What DIU is, and what it is not
The Defense Innovation Unit was stood up in August 2015 as DIUx, a small outpost in Mountain View created under Secretary of Defense Ash Carter to buy commercial technology on commercial timelines. It dropped the experimental "x" in 2018. In April 2023 it was elevated again, and now reports directly to the Secretary. That reporting line matters more than the office square footage. It means DIU can carry a commercial capability toward a service program without asking a service acquisition executive for permission to exist.
DIU is not a research agency. It does not fund science, does not award grants, and does not run the SBIR program. If a company wants a Phase I to study whether an idea might work, DIU is the wrong door. DIU buys prototypes of things that already function, from companies that already sell them somewhere, on behalf of a military end user with a funded problem who wants a fielded answer inside a year or two.
The organization works out of Mountain View, Washington, Austin, Boston, and Chicago, and has extended reach through OnRamp Hub partnerships so companies outside the coastal technology centers have a local contact. Two aligned programs sit in the same structure: the National Security Innovation Network, which builds the academic and startup pipeline, and National Security Innovation Capital, created by the FY2021 National Defense Authorization Act to put non-dilutive money into dual-use hardware companies.
How the work is organized
DIU groups its buying into technology portfolios. Each has its own program managers, its own relationships with military customers, and its own reading of what industry can deliver right now. The portfolio map is the most useful document for a company deciding whether it belongs in the room.
- Artificial Intelligence and Machine Learning. Decision support, computer vision, predictive maintenance, and the data engineering underneath all three.
- Autonomy. Uncrewed air, ground, surface, and subsurface systems, plus the autonomy software that runs on them.
- Cyber & Telecommunications. Network defense, resilient communications, spectrum, and secure transport.
- Space. Commercial imagery, space domain awareness, on-orbit services, and ground software.
- Energy. Microgrids, storage, and installation resilience for fixed and expeditionary sites.
- Human Systems. Health, readiness, training, and talent management technology.
- Contested Logistics. Sustainment planning, distribution visibility, and maintenance under degraded conditions.
- Advanced Materials and Manufacturing. Production capacity, additive manufacturing, and supply chain qualification.
Software work is not confined to the AI portfolio. All eight areas generate requirements that are, underneath, a data problem: fusion, records, inference, or pipelines. A firm that reads only the AI listings will miss most of its own market.
Software and data content by DIU portfolio
Editorial weighting from published solicitations and practitioner reading. Illustrative, not a measured statistic.
Where the money actually sits
For most of its history DIU ran on a budget small enough to be a rounding error inside the Department. That changed. In fiscal year 2024 Congress appropriated close to a billion dollars to DIU, roughly ten times the prior level, with the largest share aimed at a Hedge Portfolio meant to buy proven commercial capability at scale rather than one prototype at a time. DIU stopped being purely a matchmaker and started being a buyer with its own wallet.
That wallet still rarely pays for the whole thing. The normal pattern is that a military customer brings the requirement and the transition money while DIU brings the authority, the speed, and part of the prototype funding. One adjacent pot matters as well: the Accelerate the Procurement and Fielding of Innovative Technologies program, administered by the Office of the Under Secretary of Defense for Acquisition and Sustainment, exists to buy production quantities of technologies that finished a prototype and had nowhere to go.
Two named efforts show how large the buys get. Replicator, announced in August 2023 by Deputy Secretary of Defense Kathleen Hicks, pushed thousands of attritable autonomous systems into the field on a two-year clock; Replicator 2, announced in September 2024, turned the same machinery on counter-uncrewed-aircraft defense of installations. Blue UAS, DIU's vetting effort for uncrewed aircraft that comply with Section 848 of the FY2020 National Defense Authorization Act and the American Security Drone Act of 2023, produces a cleared list the whole Department buys from. Neither is a software program on its face. Both consume enormous amounts of software.
The commercial solutions opening, in law and in practice
Almost everything DIU buys starts with a Commercial Solutions Opening. The authority began as a pilot in Section 879 of the FY2017 National Defense Authorization Act, Public Law 114-328, was made permanent in the FY2022 Act, and now lives at 10 U.S.C. § 3458. A CSO lets the government publish a problem rather than a specification, take short solution briefs instead of full proposals, evaluate them through peer or scientific review, and award to more than one company without a standard source selection.
In practice DIU publishes an Area of Interest, a short public document that states the operational problem, the environment, and the evaluation factors. It exists because a military customer asked for it and put money behind it. That is the fact most companies miss. An AOI is not market research. By the time it is public, someone in a program office has already decided they want to buy something.
Companies respond with a solution brief. Not a technical volume, not a fifty-page narrative, not a capability statement. A brief is typically about five pages or a short slide deck, due in two to four weeks, read by people who look at a great many of them. DIU has consistently targeted roughly sixty to ninety days from the close of an AOI to award, and reports its performance against that target publicly.
Area of interest to prototype award
The other transaction, and the clause that matters
A CSO can end in a Federal Acquisition Regulation Part 12 commercial contract, but DIU usually awards a prototype other transaction agreement under 10 U.S.C. § 4022, the provision formerly numbered 2371b. An OT is not a FAR contract. Most of the FAR does not apply. Cost accounting standards do not apply. Intellectual property terms are negotiated rather than imposed, which is why commercial companies sign one when they would never sign a standard defense contract.
Section 4022(d) sets the entry conditions. At least one nontraditional defense contractor or nonprofit research institution must participate to a significant extent, or all significant participants must be small businesses or nontraditional contractors, or at least one third of the cost must come from non-federal sources. A nontraditional defense contractor is defined at 10 U.S.C. § 3014 as an entity that has not performed, for at least one year, any contract subject to full cost accounting standards coverage. Many capable engineering firms qualify without knowing it.
The clause that decides whether any of this was worth doing is 10 U.S.C. § 4022(f). If competitive procedures were used to award the prototype, and the participants successfully complete it, the government may award a follow-on production contract or transaction without further competition. Read that twice. The entire strategic value of a DIU prototype is that it is a legal on-ramp to sole-source production, so everything in the prototype phase should be built to satisfy the words "successfully completed."
Scale is bounded by statute. Prototype projects above one hundred million dollars require a written determination by a designated senior official and notice to the congressional defense committees; projects above five hundred million dollars require determination at the Under Secretary level. Most DIU prototypes sit far below both lines, which is part of why the awards move quickly.
| Path | Instrument | What it rewards | Transition route |
|---|---|---|---|
| DIU commercial solutions opening | Prototype OT under 10 U.S.C. 4022 | Working product, commercial traction, a named military user | Follow-on production without further competition, 4022(f) |
| SBIR Phase I and II | FAR contract or grant | Technical merit, feasibility, commercialization plan | Phase III sole source under 15 U.S.C. 638(r) |
| Consortium OT | Prototype OT through a managed consortium | Membership, teaming, responsiveness to a call | Follow-on production under the same statute |
| FAR Part 12 commercial | Commercial item contract | Catalog pricing, established commercial sales | Options, then recompete or a schedule order |
| GSA schedule order | Task order against a multiple award schedule | Pre-negotiated labor rates and past performance | Follow-on task orders |
The data DIU holds and publishes
Open and closed areas of interest. The solicitations page on diu.mil carries every open AOI with its due date, and closed AOIs stay readable. Twelve months of closed AOIs is the best available training set for what a DIU problem statement looks like, how narrow the evaluation factors get, and which portfolios are actually spending.
Annual reports. DIU publishes an annual report with portfolio activity, award counts, and transition statistics. Its reported transition rate, the share of completed prototypes that reach a production agreement or a program of record, has run in the vicinity of one third to forty percent. That is high for defense prototyping and low enough to be honest about.
Blue UAS lists. The cleared list and the framework list are both public, and both matter beyond aviation, because they define which airframes and components the rest of the Department is permitted to buy.
Award announcements. Other transactions are not captured in the Federal Procurement Data System the way FAR contracts are, and the Government Accountability Office has reported repeatedly that departmental data on other transaction agreements is incomplete. DIU press releases and portfolio pages are often the only public trace of who won what, so competitive research on DIU cannot be done from USAspending alone.
What gates entry
Speed at DIU is real, but it sits on top of the same compliance floor as everything else the Department buys. Five gates decide whether a company can be considered at all.
- An active SAM.gov registration with a Unique Entity ID and CAGE code. Other transactions still require it, and a lapsed registration kills an award that was otherwise ready to sign.
- Controlled unclassified information handling. Where DFARS 252.204-7012 is flowed into the agreement, NIST SP 800-171 controls and a system security plan come with it. The Cybersecurity Maturity Model Certification program rule at 32 CFR part 170 took effect December 16, 2024, and the companion contract-clause rule began phasing requirements into solicitations through 2025.
- Supply chain prohibitions. Section 889 of the FY2019 National Defense Authorization Act, Public Law 115-232, and its clause at FAR 52.204-25 bar covered telecommunications equipment. Section 848 and the American Security Drone Act govern uncrewed systems and their components.
- Export control. Defense articles and technical data fall under the International Traffic in Arms Regulations at 22 CFR parts 120 through 130, with registration through the Directorate of Defense Trade Controls. Access to militarily critical technical data runs through the Joint Certification Program and DD Form 2345, a real gate that stops firms at the document-request stage.
- Ownership and capital screening. Foreign ownership, control, or influence is examined, and the capital table of a venture-backed company is part of that diligence. Classified work adds facility and personnel clearances under the National Industrial Security Program rule at 32 CFR part 117.
Most areas of interest are unclassified at the solution brief stage, and many prototypes execute at the controlled unclassified level. The clearance question usually arrives later, attached to the specific data the customer needs processed, which is a friendlier sequence for a small firm than the reverse.
Where a small engineering firm realistically fits
Prime on a software or data area of interest. This is genuinely open. Software AOIs require no manufacturing capacity, no capital equipment, and no production line. The evaluation turns on whether the thing works, whether the team can execute, and whether the customer wants it. DIU was built to award nontraditional performers, and a brief from a small engineering firm competes on the same page as one from a company with a thousand employees.
Technical partner to a commercial product company. Many companies entering the defense market have a working product and no idea how to move data through an accredited boundary, satisfy a system security plan, run inference where the data cannot leave, or produce the evidence a program office needs to accept a model. That gap is an engineering subcontract, and it is where a firm with federal delivery experience earns its seat on a strong team.
Transition engineering behind a prototype that already won. A prototype in month fourteen of eighteen with a production decision ahead of it is a company under pressure. Integration into a system of record, authority to operate work, pipeline hardening, and test evidence are what "successfully completed" is made of, and they are exactly the work that gets subcontracted.
What does not work is arriving with a capability statement and no product. DIU reads briefs against an operational problem, so a company that cannot show something running has nothing to submit.
What a strong solution brief does
It answers the AOI's evaluation factors in the AOI's own order and vocabulary. Reviewers read dozens of briefs against a written rubric. A brief organized around the company's preferred story instead of the government's stated factors forces the reviewer to do the mapping, and reviewers do not do the mapping.
It shows a working artifact: a screenshot of a real interface, a measured result on a named dataset, a latency number on named hardware. Specific numbers beat adjectives, and a claim tied to a benchmark someone else controls beats a claim tied to one you generated yourself.
It names the operational fit. Who uses this, in what unit, doing what task, and what is different about their day afterward. It is also honest about maturity, because overstating readiness is the fastest way to lose a prototype that would otherwise have been won. State what exists, state what the prototype will build, and let the gap be visible and small.
The first step
Confirm the SAM.gov registration is active, with the Unique Entity ID and CAGE code current and the representations complete. Then open the solicitations page on diu.mil, read every area of interest closed in the last twelve months, and sort them by portfolio. That single afternoon of reading will say more about a firm's odds than any amount of conference attendance. Pick one open AOI where the fit is genuine, write the brief to its evaluation factors, and submit before the deadline rather than on it.
If nothing open fits this quarter, get onto a team that has already won one. Prototype performers need federal data engineering, accreditation support, and production-grade machine learning far more often than they need another partner with a slide deck.
Common questions about how DIU buys
Does DIU award SBIR contracts?
No. DIU buys through commercial solutions openings and other transaction agreements. SBIR is a separate statutory program run by the components. The two combine well: a firm holding a completed SBIR Phase II can be sole-sourced into Phase III production under 15 U.S.C. 638(r), and that authority can carry work a DIU prototype started.
Does a company need a security clearance to submit a solution brief?
Usually not. Most areas of interest are unclassified at the brief stage. Clearance and facility requirements attach later, to specific data or specific customer environments, and are stated in the AOI when they apply.
Is an other transaction worse for intellectual property than a FAR contract?
Generally the opposite. OT data rights are negotiated between the parties rather than dictated by standard DFARS clauses, which is one reason commercial companies accept them. Negotiate deliberately, define the boundary between background and foreground technology in writing, and get the terms reviewed before signature.
Frequently asked questions
A solicitation method codified at 10 U.S.C. 3458 that lets the government publish a problem instead of a specification, accept short solution briefs, evaluate them through peer or scientific review, and award to one or more companies without a conventional source selection. It can result in a commercial contract or an other transaction agreement.
Under 10 U.S.C. 3014, an entity that has not performed any contract or subcontract subject to full cost accounting standards coverage for at least one year prior. Nontraditional participation is one of the conditions that makes a prototype other transaction available under 10 U.S.C. 4022(d).
Under 10 U.S.C. 4022(f), if competitive procedures were used for the prototype and the participants successfully completed it, the government may award a follow-on production contract or transaction without further competition. Documenting successful completion against the agreed prototype objectives is the practical requirement.
DIU targets roughly sixty to ninety days from the close of an area of interest to award and reports its performance publicly. The full path from AOI publication through solution brief, pitch, proposal, and negotiation typically runs three to five months.
DIU annual reports, portfolio pages, and press releases. Other transactions are not recorded in the Federal Procurement Data System the way Federal Acquisition Regulation contracts are, and the Government Accountability Office has reported that departmental data on other transaction agreements remains incomplete, so public award research has real gaps.