DAF26BZ05-NV027: Integrating Neural Architectures for Brain-Inspired AI with Low SWAP

Below is a brief summary. Please check the full solicitation before applying (link in resources section).

Quick Answer

The Department of the Air Force is offering a Phase I SBIR award of up to $250,000 for companies developing brain-inspired AI architectures that combine multiple neural network types into a single low SWAP system. This is topic DAF26BZ05-NV027 under the DoW FY26 SBIR BAA Release 5. Proposals are due September 23, 2026 through DSIP.

What This Solicitation Is

This is a standard Phase I SBIR topic, meaning applicants are proposing a feasibility study. It falls under the Applied Artificial Intelligence critical technology area, spanning an unusually broad set of component priority areas: Trusted AI and Autonomy, Advanced Computing and Software, Quantum Science, Space Technology, Integrated Sensing and Cyber, and Microelectronics. It carries a projected CMMC Level 2 (Self) requirement. This topic description does not carry the standard ITAR/EAR restriction language found on other topics in this release, though applicants should still confirm export control applicability with the contracting officer given the DoD application space.

What DAF Is Looking For

Today's AI systems mostly rely on a single, monolithic neural network architecture. DAF wants research that instead combines multiple neural network paradigms, things like convolutional networks, recurrent networks, transformers, spiking neural networks, and other biologically-inspired models, into one unified system modeled loosely on how the human neocortex organizes cognitive processing.

The goal is a system that emulates properties like hierarchical organization, contextual attention, working memory, and decision-making under uncertainty, while achieving low Size, Weight, and Power so it can run on resource-constrained platforms like wearables, drones, or autonomous vehicles. DAF wants to see the resulting system outperform state-of-the-art single-architecture approaches on cognitive tasks such as object recognition, natural language processing, or spatial reasoning.

Phase I work centers on a literature review of existing integration strategies, a conceptual design for the neocortex-inspired integrated architecture, feasibility demonstration of key components through simulation or small-scale prototypes, and identification of candidate low SWAP hardware approaches such as neuromorphic computing or spiking neural networks.

Phase II moves to a fully functional prototype, extensive testing against relevant DoD tasks and benchmarks, and hardware and software optimization to minimize SWAP while preserving performance. Phase III targets TRL 6 entry, aimed at autonomous systems for both military use, specifically tactical air dominance and Advanced Battle Management System applications, and commercial robotics and healthcare applications.

Funding and Timeline

Award maximum is $250,000 with a maximum period of performance of 6 months. Proposals exceeding either figure will not be considered. The technical volume page limit is 20 pages.

DAF also provides up to $6,500 in Technical and Business Assistance funding per Phase I award, on top of the cost proposal total.

Proposals are due September 23, 2026 through DSIP. DAF anticipates evaluation and selection within roughly 90 calendar days of solicitation close.

Who Should Apply

This topic fits companies with deep neural network research experience across multiple architecture types, ideally with some background in neuromorphic computing or spiking neural networks specifically, since those are called out as candidate low SWAP hardware directions. Companies coming from a computational neuroscience or cognitive architecture background have a natural edge given the neocortex-inspired framing. This is fundamental research at an early TRL, so a strong conceptual design and simulation-level proof of feasibility matters more than working hardware at this stage.

Eligibility and Compliance Notes

At least two thirds of the research effort must be performed by the awardee, measured by direct and indirect costs. All research and development work must happen in the United States except in rare, specifically approved circumstances.

Small businesses majority owned by multiple venture capital operating companies, hedge funds, or private equity funds are eligible to submit and receive awards under this Phase I BAA release.

Only one principal investigator can be designated per proposal, and their technical resume with publications must be included.

DAF runs a foreign risk evaluation and due diligence review on every proposal under 15 U.S.C. 638(vv).

Frequently Asked Questions

What is the proposal deadline?
September 23, 2026, submitted electronically through DSIP.

How much funding is available?
Up to $250,000 for a Phase I award with a maximum period of performance of 6 months, plus up to $6,500 in additional Technical and Business Assistance funding.

What TRL does Phase III begin at?
TRL 6, targeting tactical air dominance and Advanced Battle Management System applications.

Are venture capital or private equity backed companies eligible?
Yes. This Phase I BAA release does not exclude VC, hedge fund, or private equity owned small businesses.

What is the page limit for the technical volume?
20 pages.

What hardware approaches are called out as candidates for low SWAP?
Neuromorphic computing and spiking neural networks are specifically mentioned as directions to define in Phase I.

What happens in Phase III?
Transition planning involves coordination with AFRL/RI and potential engagement with DAF customers, targeting both military autonomous systems applications and commercial robotics and healthcare applications.