National Science Foundation funds UT San Antonio and Texas A&M to adapt large language models on machines that can't afford to run them
The National Science Foundation made two matched standard grants on August 14th, 2026. They fund a subspace optimization framework for resource efficient and private fine tuning of large language models. One went to the University of Texas at San Antonio and one to Texas A&M University, each obligating the same amount. The award record states the problem as reaching the performance of full parameter fine tuning where computing resources are limited and training data are sensitive or off limits to share. It names small organizations, universities and healthcare institutions among those the current requirements shut out. Both awards run from January 1st, 2027 to December 31st, 2029.
How to take part
The award records are public at the National Science Foundation award pages under award numbers 2619079 and 2619080. No comment window, hearing or application step is stated in either record.
Where
1 county. In San Antonio-New Braunfels.
Timeline
- decided
Award date on both National Science Foundation award records
- Today
- effective
Start date of both awards
72 days out
How this decision moved
One dated line per check, oldest first. A line that says nothing changed means somebody looked and it had not.
- 2026-08-23
Admitted. Both award records were read directly and carry the same title, the same obligated amount and the same term.
- 2026-08-26
Checked and unchanged. The decision still stands as decided.
- 2026-08-29
Checked and unchanged. Both matched grants still stand, one at the University of Texas at San Antonio and one at Texas A and M. The stated problem is still reaching the accuracy of full parameter fine tuning on machines that can't carry it.
- 2026-09-01
The joint Texas A and M and UT San Antonio language-model efficiency award remains in the federal record.
- 2026-09-02
Checked and unchanged. The decision still stands as decided.
- 2026-09-05
The San Antonio and College Station award on adapting large models for machines that can't afford to run them still stands as made.
- 2026-09-08
The UT San Antonio and Texas A&M award for adapting large language models on constrained machines still stands as made.
- 2026-09-09
The UT San Antonio and Texas A and M work adapting large language models is still funded as awarded.
- 2026-09-12
Checked and unchanged. The decision still stands as decided.
- 2026-09-13
Checked and unchanged. The decision still stands as decided.
- 2026-09-16
The award to UT San Antonio and Texas A&M for adapting large language models on constrained machines is still in force as made. No change has been posted.
- 2026-09-19
The award stands, and the problem it is aimed at is still adapting large models on machines too small to run them comfortably.
- 2026-09-23
The award still stands, and the problem it names is still running a large model on hardware that can't afford one.
- 2026-09-26
The San Antonio and College Station award still stands, still aimed at running large language models on modest hardware.
- 2026-09-29
The award to UTSA and Texas A&M still aims at running large language models on hardware too small to host them whole.
- 2026-10-02
UTSA and Texas A&M are still funded to fit large language models onto hardware too small to hold them whole.
- 2026-10-05
The NSF award to UT San Antonio and Texas A&M for adapting large language models still stands.
- 2026-10-08
The NSF award for UT San Antonio and Texas A&M to adapt large language models is still active.
- 2026-10-11
The NSF award for adapting large language models at UT San Antonio and Texas A&M remains active.
- 2026-10-15
The award to UT San Antonio and Texas A&M for adapting large language models on smaller machines is still active.
- 2026-10-18
The award to UT San Antonio and Texas A&M for adapting language models on smaller machines is still listed.
- 2026-10-21
Checked and unchanged. The decision still stands as decided.
The evidence
Every fact above rests on one of these. The words are the source's own.
These requirements can limit the ability of small organizations, universities, healthcare institutions, and other resource-constrained users to benefit from recent advances in AI.National Science Foundation, Award Abstract 2619079 Primary source, official · api.nsf.gov
The scientific problem addressed by this project is how to achieve performance comparable to full-parameter fine-tuning when computing resources are limited and training data are sensitive or cannot be shared.National Science Foundation, Award Abstract 2619079 Primary source, official · api.nsf.gov
The research contributes to societal benefits by broadening access to advanced AI tools, strengthening the protection of sensitive information, and reducing the financial and energy costs of model adaptation.National Science Foundation, Award Abstract 2619079 Primary source, official · api.nsf.gov
"id":"2619079","initAmendmentDate":"08/14/2026"National Science Foundation, Award Record 2619079, award fields Primary source, official · api.nsf.gov
"id":"2619080","initAmendmentDate":"08/14/2026"National Science Foundation, Award Record 2619080, award fields Primary source, official · api.nsf.gov
"awardeeName":"University of Texas at San Antonio"National Science Foundation, Award Record, award fields Primary source, official · api.nsf.gov
"fundsObligatedAmt":"120000"National Science Foundation, Award Record, award fields Primary source, official · api.nsf.gov
"awardeeName":"Texas A&M University"National Science Foundation, Award Record, award fields Primary source, official · api.nsf.gov
Questions about this decision
Answered from the record itself. Every answer is assembled from stored fields, so an answer the record has no basis for is left out rather than guessed.
What is this decision?
The National Science Foundation made two matched standard grants on August 14th, 2026. They fund a subspace optimization framework for resource efficient and private fine tuning of large language models. One went to the University of Texas at San Antonio and one to Texas A&M University, each obligating the same amount. The award record states the problem as reaching the performance of full parameter fine tuning where computing resources are limited and training data are sensitive or off limits to share. It names small organizations, universities and healthcare institutions among those the current requirements shut out. Both awards run from January 1st, 2027 to December 31st, 2029.
Who decides it?
United States National Science Foundation decides. The record names the deciding body for every entry it carries.
Can the public take part?
The award records are public at the National Science Foundation award pages under award numbers 2619079 and 2619080. No comment window, hearing or application step is stated in either record.
Where in Texas does it apply?
It covers Bexar County.
Has it been decided?
It has been decided. The dates on the item page carry when.
What happens next?
The next dated step is effective, on January 1st, in 72 days.
When did it start?
The earliest date on its record is August 14th, 2026.
What kind of decision is it?
It is filed under research and science.
What sources back it?
Two sources back it. Two of them are primary.
Is it on the ERCOT grid?
Yes. It sits inside the ERCOT interconnection.
When was it last checked?
Every fact on it was last verified against its source on October 21st, 2026.
Cite this
Texas AI Docket, National Science Foundation funds UT San Antonio and Texas A&M to adapt large language models on machines that can't afford to run them. Tracked since August 14th, 2026. Last verified October 21st, 2026. https://texasaidocket.com/item/tx-2026-0092/. Reuse permitted under CC BY 4.0 with attribution. The same entry is in the docket JSON as item tx-2026-0092.
Beat
Filed under Research and science, with every other decision on that beat.