DARPA's Defense Sciences Office Just Opened Four FY26 SBIR Topics Closing July 22 — Engineering Sleep, Air-Independent Batteries, Rydberg Sensors, and Real-Time Pathogen Prediction

DARPA's Defense Sciences Office (DSO) is the agency's hardest office to characterize, and that is by design. Where the Microsystems Technology Office owns semiconductors and the Biological Technologies Office owns engineered biology, DSO is DARPA's portfolio for the foundational science that does not yet have a home — the "what if the laws of physics let us do that" bets that, when they pay off, spawn entire new technical fields. Its SBIR topics are correspondingly eclectic and, even by DARPA standards, technically severe. The FY26 DSO drop under Broad Agency Announcement series DPA26BZ03 is a perfect specimen: four topics, four nearly unrelated scientific domains, one shared deadline.

All four topics open June 24, 2026 and close July 22, 2026 at 12:00 PM Eastern. They use DARPA's standard SBIR structure — a roughly six-month Phase I to establish feasibility, followed by a Phase II of up to two-plus years for prototype development and demonstration. Read together, they are not a coherent roadmap the way an MTO drop is; they are four independent shots at four hard problems DSO program managers think are newly tractable. Below, what each one is actually asking for — and the kind of small business that can credibly answer.

This post is the deep analysis; for the quick brief see Granted News.

Topic DPA26BZ03-DV012: Engineering Sleep for Cognitive Performance

The ask is a wearable, non-invasive, closed-loop system that enhances the restorative functions of sleep. Parse each adjective, because each one is a wall. Non-invasive rules out the implanted electrodes and pharmacological interventions that dominate the sleep-neuroscience literature. Closed-loop means the device must sense sleep state in real time and intervene during sleep — most plausibly by detecting slow-wave sleep and delivering precisely timed acoustic, thermal, or electrical stimulation to deepen it — rather than simply tracking sleep for a morning report. Restorative functions points at the specific physiology of slow-wave and REM sleep: memory consolidation, glymphatic clearance of metabolic waste, and hormonal recovery.

The military rationale is obvious and acute. Warfighters routinely operate on degraded, fragmented sleep, and cognitive performance collapses predictably under sleep debt. A device that lets a soldier extract more recovery from four hours than they currently get from six is a direct force-multiplier. The commercial parallel — consumer sleep-tech, shift-worker recovery, elite athletics — is enormous, which makes this a rare DARPA topic with a clear dual-use commercialization path baked in.

Who wins: teams that already have a working closed-loop neurostimulation or sensing platform and can show real EEG/physiological validation data, not slideware. The differentiator will be the closed-loop control algorithm and clinical-grade sleep-staging accuracy, not the wearable form factor. If your Phase I proposal cannot point to existing slow-wave detection performance, you are starting a lap behind.

Topic DPA26BZ03-DV013: ExCAIPE — Expeditionary Closed and Air-Independent Power and Energy

ExCAIPE asks for closed, electrically rechargeable, high-energy-density batteries that operate without drawing on external air. The "air-independent" specification is the entire problem. Many of the highest energy-density electrochemistries — metal-air chemistries above all — achieve their density precisely by using ambient oxygen as a reactant, which is wonderful until you need to operate underwater, in a sealed vehicle, in space, or at altitude where the air is gone. A sealed system that keeps metal-air-class energy density while carrying its own oxidant, and remains electrically rechargeable rather than single-use, is a genuinely hard cell-chemistry and systems-integration challenge.

The expeditionary framing matters: DSO wants power for contested, austere, off-grid operations — undersea systems, autonomous platforms, dismounted units far from any logistics tail. Energy is a logistics problem before it is an electrochemistry problem, and the program is buying logistics independence.

Who wins: battery and electrochemistry shops with a differentiated cell chemistry and, critically, a credible story for the system — oxidant storage, thermal management, rechargeability cycle life. A pure materials lab with a high-density cell but no path to a sealed, rechargeable, manufacturable pack will struggle against teams that think at the system level. This topic rewards integrators.

Topic DPA26BZ03-DV011: MANTRAS — Manufacturing Technologies for Rydberg-Based Atomic Sensors (SBIR XL)

MANTRAS is the topic that exposes where a real technology is stuck. Rydberg atomic sensors — which use atoms excited to high-energy "Rydberg" states as exquisitely sensitive, self-calibrating RF field detectors — are one of the most promising emerging sensing modalities for electronic warfare and communications, because they can measure radio-frequency signals with a sensitivity and broadband reach that conventional antennas cannot match. The science works in the lab. The problem, which this topic names directly, is manufacturing: building a low-SWaP (size, weight, and power), ruggedized, manufacturable Rydberg sensor for real-time RF measurement, rather than a delicate benchtop physics apparatus.

The "SBIR XL" designation signals this is a larger, more ambitious topic than a standard SBIR slot — appropriate, because the gap here is the unglamorous, capital-intensive work of turning a quantum-sensing demonstration into a producible device. That is exactly the valley where promising defense tech dies.

Who wins: quantum-sensing and precision-photonics companies that have already built a Rydberg or comparable atomic-sensing prototype and are now fighting the packaging, vapor-cell manufacturing, laser-stabilization, and ruggedization battle. The proposal differentiator is a credible manufacturing and yield story, not another sensitivity record.

Topic DPA26BZ03-DV014: Real-Time Pathogen-Host Interactome Prediction

The fourth topic asks for the capability to rapidly characterize host–pathogen interactions directly from pathogen protein sequences, to prioritize medical countermeasures. In plain terms: given a novel or engineered pathogen's protein sequence, computationally predict how its proteins will interact with human host proteins — the "interactome" — fast enough to triage which countermeasures to pursue before the pathogen has been characterized experimentally.

This is a biodefense response-time play, riding the wave of protein-structure and protein-interaction prediction unleashed by the AlphaFold generation of models. The defense logic is the threat of a fast-moving, possibly engineered outbreak where the weeks normally spent on wet-lab characterization are weeks you do not have. A predictive interactome model that narrows the countermeasure search space in hours is a strategic capability.

Who wins: computational biology and ML teams with demonstrated protein-protein interaction prediction performance and the data infrastructure to back it. This is the topic where a small, sharp AI-for-biology startup with the right model and benchmark results can outcompete a larger but less specialized shop. Validation data on held-out host-pathogen pairs will be the currency.

How to play the July 22 deadline

Four weeks is a short fuse for a DARPA SBIR, and DSO topics are written for performers who are already deep in the relevant field — these are not topics you can credibly enter cold. A few cross-cutting moves:

• Match the topic to existing evidence, not aspiration. Every one of these four rewards a team that can point to prior validation data — EEG performance, cell cycle-life, sensor prototypes, interactome benchmarks. Phase I is about de-risking the specific approach, and proposals grounded in real prior results win.
• Mind the new reauthorization rules. The 2026 SBIR/STTR reauthorization (signed April 13) brings per-company proposal caps in FY2027 and tougher eight-watchlist foreign-risk screening. DoD is implementing reforms first. If your cap table or workforce has foreign ties, get your documentation in order before you submit.
• Think past Phase II. ExCAIPE and the sleep system have the clearest dual-use commercialization paths; the new Strategic Breakthrough Award ($30M post–Phase II) is the long-game prize for whichever of these matures fastest with private co-investment.
• Engage the program manager early. DSO PMs write topics around specific technical bets. The proposals that resonate are the ones that demonstrably understand the bet, not the ones that restate the topic language.

DSO topics are where DARPA places its weirdest, highest-ceiling wagers. Engineering sleep, sealed high-density power, manufacturable quantum sensors, and predictive biodefense are four very different futures — and on July 22, four very different small businesses will find out whether DARPA wants to fund theirs. Granted tracks every live DARPA SBIR topic across all offices, with deadlines and full solicitation detail, so you can match your evidence to the right bet before the window closes.