The supplier base is attracting new firms, but entry, production transition and repeat revenue measure different things. Reading them together gives a more useful picture of commercial adoption.
A CSIS report published on 6 July 2026 describes renewed growth in the number of firms entering the US defence industrial base. Drawing on federal contract data, authors Jerry McGinn and A.J. Dilts report roughly 5,000 new entrants in each of fiscal 2024 and fiscal 2025 after several years of decline. They also report US$122.6 billion in obligations to nontraditional vendors in fiscal 2025, compared with US$60.9 billion in 2015. These are participation and funding measures, with different populations and units. Neither directly measures recurring product revenue. The CSIS report
For a founder, the next question concerns the path from that first recorded transaction to a business that customers choose again. For a potential industrial partner, it concerns how much of an entrant's apparent traction can support a delivery commitment. Both require a closer view than a market-wide count.
The practical interpretation developed here is BDI's analysis. CSIS provides evidence that entry has increased; it does not provide a survival rate for every new supplier or establish the economics of a particular company.
An entrant is a business identified under a research definition. An award is a transaction or agreement. A project can involve several awards, and one supplier can serve several projects. Combining these units produces attractive but ambiguous growth claims.
Consider a hypothetical software company that receives an initial evaluation agreement, an extension of that evaluation and a separately funded integration task from the same organisation. It has three funding events. It may still have one customer evaluating one product. Treating the events as three customer wins would exaggerate diversification; treating all three as meaningless would overlook useful evidence that the customer continues to fund work.
The more revealing account states what changed. Did the integration task connect the software to the customer's ordinary workflow? Was its purpose further experimentation or a supported service? Did the customer accept the deliverable? Is the next purchase controlled by the same budget owner? Those distinctions explain why a second transaction may represent substantial progress in one case and little change in another.
Obligations add a further dimension. They identify committed funding in the relevant records, rather than the timing of a supplier's revenue recognition or cash receipts. A commercial assessment should retain that label rather than relabelling every obligation as sales.
GAO's February 2025 review of the Defense Innovation Unit reported 450 prototype awards over fiscal 2016–2023. DIU reported that 51% of completed prototypes transitioned to production. The denominator is completed prototypes, rather than all companies, all awards or every project that started. GAO also identified weaknesses in DIU's ability to assess progress against its newer strategic objectives. GAO's DIU review
This is a useful comparison with the CSIS entrant count, not a conversion factor to apply to it. The studies cover different organisations, periods and populations. Multiplying the entrant figure by DIU's transition percentage would create an unsupported estimate of successful businesses.
A second GAO study, published in September 2025, examined small-business research programmes. It used progression from Phase I to Phase II as an indicator of movement towards commercialisation. That progression concerns research and prototype development; it is not equivalent to sustained commercial sales. The study also found that agencies labelled open and conventional topics inconsistently, complicating comparisons between categories. GAO's award-data study
The common lesson is methodological. Before comparing success rates, identify the event being counted and the group that had an opportunity to experience it.
A business with its first award late in fiscal 2025 has had less time to obtain repeat work than one entering early in fiscal 2024. A simple comparison of their follow-on awards would mix commercial performance with time available.
A useful cohort view follows companies from a defined first event and compares them after similar observation periods. It also records transactions that remain unresolved. An unfinished evaluation should remain unfinished in the dataset, rather than being silently classified as a failure or assumed to become a production award.
This matters especially when selecting companies for partnership research. An early entrant with a small accepted delivery and a funded expansion may offer clearer evidence than a newer entrant with a much larger headline ceiling. Yet the newer entrant may have a better product or broader future opportunity. The records support a description of demonstrated progress; they do not settle every judgement about potential.
Corporate changes need similar care. A new legal entity created through a restructuring can look different from a newly established product business. An acquisition can move repeat work into another parent's reporting. Tracking the operating business alongside its public identifiers reduces false conclusions about entry and disappearance.
A repeat purchase is commercially stronger when delivery becomes more predictable. If each new customer requires a largely separate engineering effort, growing revenue may conceal a services business with limited product reuse.
For the hypothetical software supplier, a second customer using the same supported release provides different evidence from a second customer funding an incompatible version. Both can generate revenue. The first may support a common product roadmap; the second may increase the number of engineering obligations the company must sustain.
The decisive evidence sits in work performed: the proportion of integration reused, the support burden after acceptance, the number of customer-specific releases and the time needed to introduce an upgrade. Public announcements rarely disclose all of this. A publication can identify the information gap without inventing a margin estimate.
Hardware businesses face an analogous distinction between making another prototype and repeating a controlled production process. Supplier qualification, acceptance records and dependable component availability can matter more to the next contract than a previous demonstration's visibility.
Established groups may provide customer access, integration resources and support infrastructure. Whether that improves the entrant's business depends on the work and commercial rights retained.
A startup can become a reliable recurring supplier to one prime without developing direct government sales. That is a legitimate business model. Its risks include customer concentration and dependence on the prime's programme timing. Conversely, a joint announcement with several large groups may provide little recurring revenue if no funded integration or support responsibility follows.
Our analysis of Ifri's startup–prime partnership research examines the allocation of that work. The related discussion of joint procurement and commercial standardisation considers whether several buyers create a common deliverable or several national variants.
The same reasoning applies outside strictly military customers. Exail's repeat order from OMS illustrates why the identity of a repeat customer and the purpose of a second purchase deserve attention. A civil market can provide product and production experience without proving military acceptance.
The most informative public record follows a small number of connected facts: the initial funded use, the accepted deliverable where disclosed, the next funded use and the buyer responsible for it. Dates and original source links make the sequence reviewable.
Where amounts are available, they should retain their stated basis. An award ceiling, funded obligation, group order intake and recognised revenue belong in separate fields. Where outcomes are undisclosed, a blank is more useful than a fabricated completion estimate.
This approach also makes editorial coverage more durable. A later purchase can update the company's commercial history without rewriting the meaning of the earlier award. Readers can see whether the product gained another customer, expanded at an existing customer or remained within development work.
CSIS's figures make the new-entrant population worth watching. The strongest business intelligence comes from following what those firms subsequently deliver, who purchases again and whether the work becomes easier to repeat. That turns an entry statistic into evidence that founders, partners and commercial teams can actually use.
The public CSIS HTML report was accessible and reviewed on 6 September 2026, with the reported entrant count checked in its New Entrants and Nontraditional Firms section. Federal contract data were not independently recomputed. Conclusions about repeat business and supplier economics are BDI analysis.
Suggest a correction