The DARPA BAA Guide for Startups

DARPA (Defense Advanced Research Projects Agency) funds high-risk, high-payoff research through Broad Agency Announcements (BAAs) and a dedicated SBIR program. BAA awards range from $500K to $10M+ with no fixed phases. DARPA SBIR awards follow the standard Phase I ($250K) / Phase II ($1.8M) structure. DARPA wants 10x breakthroughs, not incremental improvements -- and it actively seeks commercial technology companies, not just defense primes and universities.

DARPA is the most misunderstood funding agency for startups. Most founders assume it's only for military contractors with security clearances. That's wrong. DARPA specifically wants non-traditional companies bringing commercial innovation to national security problems. But the application process is different from anything else in the federal system -- more relationship-driven, more open-ended, and more demanding.

DARPA's two funding paths

DARPA funds research through two distinct mechanisms. Understanding which one fits your company is step one.

If you're new to DARPA: start with SBIR. The amounts are smaller but the process is more structured, the topics are defined, and you don't need a prior PM relationship. Use SBIR to build a track record, then pursue BAA opportunities.

If your technology is genuinely breakthrough: BAAs are where the real money is. 96% of DARPA's budget flows through BAAs, not SBIR.

DARPA's six technical offices

DARPA has about 220 employees and ~100 program managers across six offices. Each office has its own BAA and funds different technology domains.

Each office publishes its own office-wide BAA on darpa.mil/research/opportunities/baa. These are standing solicitations that accept abstracts on a rolling basis, typically for 12+ months.

For most tech startups, I2O (AI, cyber, software) and BTO (biotech) are the highest-probability entry points. MTO matters for hardware/semiconductor companies. DSO, STO, and TTO tend to fund larger, more complex system-level programs.

How to respond to a DARPA BAA

The BAA process is different from every other federal solicitation. It's more like a conversation than a form submission.

Step 1: Find the right BAA and PM

Before writing anything, identify which BAA and which program manager (PM) aligns with your technology.

• Office-wide BAAs: Each technical office has a standing BAA accepting abstracts on rolling deadlines. These are the most accessible entry point.
• Program-specific BAAs: Individual programs publish their own BAAs with more specific technical requirements.
• Where to look: darpa.mil/research/opportunities/baa for office-wide BAAs. SAM.gov for all DARPA opportunities. dodsbirsttr.mil for SBIR/STTR topics.

Step 2: Email the program manager

This is the step most startups skip -- and it's the most important one. Unlike NSF or NIH, DARPA PMs are deeply involved in shaping research directions. They expect to hear from potential proposers before seeing a formal submission.

What to send: a 1-paragraph, non-confidential description of your technology and proposed approach. Not a pitch deck. Not a 5-page white paper. One paragraph: what you'd do, why it matters for DARPA's mission, and what makes your approach novel.

What to expect: the PM will either (a) encourage you to submit an abstract/white paper, (b) redirect you to a different PM or program, or (c) tell you it's not a fit. All three outcomes are valuable. Option (c) saves you weeks.

Step 3: Submit an abstract or white paper

Most BAAs require a brief abstract (2-5 pages) before a full proposal. This is a screening step -- DARPA reviews for relevance and innovation, then invites promising abstracts to submit full proposals.

Your abstract should cover:

• Technical approach: what you'd do and why it's novel (not incremental)
• Relevance to DARPA's mission: how this solves a national security problem
• Team qualifications: why your team can execute this
• Milestones: what you'd demonstrate and in what timeframe

Step 4: Full proposal (by invitation)

If your abstract is invited, you'll submit a full technical and cost proposal. This is negotiated with the PM -- the scope, budget, and milestones are typically refined through discussion before final submission.

Full proposals are evaluated on:

• Technical innovation and merit -- is the approach genuinely novel?
• Contribution to DARPA's mission -- does this advance national security capabilities?
• Team and execution -- can you actually deliver?
• Transition potential -- how does this become an operational capability?

The bar is high. DARPA's standard is "10x improvement," not "somewhat better." If your technology is an incremental optimization, DARPA isn't the right agency.

What DARPA actually expects from startups

DARPA actively recruits non-traditional companies. The Small Business Programs Office specifically targets commercial tech companies. Many programs prefer startups because they bring speed and market discipline that traditional defense contractors don't.

Most programs are unclassified. Especially at the research stage. If your proposal is selected for work that involves classified information later on, DARPA can sponsor the credentialing process. You don't need anything like that to apply.

Individual performers win awards. A startup with a strong technical team and a genuine breakthrough can compete against large prime contractors, especially for research-phase work. You don't need a team of 50 or a university partnership.

BAA competition works differently than SBIR. Office-wide BAAs are open-ended -- fit matters more than head-to-head scoring against other proposals. That said, PMs have fixed program budgets, so you're still competing for finite resources. The difference: if your technology genuinely solves a problem in a PM's portfolio, that PM can fund you without waiting for a scored competition cycle.

Should your startup apply to DARPA?

Be honest with yourself on these criteria.

DARPA is a fit if:

• Your technology represents a genuine 10x improvement (not incremental)
• You can articulate a national security or dual-use application
• You're comfortable with a relationship-driven process (PM engagement is required, not optional)
• You can handle government contracting (reporting, milestones, potential classification)
• Your technology is at least at the concept/feasibility stage

DARPA is probably not a fit if:

• Your innovation is "faster/cheaper/better" -- that's incremental, not breakthrough
• You can't identify a defense, intelligence, or national security use case
• You need structured guidance (DARPA PMs set direction, but you drive the approach)
• You're purely consumer-facing with no dual-use angle
• You can't wait 3-12 months for an award decision

The realistic assessment

For most first-time applicants, DARPA is the wrong starting point -- NSF SBIR is lower friction and has the highest success rate. The DARPA process is less structured, the relationship component is heavier, and the competition includes well-funded university labs and defense primes with decades of DARPA experience. Build a track record with standard SBIR first, then pursue DARPA with credibility.

If your technology is genuinely breakthrough and you can identify a PM whose portfolio it fits, DARPA can be transformative. The awards are larger, the relationships are deeper, and a successful DARPA program opens doors to Phase III contracts and government procurement that standard SBIR can't match.

BAA timeline (what to expect)

The BAA process moves slower than most startups expect:

DARPA SBIR follows the standard DoD timeline (~6 months for Phase I decisions). If you need faster capital, NSF or AFWERX are better bets.

The DARPA SBIR path

If BAAs feel like too much for a first engagement, DARPA's SBIR program is a more accessible entry:

• Phase I: ~$250K for 6 months of feasibility research
• Phase II: ~$1.8M for 24-36 months of prototype development
• Enhancement: Up to $500K in 1:1 matching funds for transition to a military customer (verify current amounts at dodsbirsttr.mil)
• Topics: Published in DoD SBIR solicitation cycles at dodsbirsttr.mil

DARPA SBIR topics tend to be more technically ambitious than other DoD branches. They're looking for the same 10x innovation bar in a smaller package. But the process is more structured and the amounts are predefined.

The white paper: structure, length, and what actually matters

Most BAAs ask for an abstract or white paper of 2-5 pages. This is the document that gets you invited to submit a full proposal -- or quietly filed away. Here's what separates the papers that advance:

Page 1: The hook. State the problem, your approach, and why it matters in the first half-page. PMs read dozens of white papers per week. If they don't understand your proposal's value by the end of page 1, they won't finish reading.

Pages 2-3: Technical approach. This is the meat. Describe what you'd do, how it works, and what makes it novel. Be specific about the mechanism, not vague about the vision. "We use adversarial training with domain-specific threat models to generate robust defenses against LLM prompt injection" is better than "we apply AI to cybersecurity."

Include:

• The technical innovation (what's genuinely new)
• Why existing approaches fall short (the gap)
• Preliminary results or proof points (data beats claims)
• Key risks and your mitigation strategy

Page 4: Team and execution. Briefly describe key personnel (3-5 people), their relevant experience, and your organization's capabilities. If you have prior DARPA, DoD, or IC experience, highlight it. If you don't, emphasize relevant commercial execution.

Page 5: Milestones and budget summary. A high-level timeline (12-18 months typical for Phase 1), 3-5 milestones, and a rough budget range. The PM uses this to gauge scope and feasibility, not to negotiate exact terms.

What NOT to include: Company marketing material, irrelevant past performance, excessive background on known problems, or requests for feedback ("we'd welcome the opportunity to discuss..."). The white paper is a proposal, not a cover letter.

How to engage a DARPA PM (with real examples)

PM engagement is the single biggest differentiator between startups that win DARPA funding and those that don't. Here's the practical playbook:

Finding the right PM. Each BAA lists the PM(s) who manage it. For office-wide BAAs, multiple PMs cover different technical areas. Match your technology to the right person by reading their published program descriptions and any conference talks or papers they've authored.

The initial email. Keep it under 200 words. No attachments. No NDA requests. Include:

1. One sentence on who you are
2. One sentence on what you've built
3. Two sentences on how it connects to the BAA's technical area
4. A question: "Would this be worth a white paper submission?"

Example (good):

Subject: Abstract inquiry -- [BAA number] -- adversarial ML for tactical networks

Dr. [PM Name], we're [Company], a 12-person startup that developed [specific technology] for commercial network security. Our approach uses [brief technical description] and has demonstrated [specific result] in production environments. We believe this could address [BAA technical area]'s need for [specific capability]. Would a white paper on this approach be relevant to your program? Happy to provide more detail if helpful.

What happens next. The PM will typically respond within 1-2 weeks. Possible outcomes:

• "Yes, submit a white paper" -- proceed with confidence
• "Interesting, but talk to Dr. [Other PM] instead" -- redirect (still positive)
• "Not aligned with current priorities" -- save yourself weeks of writing
• No response after 2 weeks -- follow up once, then move on

After the white paper. If your paper is selected, expect a phone call or meeting with the PM to discuss scope refinements. This is a negotiation -- the PM may want to expand or narrow your proposed work. Be flexible. The PM's input makes your full proposal stronger and signals that they're invested in your success.

Office-by-office: what each technical office funds (with examples)

DARPA's six technical offices each have distinct research cultures and different relevance for startups:

For AI/ML startups: Start with I2O. They have the most relevant BAAs and the most experience working with commercial AI companies.

For biotech startups: BTO is the primary office, but consider ARPA-H as well -- ARPA-H specifically funds health applications while BTO funds defense-relevant biology.

For hardware startups: MTO (for chips/sensors/photonics) or TTO (for autonomous systems and robotics). These offices have longer development timelines but larger awards.

For space/comms startups: STO is the primary office. Also consider DIU for more mature commercial space solutions and SpaceWERX for Space Force-specific applications.

Before and after: what a weak vs. strong abstract looks like

Weak abstract opening (actual pattern we see):

"In today's rapidly evolving threat landscape, cybersecurity challenges pose significant risks to national security infrastructure. Our company, CyberTech Solutions, has developed an innovative approach to addressing these challenges using cutting-edge artificial intelligence and machine learning techniques. We propose to leverage our proprietary platform to enhance defensive capabilities across the Department of Defense..."

What's wrong: Generic problem statement, buzzword-heavy, no specifics, reads like marketing copy. The PM learns nothing about what you'd actually do.

Strong abstract opening:

"Current network intrusion detection systems generate 11,000+ alerts per day per SOC, with a 97% false positive rate (SANS 2025 Survey). Analysts investigate less than 5% of alerts, meaning real threats routinely go undetected. We propose a causal reasoning engine that reduces false positives by 85% by modeling network behavior as directed acyclic graphs rather than statistical anomaly patterns. In production at 3 commercial SOCs, our system processes 2M events/second with a demonstrated 0.3% false positive rate on the CICIDS2017 benchmark."

What's right: Specific problem with numbers, clear technical approach, differentiated mechanism, verifiable results. The PM knows in 4 sentences what you'd do and why it's credible.

The multi-stage timeline: from first contact to contract

Most founders underestimate how long DARPA engagements take. Here's the realistic timeline:

The SBIR path is faster. DARPA SBIR topics have fixed review timelines -- typically 4-6 months from submission to award. But the scope is more constrained and the awards are smaller ($250K Phase I vs. potentially $5M+ for a BAA).

For more on the broader SBIR landscape

DARPA is one of many paths to non-dilutive funding. If you're exploring your options:

• The Startup Founder's Complete Guide to SBIR Grants -- covers all 11 agencies
• SBIR Funding Amounts by Agency -- the cross-agency comparison
• How to Win an SBIR Grant -- cross-agency tactical guide
• Should Your Startup Apply for Government Grants? -- the decision framework

Want help figuring out if DARPA is right for you?

DARPA engagement is relationship-driven, and the first conversation matters. If you think your technology has a defense or dual-use application but aren't sure how to position it, our Strategy Review can help you assess fit and identify the right PM and office to approach.