DOE ASCR's $30M Lab Call: Four Panels, Ten Awards, and 'Super Intelligence' in the Mission Statement
November 11, 2026 · 7 min read
Granted Research Team · Editorial policy
The document's first sentence is the place to start. Describing the mission of its Advanced Scientific Computing Research program, the Department of Energy says ASCR exists "to advance applied mathematics and computer science, including super intelligence (SI) and quantum information science (QIS)." The abbreviation "SI" runs through all of LAB 27-3645, the program announcement Office of Science issued on November 9, 2026. In places where earlier ASCR solicitations would have said artificial intelligence, it says super intelligence. One of its four review panels is titled "Foundational Research for Explainable SI Reasoning."
The terminology is new. The rest of the call is a familiar ASCR exploratory-research competition. It offers about $30 million for two to three awards in each of four review panels, with individual awards expected around $3 million over three years. Each panel has its own pre-proposal deadline, and the earliest one falls on November 22, 2026, 11 days from now.
Who Can Apply, and Who Can Only Join
LAB 27-3645 is a DOE national laboratory announcement. Only DOE labs can lead a proposal. FFRDCs sponsored by other federal agencies "are not eligible to submit," and other federal agencies and their FFRDCs cannot be named as subawardees either.
Universities and companies take part through subawards. The announcement does not accept collaborative proposals, the SC format in which each institution submits its own linked application. Multi-institutional teams must submit "one proposal from a designated lead institution with all other team members proposed as subawards," and "the lead institution performs a greater portion of the scientific and technical work than any other team member." The explainable SI panel goes further: "Multi-lab collaborations are strongly encouraged, and the team could include non-DOE lab partners via subcontracts."
For a faculty member in numerical linear algebra or quantum networking, that changes the approach. You do not apply. You get onto a lab's team before the lab's pre-proposal deadline, and the lab has to do most of the work.
Other limits apply to everyone:
- Three per panel. A lab may lead no more than three pre-proposals or proposals per panel. There is no limit on subaward appearances.
- Two per PI. A PI on any pre-proposal or proposal can be senior or key personnel on no more than two subawards or separate submissions to each panel.
- Permanent staff only as PI. "Individuals in term-limited appointments, whether as adjunct, visiting faculty, fellows, or similar appointments, are not eligible to be proposed as a PI." Lab postdocs and term staff can be on the team but cannot lead it.
- $1 million a year in total. Every panel caps requests at $1,000,000 per year "in total across all institutions" for a three-year award. A university subaward comes out of that figure, not on top of it.
The Four Panels and Their Calendars
| Panel | Pre-proposal due | Encourage/discourage by | Full proposal due | Max awards | Program manager(s) |
|---|---|---|---|---|---|
| Analog computing and computation through natural systems | Nov 22, 2026 | Dec 11, 2026 | Jan 24, 2027 | 3 | Marco Fornari |
| Sparse linear algebra on modern and emerging supercomputers | Feb 19, 2027 | Mar 12, 2027 | Apr 12, 2027 | 2 | David Rabson, Xujing Davis |
| Explainable SI reasoning | Mar 8, 2027 | Mar 30, 2027 | May 7, 2027 | 2 | David Rabson, Xujing Davis |
| Distributed quantum computing | Apr 5, 2027 | Apr 27, 2027 | May 31, 2027 | 3 | Marco Fornari |
Add up the maximum award counts and you get ten awards. At $3 million each, that is the $30 million total. In practice, every encouraged pre-proposal is competing for one of two or three slots. Only applicants whose pre-proposal is encouraged may submit a full proposal.
The pre-proposal is short: three pages, and the first page "must specify at least one scientific hypothesis." Full proposals allow a 10-page project narrative. Funded teams must attend an annual PI meeting and should budget travel for it.
Analog computing (due first)
The analog panel builds on ASCR's Basic Research Needs for Analog Computing for Science report. It starts from the observation that analog computing "lacks a unifying paradigm for algorithmic design and abstraction," which makes it hard to judge whether results are reliable and reproducible. ASCR wants mathematical and computer-science abstractions "linking physics, algorithms, benchmarking tools, and architecture," along with integration into HPC systems. Labs that have been studying neuromorphic, photonic, or physics-based computing substrates have 11 days to put a three-page hypothesis together.
Sparse linear algebra
This is the most traditional panel. It asks for new sparse algorithms that make better use of the GPUs in current supercomputers, or that target dataflow architectures, ASICs or neuromorphic hardware. The announcement sets a clear limit: "the focus is on ideas that contribute to applied mathematics." Hardware design, memory management and runtime systems are "outside the scope of this panel." A proposal that reads like systems software will be discouraged.
Explainable SI reasoning
This panel's text is the most specific and the most demanding. ASCR says current methods "produce plausible and stochastic outputs through opaque mechanisms that are prone to hallucination," and asks for ways of "replacing or augmenting standard autoregressive and black-box SI" with reasoning that is verifiable and interpretable. It suggests causality, formal proof methods, uncertainty quantification and built-in scientific constraints as possible directions. It names chain of thought, causal tracing, logit attribution and mono-semanticity as acceptable starting points, but requires novelty that goes "beyond them" to "overcome the fundamental representational and reasoning limits of purely statistical token prediction."
The exclusion is just as explicit: "Work primarily focused on deploying existing models to domain applications, building domain workflows, or developing applied platform infrastructure is out of scope." This is a foundations call. Proposals must include "concrete approaches for validating and quantifying" reliability. University interpretability and formal-methods groups should find a lab partner for this one.
Distributed quantum computing
The last panel follows the Basic Research Needs in Quantum Computing and Networking report. It asks for software toolchains, protocols and APIs for distributed heterogeneous quantum computing, as well as distributed shared qubit memory, program partitioning, and error-corrected quantum communication. Hardware components such as quantum transduction and variable-delay quantum buffers "may be also considered." Cryptography, quantum cybersecurity and the quantum internet are out of scope. Proposals should connect to ASCR's effort to deploy "an end-to-end software toolchain for scalable fault-tolerant quantum computation."
Where This Sits in DOE's AI Reorganization
LAB 27-3645 comes out of an Office of Science that has been reorganized. In November 2025, DOE created an Office of Artificial Intelligence and Quantum under the Under Secretary for Science, and the administration's fiscal 2027 request proposed about $1.2 billion for it to coordinate Genesis Mission work. ASCR remains its own program, with a fiscal 2027 request of about $1.1 billion. The Genesis Mission's main Office of Science solicitation, DE-FOA-0003612, closes December 17, 2026 on the Office of Science open FOA list.
Measured against that, $30 million is small. Its value is in the subjects it funds. Explainable reasoning and analog computing are foundational areas that the large AI-for-science calls tend to pass over in favor of applications. For lab researchers whose work does not fit the Genesis Mission's application focus, these panels may be the best fit at DOE this year.
One minor point: the announcement's own Key Facts table lists the number as "LAB 37-3645." The cover page, the Office of Science lab announcements list and every other reference say LAB 27-3645. Use 27-3645 in PAMS.
Practical Moves for Labs and University Partners
Lab PIs targeting the analog panel: you have until November 22 to submit three pages through your authorized institutional representative, not on your own. Confirm today that your lab's submitting office knows the pre-proposal is coming. Write the hypothesis on page one first, since a missing hypothesis is an easy way to get discouraged.
Lab PIs targeting the later panels: use the time to settle team structure. Under the $1 million annual cap, two university subawards of $200,000 a year leave $600,000 for the lead lab, which keeps you within the "greater portion" rule. Three or four subawards will not. Tell your program manager early about teaming plans. The announcement "strongly encourages" this.
University faculty: look for a co-author at a lab, not a sponsor. The best fits are applied mathematicians already working with Frontier, Aurora or El Capitan users on sparse solvers, formal-methods and interpretability researchers for the SI panel, and quantum networking groups in the DOE quantum research centers. Plan on a subaward of $150,000 to $300,000 a year and on the lab leading the writing.
Everyone: do not overstate the SI language. The explainable reasoning panel asks for mathematics and computer science that make models trustworthy for science. It does not ask for claims about building superintelligence. Proposals that read as promotion rather than hypothesis-driven foundations will lose to ones that propose falsifiable work.
The panels are staggered from now through May, so most teams have time to prepare well. The analog teams do not. Whichever panel you aim for, Granted can help turn a lab-university collaboration into a three-page pre-proposal with a clear hypothesis before the encourage/discourage decision.