI built on the assumption of large computing resources and moving it onto devices limited in both power and computation. Then, moving from a device on its own to several acting in concert.
The motto recorded for the institute is big AI for small devices. Its mission also names full-stack design interacting closely with both classical and quantum computing. Unless the design runs coherently from top to bottom, the conditions on the right cannot be met within the resources available.
Why it matters
Debate about AI gathers around large models and large computing resources, while the devices people actually touch face tight constraints. Running at the edge changes both model design and evaluation criteria. That the move from standalone devices to coordinated systems is itself the subject of research indicates the premise of designing products one unit at a time is shifting.
FAQ
What does big AI for small devices mean?
Why does quantum computing appear?
Why is the obligated amount lower than the estimate?
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: 2625580