The Question Behind the Pipeline
Chips are capital-intensive. The traditional path for a fabless startup is venture pre-seed at 20 to 30% dilution before any silicon exists, and another round before tape-out. The founder ends up minority-owning the thing they invented, has board obligations to people whose timelines do not match silicon's, and is on the venture treadmill from day one.
The other path is non-dilutive: SBIR, STTR, NSF, ONR, DARPA, AFWERX, and the alphabet soup of federal programs that exist specifically to fund early-stage R&D in companies the venture market either ignores or values badly. Phase I Base plus Phase I Option plus Phase II for a Navy STTR can reach $2.24M without dilution. Add a parallel NSF SBIR ($305K Phase I, $1M Phase II) for the civilian side and a single solo founder can be funded through Phase II silicon without selling a share.
The catch is that you cannot just submit. To be eligible to submit, you have to be a particular kind of registered entity in a particular kind of compliance posture, and the chain of prerequisites that gets you there is long, opaque, and full of dependencies you only discover after you have hit them.
This post is a walk through that chain. Not a complaint and not a victory lap. Documentation, written down so the next solo founder can navigate it faster than I did.
March 5 to June 3
Hard dates so the rest of the post has scale.
Lattrex was incorporated in Clerky on March 5, 2026. As of today, June 8, 2026, the company has an active SAM.gov registration, a CAGE code, a DSIP account in good standing, a PIEE account with sub-accounts in SPRS, an active research partnership with Virginia Tech under a signed institutional MNDA, a CMMC Level 2 self-assessment uploaded to SPRS, a foreign-affiliations disclosure on file, a Fraud Waste and Abuse training certification, a signed Allocation of Rights agreement with the research institution, a Fundamental Research Disclosure, a 10-page Technical Volume, a Cost Volume that clears every cap to the dollar, and a Navy STTR Phase I proposal in the portal under topic DON26TZ01-NV007.
That is 90 days. Solo. From "no company exists" to "federal grant proposal with a university partner is in." If you take nothing else from this post, take that the timeline is doable. The rest of this post is what is between those two endpoints.
The Interdependency Chain
The pipeline reads like a list, but it lives like a dependency graph. Every node blocks the next one, and you do not find out a node is blocked until you are standing in front of it asking why nothing is happening.
Walk an example in order. To submit a federal proposal, I needed a CAGE code. To get a CAGE code, I needed a completed SAM.gov registration. To register on SAM.gov, I needed business banking, an EIN, and a UEI. To open business banking, I needed the company registered to do business in Tennessee. To register in Tennessee, I needed a certificate of good standing from Delaware, because Lattrex is a Delaware C-corp operating in Tennessee. To get the Delaware document, I submitted the request and waited.
Three weeks. No response, no status, no acknowledgment. Eventually I learned that the request was sitting in Delaware's normal-processing queue with no published ETA, and that if I had paid $50 for expedited handling I could have had the document the next day. The fast-track option existed the whole time. It was not on the form I had filled out. It was visible only if you already knew to look for it.
That single chain is the whole post in miniature. Each prerequisite is invisible until the previous one clears, and the system quietly offers a $50 fast path it never tells you about. Every founder who has done this has a version of this story. Non-founders have no idea it exists.
The reframe to land is that this is not a complaint. It is what the gauntlet actually looks like, and the point is that it can be cleared with discipline. Nobody hands you a map. You assemble the map yourself, one blocked dependency at a time.
The Paperwork Stack, In Order
The full list. I am numbering them because order matters: most are gating prerequisites for the next one, and skipping forward is how you discover you missed something at the bottom of a queue 4 weeks ago.
1. Form the company. Delaware C-corp for the legal treatment and equity flexibility that federal funding and any future equity transaction will want. Not an LLC, not an S-corp. The form-of-entity decision is its own rabbit hole; the short version is that Delaware C-corp is the path of least friction for everything downstream.
2. Foreign-qualify in your operating state. Lattrex is a Delaware corporation operating in Tennessee. To do business in Tennessee, the company has to be foreign-qualified there, which means filing with the Tennessee Secretary of State and getting authorization to operate as an out-of-state entity. This is the dependency that triggered the 3-week Delaware certificate-of-good-standing wait described above.
3. EIN. IRS Form SS-4. Federal tax ID. Free, fast, prerequisite for everything financial that follows.
4. Business banking. Requires EIN and incorporation docs. Has to be the corporate entity's account, not a personal account labeled "business." Banks have their own onboarding queues. Mine took 2 weeks to fully provision wire transfers, which was the binding constraint on later cash flow.
5. UEI. Unique Entity Identifier, issued by SAM.gov directly. Used to be DUNS issued by Dun and Bradstreet; the federal system moved off DUNS a few years ago and the UEI is now native to SAM. Free. Prerequisite for the SAM registration itself.
6. SAM.gov registration. The System for Award Management is the master registration for any federal contracts. It takes weeks to complete and to clear approval. The form asks for UEI, EIN, business banking, incorporation paperwork, NAICS codes, business size representations, and an officer who can attest under penalty of perjury that everything submitted is accurate. The certification is not a formality. False statements have actual criminal consequences. Annual renewal required.
7. CAGE code. Commercial and Government Entity code, assigned during SAM registration by DLA. Required for any DoD work. You do not apply for it separately; it falls out of SAM registration as a side effect, but it is the actual artifact most DoD systems index against.
8. DSIP account. Defense SBIR/STTR Innovation Portal. Where DoD SBIR/STTR proposals are submitted. Account creation requires a SAM-registered entity in good standing. You cannot create a DSIP account before the SAM registration clears, which means DSIP setup is gated on the multi-week SAM queue.
9. PIEE account. Procurement Integrated Enterprise Environment. Where DoD contract administration happens once you have an award. Multiple sub-accounts live inside PIEE: SPRS, WAWF, others depending on the contract. You set up PIEE before award because some pre-award compliance gates submit through it.
10. The two SPRS submissions that look like the same thing. This one almost nobody writes down clearly. Inside PIEE, the Supplier Performance Risk System collects two related but distinct submissions:
The first is the NIST SP 800-171 self-assessment score (DFARS 252.204-7019 and -7020). You compute your score against 110 controls and upload it. The system accepts any number, including a negative one. Mine started deeply negative because I had not yet provisioned the secure environment. That is fine. You upload it anyway, and that upload is what makes you eligible to receive a contract conditional on improving it.
The second is the CMMC Level 2 affirmation (DFARS 252.204-7021). This is a separate submission, governed by a separate regulation, with different acceptance rules. You literally cannot complete this affirmation until your self-assessment score is above a threshold that you cannot hit at the start.
So you have two protocols, two regulations, that look like the same requirement and are not. One you must do immediately; one you eventually must do but physically cannot do yet. Working out which one gates submission (the first) versus which one is a forward commitment (the second) took real reading. This is the kind of thing that is obvious only after you have already been confused by it.
11. CMMC Level 2 self-assessment. 110 controls from NIST SP 800-171. Self-implemented, with a System Security Plan, a Plan of Action and Milestones, an enclave definition that says exactly where Controlled Unclassified Information is allowed to live, and a stack of supporting security policies. Documentation per 32 CFR 170.16. POA&M per 32 CFR 170.21. The CMMC documentation alone runs well over 100 pages of artifacts and has its own working directory. This is the single largest non-technical workload in the pipeline.
12. Sponsoring SYSCOM identification. For DON26TZ01-NV007 the sponsor is ONR. Every signal points the wrong way: the technical points of contact sit physically at NUWC Newport, every email comes from a NUWC address, the eventual integration site is NUWC Newport. So every instinct tells you the sponsoring command is NUWC. It is not. The sponsor is ONR, and you only find that by reading Table 2 of the binding BAA document. Getting it wrong in the proposal text is a credibility hit with reviewers who know the org chart cold. It is the kind of error that is invisible going in and obvious in hindsight.
13. Topic question period engagement. Before submission opens, DoD allows technical questions to the TPOC on each topic. Smart questions are free intelligence about how the sponsor thinks about the problem. Dumb questions are reputational damage with the same TPOC who will eventually score your proposal. I asked 6 across 3 weeks. Most were useful. One was over-reach, asking the TPOC to constrain an architectural interpretation, and she correctly declined to be drawn into that. Lesson: questions are tools, not interrogation. Use them for what the sponsor is allowed to answer in writing.
14. Allocation of Rights agreement. STTR requires a documented IP allocation between the small business (Lattrex) and the research institution (Virginia Tech MICS). Navy provides a mandatory template at navysbir.com. The template is a 4-page model with fill-in-the-blank lines roughly a quarter-inch tall, into which you are supposed to fit full business addresses, names and addresses of both parties, and descriptions of background and foreground intellectual property. It physically cannot hold what it asks for. The tell that you are allowed to reformat it preserving structure and clause numbering, but expanding the fields to legible size, is a parenthetical on the template that reads "(This is only a model)." Even my university research partner, who has run federal grants for years, was unsure whether the AoR had to be signed and submitted with the proposal or only after award. The answer is that it is required in the submission package, and a missing one is a compliance-review rejection. If a seasoned PI is unsure, a first-time solo founder has no chance of knowing it cold.
15. Fundamental Research Disclosure. Required because Lattrex is partnered with a university. Filed in Volume 5.
16. Foreign affiliations and relationships disclosure. DSIP Volume 7. Mandatory. Foreign Person status, foreign ownership, foreign funding, foreign business relationships, foreign academic ties. False statements have actual criminal consequences. Most of the fields for Lattrex were "none," which is the easy case. For founders with international academic backgrounds or pre-Lattrex foreign employment, this volume gets harder.
17. Fraud, Waste, and Abuse Training certification. DSIP Volume 6. Mandatory. A free online training that takes 20 minutes, but missing the certification kills the submission. The first time I built my submission package I almost forgot it.
18. Company Commercialization Report. DSIP Volume 4. For a first-time submitter, the form is short and largely "no prior awards." For repeat submitters, history of previous awards and commercialization metrics.
19. Cover Sheet. DSIP Volume 1. Auto-populated from your DSIP account if registration is done right. The address has to match SAM, the CAGE record, and the incorporation paperwork exactly. Any mismatch is a rejection vector. If your address changed mid-pipeline, you need to fix it everywhere before you can submit anywhere.
20. Technical Volume. DSIP Volume 2. 10 pages hard cap including Phase I Base and Phase I Option. Navy template at navysbir.com, mandatory format. The only volume that contains the actual technical pitch. Everything else is paperwork that exists to qualify you to submit this single document.
21. Cost Volume. DSIP Volume 3. A workbook download. Phase I Base cap $140K. Phase I Option cap $100K. Combined Phase I cap $240K. TABA up to $6,500 in addition to the Base cap. Cost-share field must be exactly $0.00 for Navy SBIR Phase I. STTR work-share rules: small-business minimum 40% of total cost, research-institution minimum 30%, both holding separately in Base and in Option, not just in combination.
The Gates Are Binary
DSIP enforces every cap to the dollar. There is no almost-compliant.
When I reconciled the Phase I Option budget against Virginia Tech's subaward numbers, the combined Lattrex plus VT total came to $100,267. The Option cap is $100,000. I was $267 over. On a $240,000 proposal, $267 is a rounding error in any sane accounting. Compliance review is not sane accounting. $267 over is rejected, full stop, before a human reads the technical volume.
The fix was trimming my own founder hours from 520 to 517 to drop back under the line. Three hours of my time, $267, between rejected and reviewed.
The lesson is that the system does not grade on a curve and you can be $267 and one unchecked cell away from rejection without knowing it. Reconcile every cap to the penny before you click submit. Then reconcile again.
Cite the Binding Document
Partway through the cost volume I was citing a 7% profit/fee cap as a hard Navy STTR requirement. When I went to source it for the cost narrative, I searched the actual binding BAA document end to end. The 7% figure is not in it. The number came from general SBIR and FAR background knowledge floating around the SBIR forums and the consulting ecosystem, not from the solicitation that actually governed my proposal. The real fee ceiling for a fixed-price contract comes from FAR weighted-guidelines analysis, not a flat BAA number.
That was a small error and an honest one. The lesson scales. The SBIR forums and the conventional wisdom are full of numbers that are not in your BAA. They are right for someone else's solicitation under a different program in a different year, or they were right 3 cycles ago and the BAA was updated without the forums catching up. Read the binding document. Cite the binding document. When somebody on a forum tells you a number, ask which BAA they are quoting and check yours.
The Real Ratio
The technical volume is 10 pages. That is the only document where you make the actual pitch for the research you want funded.
The eligibility paperwork that lets you submit those 10 pages includes the CMMC documentation (well over 100 pages of artifacts), the foreign-affiliations disclosure, the Fraud Waste and Abuse certification, the Company Commercialization Report, the Cost Volume workbook, the Allocation of Rights agreement, the Fundamental Research Disclosure, the Cover Sheet, the SAM record, the SPRS submissions, the institutional and personal NDAs that govern the research partnership, and the supporting documents volume.
The ratio of eligibility paperwork to technical pitch is closer to 15 or 20 to 1, not 5 or 10 to 1. Said with specifics rather than as a multiplier, the asymmetry is staggering. The interesting part of the proposal, the part that actually argues why Kyttar should exist, is roughly 5% of the total submission package.
What the Payouts Look Like
For context on what is at the end of the gauntlet.
Phase I Base ($140K). 6 months. Partial founder salary and a research-institution sub-contract. Lattrex's planned split with Virginia Tech is roughly 60% Lattrex, 35% MICS, 5% materials and travel.
Phase I Option ($100K). 6 more months. Exercised at sponsor discretion based on Phase I performance.
Phase II ($2M total). 18 months. This is the silicon money. Multi-project shuttle tape-out at an advanced node becomes financially feasible at Phase II scale, which is the bridge from prototype to demonstrator on the path to Phase III.
Phase III. Direct or sole-source follow-on contracts. Production scaling. Actual customer revenue. The thing all the prior phases were in service of.
In parallel, the NSF SBIR civilian track is $305K Phase I and $1M Phase II, with no CMMC and no FCL requirements. NSF SBIR is on a separate review timeline and a separate IP regime, so winning one does not block the other.
If both tracks award, Lattrex is funded through 2028 without selling a share.
What This Signals
It is fashionable to dismiss federal contracting as paperwork theater, and a non-trivial fraction of it is. But the discipline required to clear the eligibility pipeline is the same discipline required to ship product to a federal customer in Phase III. The pipeline is not separate from the engineering; it is a filter for whether the company can execute on the operational half of the job.
Most chip startups die between "interesting architecture" and "ships product," and they usually do not die on the engineering. They die on the operational execution: the missed compliance gate, the misread BAA, the un-renewed SAM registration, the conflated regulation, the $267 budget overage that nobody caught. The federal funding pipeline is structured to filter out companies that cannot handle that. Going through it is its own form of validation.
What Comes Next
The Navy STTR Phase I proposal is in. Decision in roughly 90 days. The NSF SBIR Pitch has also been submitted and waiting for feedback on whether or not to move on to the full proposal. The chip is at fab, with first silicon expected back the first week of November.
Between now and November my work shifts from engineering to customer discoveryand foundry conversations for a potential Phase II contract. The next post on this blog is going to be about customer discovery: what I am trying to figure out about who needs Kyttar and what problems it actually solves.
The chip startups that die do not usually die on the engineering. They die on the operational execution, the part nobody photographs. Going from incorporation to a submitted federal proposal in 3 months, alone, is the same muscle that ships silicon. On to the next challenge.