DARPA's SBIR Release 5 Wants Concrete Printed on the Seafloor, ISR That Understands What It Sees, and Cell Lines From Any Species. Topics Open August 26.

DARPA's FY26 SBIR/STTR machine now runs on a metronome. Since the Department of War moved to pre-releasing new topics on the first Wednesday of every month, the agency's solicitations have stopped being unpredictable annual events and become a schedule you can plan a company around. Release 5 pre-released on August 5, 2026, opens for submissions on August 26, and closes September 23 at 12:00 PM ET — the same tight four-week window that defined Release 4's July drop.

That cadence is the story before the topics are. The three-week gap between pre-release and open is not dead time — it is the single most valuable window in the DARPA SBIR process, because during pre-release you can talk directly to the topic authors. Once the topic opens, that channel closes and communication is restricted. Small businesses that treat the pre-release as a research phase — reading the topic, calling the technical point of contact, pressure-testing their approach against what the program manager actually wants — consistently write sharper proposals than those who start writing on open day. Release 5 pre-released August 5. If you are reading this on the 6th, you have the full window in front of you.

Here is what DARPA is buying this month.

The five SBIR topics

Release 5 leans unusually hard into the physical and maritime — three of the five SBIR topics touch the undersea domain in some way, a clear signal about where Strategic Technology and Defense Sciences are pointing.

Hoboken — SBIR XL (DPA26BZ04-DV018, Strategic Technology Office). DARPA wants subsystems for an integrated underwater 3D concrete printing system that builds structures on the seafloor using local sediments and seawater rather than materials shipped in from the surface. This is in-situ resource utilization for the ocean floor — the maritime cousin of printing lunar habitats from regolith. The hard problems are chemistry (getting concrete to cure reliably in cold seawater from variable local feedstock), robotics (deposition at depth), and autonomy. It is a natural fit for teams in marine materials, cementitious chemistry, or subsea robotics who can own one subsystem rather than the whole stack.

Semantically-Aware ISR (DPA26BZ04-DV019, Strategic Technology Office). This funds mission-aware semantic communications for low-SWaP tactical ISR platforms operating in degraded or contested communications environments. The idea is ISR that transmits meaning rather than raw bits — a sensor that understands which parts of what it sees matter to the mission and prioritizes those under bandwidth collapse. This is squarely an AI/edge-inference play: you are compressing and prioritizing at the semantic layer on power- and size-constrained hardware. Teams working on edge ML, learned compression, or resilient tactical networking should look hard at this one.

Universal Cell Culture Platform (DPA26BZ04-DV020, Biological Technologies Office). The full title — "Rapid Establishment of Species-Agnostic Cell Lines and Autonomous Culture Operations" — describes the ask precisely. DARPA wants a platform that can rapidly establish viable, reproducible cell lines from diverse species, with autonomous culture automation doing the hands-on work. This is a biotech-meets-robotics topic: the value is in generalizing cell-line establishment across organisms (not just the handful of well-characterized human and mouse lines) and removing the human bottleneck from culture maintenance. It rhymes directly with the autonomous-laboratory thread running through the DOE Genesis Mission SBIR call — a company with an autonomous-culture platform has two federal buyers pursuing the same capability at once.

Open Architecture Platform for Underwater Vehicles (DPA26BZ04-DV021, Defense Sciences Office). DARPA wants a modular, open-architecture autonomous underwater vehicle (AUV) supporting rapid payload adaptation and fleet-level collaboration. The operative word is open — the government is explicitly trying to avoid vendor lock-in by funding a platform where payloads and behaviors can be swapped in by third parties. If your strength is the vehicle, the modular interface standard is the deliverable; if your strength is payloads or multi-vehicle autonomy, this topic is building the bus you will eventually plug into.

The STTR topic: tooling for algorithm discovery

Release 5 also carries a single STTR topic, and STTR's structure is the reason to pay attention to it.

SPEED DIAL (DPA26TZ05-DV003, Defense Sciences Office). This funds a framework and tool suite for integrating discovered algorithms into scientific workflows — infrastructure for taking algorithms that AI systems surface and actually operationalizing them in real research pipelines. It is a meta-tool, and it sits at the intersection of automated science and software engineering.

The STTR distinction matters practically: STTR requires a formal partnership with a research institution — a university or federal lab must perform at least 30 percent of the work, versus SBIR's requirement that the small business perform the majority. For SPEED DIAL specifically, that partnership requirement is a feature, not a hurdle: an academic algorithms or scientific-computing group is exactly the collaborator that makes a credible team here. If you are a small software firm without a standing academic partner, the pre-release window is also the time to line one up.

How to actually work the Release 5 window

The mechanics of DARPA's cadence reward a specific playbook, and it is worth being explicit about it.

Use the pre-release to talk to the topic author. This is the whole game. Between August 5 and August 26, the technical points of contact will answer questions. Ask them what "good" looks like, what failure modes they have seen in past attempts, and what adjacent efforts already exist. A fifteen-minute call can save you from writing a technically excellent proposal against a problem the PM has already solved a different way.

Respect the four-week clock. The window from August 26 open to September 23 close is short by design. DARPA's monthly rhythm assumes you did your positioning during pre-release and are writing, not researching, once the topic opens. Teams that start cold on open day are structurally behind.

Match your topic to your genuine core competency — and only bid one subsystem when that is the ask. Several of these topics (Hoboken, the open-architecture AUV) are explicitly component or subsystem plays. You do not need to own the entire system to win; you need to own the piece DARPA named and show you can deliver it. Over-scoping a Phase I proposal to cover the full stack is a common way to lose.

Read the topic against the broader map. Note how many of Release 5's topics — autonomous cell culture, semantic ISR, algorithm-discovery tooling — echo the autonomy-and-AI-for-science theme now running through DOE, NSF, and the broader Genesis Mission. A capability that fits a DARPA topic often fits a civilian-agency call too. Building once and selling to multiple federal buyers is the highest-leverage move a small deep-tech company can make in 2026.

Release 5 is a normal month for the new DARPA. That is precisely why it is worth taking seriously: there will be a Release 6 in September, a Release 7 in October, and the companies that win are the ones who have turned this monthly cadence into a standing pipeline rather than reacting to each drop as if it were a surprise. The first Wednesday of every month is now a date on your calendar. This month's topics close September 23 — the pre-release conversations that decide who wins are happening right now.

Want DARPA's monthly topic drops matched against your company's technical profile the moment they pre-release? Granted tracks SBIR/STTR solicitations across every agency and surfaces the ones where your capabilities actually fit — with the deadlines and pre-release windows that matter.