DARPA is not structured like the rest of DoD
Every DoD SBIR component has program offices and acquisition customers. DARPA has program managers. That is not a semantic difference. It is the single most important structural fact about DARPA SBIR. The program manager (PM) is a rotating four-to-six-year hire — often a senior researcher from industry or academia — who arrives with a research agenda, runs a portfolio of programs in support of it, and then leaves. The PM's SBIR topic is an extension of that research agenda. The review happens inside the PM's portfolio logic. And when the PM departs, the portfolio decays.
DARPA explicitly funds ideas that might fail. Proposals that guarantee success are red flags — they signal incremental work. The optimal DARPA proposal describes a genuinely hard problem where success would be transformative.
This means DARPA SBIR topics read unlike anything else in DoD. They are specific, technical, and often reach for a capability that no vendor is known to have. The PM is not trying to buy a commodity — the PM is trying to probe whether a technical approach works. For a small firm with a real technical angle, DARPA is attractive for exactly this reason. For a firm looking to sell a commercial product to a stable program office, DARPA is the wrong door.
The DARPA offices and their SBIR patterns
Information Innovation Office (I2O)
I2O is the heaviest AI/ML shop at DARPA. Topics across AI Next, Assured Autonomy, CASE, OPS-5G, REMA, TIAMAT, AI Forward and related programs produce SBIR topics on ML robustness, autonomous systems, cyber ML, and information reasoning. I2O SBIR topics often tie to a named DARPA program with multi-year structure — winning a Phase I can open a door to a broader program team performer role.
Microsystems Technology Office (MTO)
MTO focuses on microelectronics, RF, photonics, and quantum. AI-adjacent topics appear around edge inference hardware, neuromorphic compute, and ML for EDA. Overlaps with DMEA work.
Defense Sciences Office (DSO)
DSO reaches farthest into basic research. Topics cross mathematics, materials, biology, and computing fundamentals. SBIR volume is modest but topic specificity is extreme — a DSO topic is usually readable as one PM's bet.
Strategic Technology Office (STO)
STO focuses on air, sea, space, and strategic systems. Topics include spectrum, EW, mosaic warfare concepts. AI plays a supporting role.
Tactical Technology Office (TTO)
TTO focuses on platform systems — aerial, maritime, ground. Autonomy and unmanned systems SBIR topics appear regularly.
Biological Technologies Office (BTO)
Biology, neurotech, pandemic prevention. AI shows up in bioinformatics and protein topics.
Topic count and cadence
DARPA is small relative to Army or Navy. A typical DoD SBIR release includes 10 to 25 DARPA topics, versus 80 to 120 from Army. Releases align with the DoD joint calendar (26.1, 26.2, 26.3 in 2026).
Phase I at DARPA is 250 thousand dollars for 12 months (longer and larger than Army/Navy). Phase II up to 1.8 million over 24 months. The larger Phase I reflects DARPA's posture: they want firms to actually test something substantial, not deliver a paper study.
Review culture: the PM reads proposals
At DARPA, the program manager often reads proposals personally or chairs the review panel. This is fundamentally different from Army or Navy where technical review boards of program office staff read proposals. A DARPA proposal is being read by the PM who wrote the topic. The PM already has a hypothesis. The proposal is evaluated on: does this team believe the same hypothesis, can they test it, and is the test informative.
This produces a specific pattern: DARPA reviewers reward proposals that show technical insight beyond the topic language. They will discount proposals that parrot the topic back. A proposal that identifies an unstated assumption in the topic — and proposes to test it — often outperforms a proposal that hits every stated bullet.
Who wins DARPA SBIR
Firms with deep research depth
DARPA rewards scientific credibility. PhD-level technical staff, recent publications, and a clear track record of advancing a technical area go further here than at any other DoD component.
Former DARPA performers
Firms that have been on a DARPA program team in any role (prime, sub, subcontractor to a university) have massive incumbency advantage. The PM already knows them.
Specialized small firms
A 15-person firm with three PhDs in a narrow specialty can outcompete a 500-person firm with broad AI expertise. DARPA prefers depth over breadth.
Firms that can talk to program managers directly
DARPA PMs are accessible before topic release. A firm that has had two or three pre-release conversations with the PM is positioned differently than a firm that responds cold.
Risk tolerance and program length
DARPA is the only DoD component whose explicit mandate is to pursue high-risk, high-payoff research. Topics describe capability that may not work. Reviewers accept technical risk on the approach. This is freeing for firms with novel technique and punishing for firms that try to pitch a safe incremental improvement.
Programs are short. DARPA programs typically run 3 to 4 years total. Phase III within a DARPA program often means transitioning to a service — Army, Navy, Air Force — as the DARPA program closes. Firms that build a DARPA practice plan the service transition from the proposal, not after the Phase II.
What DARPA SBIR is not
DARPA SBIR is not a good fit for:
- Firms selling a mature commercial product. The PM is not buying SaaS.
- Firms whose main pitch is workflow integration. Topic writers care about technical capability, not UX.
- Firms seeking a stable program-of-record customer. DARPA programs end. The transition is the plan.
- Firms that want long, predictable procurement cycles. DARPA operates on program timelines, not contract timelines.
Practical steps for a first DARPA SBIR
- Pick a DARPA office and read every announced program in that office. PMs list their programs publicly.
- Identify a specific PM whose portfolio matches your technical area.
- Email the PM in pre-release. A one-paragraph note identifying your firm's relevant work is sufficient.
- If the PM replies with a conversation, use that conversation to understand the underlying research hypothesis.
- Write the proposal to that hypothesis, not the topic bullet list.
- Line up a Phase III transition plan: which service program office will absorb the work, which PM or staff officer has indicated interest.
Bottom line
DARPA SBIR is the sharpest instrument in DoD SBIR. For a firm with genuine research depth and a technical angle that matches a PM's portfolio, DARPA is among the best places in DoD to win. For a firm with a mature commercial product looking for a program office buyer, DARPA is the wrong door and the services are right. Read the PM portfolios, pick one, and write to the person, not the paper.
Frequently asked questions
DARPA is program-manager driven, not program-office driven. PMs write topics that reflect their research agenda, read proposals personally, and tolerate technical risk the services would reject.
Phase I is 250 thousand dollars for 12 months. Phase II up to 1.8 million over 24 months. The larger Phase I reflects DARPA's expectation of substantive technical work, not paper studies.
Information Innovation Office (I2O) is the heaviest AI/ML shop. Tactical Technology Office (TTO) and Microsystems Technology Office (MTO) produce AI-adjacent topics in autonomy and edge compute respectively.
Yes. PMs are accessible during pre-release and often earlier. A one-paragraph email identifying relevant technical work is a legitimate opening. Pre-release conversations are where positioning happens.
Firms with research depth, former DARPA performers, and specialized small firms with named PhD technical staff. Generic AI pitches lose. DARPA prefers depth over breadth.
Generally no. DARPA is a research buyer, not a commercial product buyer. Firms with mature commercial products do better with service SBIRs or AFWERX Open Topic.