Defense Tech VC Doubled to $12.3B. The Cleared Engineer Pool Is 41.

You just closed a Series A to build autonomous ISR, counter-drone, or hypersonics software. Your board deck says you need 30 cleared engineers by Q4. So does everyone else's, and the math does not work the way your investors think it does.

FT/PitchBook data shows defense tech startups raised $12.3B in H1 2026, nearly double H1 2025 and already ahead of the full $9.95B raised in all of 2025. Anduril's $5B Series H at a $61B valuation alone accounts for roughly half of the US total. But the input that actually gates hiring, the population of Americans holding an active security clearance, grew at roughly the same low-single-digit pace it has for a decade. This is a structural squeeze, not a market cycle, and most newly funded founders have no playbook for it.

The capital-to-clearance ratio just broke

Defense tech VC roughly doubled year over year while the cleared talent base grew low single digits, so every marginal dollar is chasing the same shallow pool. The bottleneck is not recruiter headcount or comp bands. It is DCSA background-investigation throughput and ODNI adjudication policy, both of which move on federal timelines, not venture ones.

Here is the dataset every defense tech founder should have taped to their monitor:

The two rows that matter are the last two. In Refolk's index of professional profiles, only 869 US software, embedded, and firmware engineers surface under standard sourcing filters against the Defense & Space industry. Cross-filter for TS/SCI language plus embedded systems or C++, and the number collapses to 41 discoverable profiles nationwide.

41

Meanwhile Anduril, Shield AI ($1.5B Series G at $12.7B), Saronic ($1.75B Series D at $9.25B), Mach Industries, Applied Intuition, and every prime from Lockheed to Booz Allen are hiring against the same 41.

Why "3.7 million cleared Americans" is the wrong denominator

The 3.7M figure ODNI publishes counts every active or current eligibility including military, IC analysts, and non-technical contractor staff, so it wildly overstates the engineering supply available to a startup. When founders tell investors "there are millions of cleared people out there," they are conflating a headcount ledger with a talent market.

Break the funnel down:

• 3.7M total active/current clearance eligibilities (FY2023, ODNI)
• 1.26M hold Top Secret
• Roughly 800K to 900K are TS/SCI-eligible (fewer than one in four of the total)
• A small minority of those are software engineers with production skills (embedded, C++, kernel, RF, sim)
• Of those, in Refolk's index, only 41 self-identify publicly with the combined TS/SCI plus embedded/C++ signal

The gap between the TS-eligible universe and the publicly discoverable software cleared talent is roughly 30,000 to 1. That is the hidden pool problem in one ratio. Cleared engineers, especially TS/SCI holders at IC agency programs, are actively discouraged from publishing employer, program, or clearance details on public profiles. So the inbound recruiting motion that works fine for a fintech Series B is systematically blind to 99%+ of the actual market.

The primes are not your only competition anymore

The traditional cleared talent buyers, Northrop Grumman, Lockheed Martin, Boeing, Raytheon, Booz Allen, CACI, have been joined by a new tier of well-capitalized startups with modern comp bands and equity upside. And in a twist, primes like Lockheed and Boeing are now co-investing alongside traditional VCs in the same startups poaching their engineers. JPMorgan's Security and Resiliency Initiative co-led Shield AI's Series G alongside Advent International.

You are not competing with a prime for cleared engineers. You are competing with your prime's own venture arm.

DCSA throughput is the real cap table

Clearance supply is governed by the Defense Counterintelligence and Security Agency's investigation cadence and ODNI adjudication policy, and both add net eligibilities in the low single digits per year. That is what "the pool didn't double" actually means at the mechanism level.

DCSA ran more than 1.7 million personnel-security actions against the cleared base in FY2023, but that number is dominated by periodic reinvestigations, upgrades, and continuous vetting, not net new grants. Historical ODNI trend data shows the total cleared population swinging between roughly minus 1.2% (FY2017) and plus 4.2% (FY2019) year over year. At the FY2019 peak, that is roughly 150K to 170K net new eligibilities added to a 3.7M base, most of them military and non-engineering.

Compare that to the roughly $6B increment in H1 2026 defense VC spend versus the full prior year. If you assume even 20% of that new capital funds engineering payroll, the sector is trying to buy tens of thousands of engineer-years against a supply function that can add a rounding error's worth of net cleared bodies. The bottleneck is not sourcing effort. It is federal vetting throughput.

Clearance supply is set by DCSA cadence, not by how hard you recruit. Capital cannot buy around a policy input.

Reciprocity is not "portable clearance"

The other misconception on founder Twitter is that a TS/SCI hire from Lockheed is drop-in ready. It is not. Three frictions kill the "we'll just hire cleared people" plan:

• SCI read-ins are per-agency. The receiving agency must indoctrinate the holder for each new SCI compartment, even if the underlying TS transfers cleanly.
• Polygraph requirements vary. A CI polygraph from one agency does not satisfy a full-scope lifestyle polygraph at another. Startups working sensitive DoD or IC programs often need the fuller poly.
• The 24-month break-in-service rule. Any gap longer than 24 months in cleared service requires a full reinvestigation, which is exactly the population (recent big-tech leavers, ex-DoD who went commercial) most startups target.

Budget 4 to 12 weeks of dead time per hire even when the clearance nominally transfers. That is real burn against a Series A runway.

Where the 869 discoverable engineers actually live

Refolk's index surfaces 869 US software, embedded, and firmware engineers in Defense & Space under standard sourcing filters, and they concentrate in a handful of programs and metros. This is the addressable outbound universe before you even get to clearance status, and it tells you where to spend recruiter hours.

The concentrations that repeat in the data:

• Applied Research Laboratories at UT Austin (ARL-UT) shows up as the single largest cleared engineering feeder
• NAVSEA Warfare Centers across Crane, Dahlgren, Newport, Port Hueneme
• Small primes and defense startups including TurbineOne, Dynetics, and STAR Dynamics
• Geographic clusters: Austin TX, Huntsville AL, Greater Boston, DC-Baltimore, Colorado Springs

When you cross-filter to the 41 publicly TS/SCI-tagged embedded/C++ profiles, the top employers shift toward US Army, Blue Origin, RTX, Deloitte, and Colossal. That mix tells you something useful: the visible TS/SCI engineers are disproportionately at organizations where clearance disclosure is culturally acceptable (Army, primes) or where the employer brand pulls people to list it (Blue Origin). At IC agencies and their direct contractors, disclosure is rare.

This is the exact gap Refolk closes for defense founders: describe the person in plain English, including proxy signals for clearance (employer history, program mentions, .mil email domains in commits, DoD SkillBridge participation), and get a ranked shortlist instead of a boolean-search dead end.

The playbook for sourcing the invisible 99%

Since clearance status is invisible to inbound sourcing, cleared engineer hiring has to be outbound and signal-based, and it should lean on veteran-transition pipelines rather than LinkedIn Recruiter boolean. This is where most newly funded founders are underinvested.

The channels and signals that actually work:

1. DoD SkillBridge, the transitioning-military internship program. Candidates arrive with active clearances and authorized industry work before ETS. They surface in Refolk's index repeatedly.
2. Veteran hiring networks: Corporate Gray, Hire Heroes USA. Higher signal density than general job boards for cleared candidates.
3. ClearanceJobs, which reported 25% YoY site traffic growth and double-digit growth in applications and views. It is the one job board where "TS/SCI required" filters actually match a self-selected cleared audience.
4. GitHub-to-clearance inference: contribution patterns on defense-adjacent OSS, commits from .mil or FFRDC email domains, contributions during known deployment windows.
5. ARL-UT and FFRDC alumni networks. ARL-UT alone is the single largest feeder in the internal data. Target their alumni, not their current staff.
6. Prime-to-startup lateral pattern: engineers who left Lockheed, Northrop, or RTX in the last 24 months (still inside the reciprocity window) are the highest-conversion targets. Miss the 24-month window and you are paying for a full reinvestigation.

ClearanceJobs' 2025 Compensation Report found more than 13% of cleared professionals switched employers in the last year for better pay. Retention is soft. The comp floor is also moving: with Anduril accounting for roughly half of US H1 2026 defense VC and hiring aggressively, mid-stage Series A to B startups are effectively competing for the residual pool at Anduril-set comp benchmarks.

30,000:1

What to stop doing

A few common founder moves that waste weeks:

• Stop posting "TS/SCI required" on LinkedIn and waiting. You will get three unqualified applications and one from a contractor.
• Stop assuming reciprocity means portable. Budget the SCI read-in and poly delta into your hiring plan.
• Stop targeting FAANG leavers first. Most are outside the 24-month break-in-service window and will need a full reinvestigation.
• Stop competing on base comp with Anduril. You will lose. Compete on program scope, mission specificity, and equity.

The right motion is outbound, signal-driven, and patient. Refolk was built for exactly this shape of problem: a small, largely invisible pool where the useful signal lives in employer history and open-source behavior rather than in stated skills.

What to tell your board about the ratio

The honest board update is that headcount plans built off the $12.3B fundraising narrative are not survivable at current clearance supply, and the founders who ship in 2026 will be the ones who moved to outbound sourcing and SkillBridge pipelines in Q1. Everyone else will spend Q3 explaining why offers are stalling.

Two numbers worth including in the deck:

• Defense tech VC in H1 2026 grew roughly 2.0x YoY to $12.3B.
• The cleared population grew in the low single digits YoY, and the publicly discoverable software subset is measured in the hundreds, not the millions.

If your hiring plan assumes the pool scales with capital, rebuild it. It does not.

FAQ

How many cleared software engineers are actually hireable in the US?

The commonly cited 3.7M cleared Americans (ODNI FY2023) is not the right denominator. About 1.26M hold Top Secret and an estimated 800K to 900K are TS/SCI-eligible, but most are military, IC analysts, or non-technical contractors. Refolk's index surfaces 869 US software, embedded, and firmware engineers publicly discoverable in Defense & Space, and only 41 publicly self-identify with TS/SCI plus embedded or C++ skills. The realistic addressable engineering pool is five figures at most, not seven.

Why can't more VC funding just create more cleared engineers?

Because clearance supply is set by federal policy, not markets. DCSA runs background investigations at a policy-limited cadence and ODNI adjudicates on federal timelines. Historical YoY growth in the cleared population has ranged from roughly minus 1.2% to plus 4.2%, so even in peak years the pool adds around 150K to 170K net eligibilities, mostly military and non-engineering. Capital cannot buy around that.

What is the fastest legitimate way to add cleared engineers to a startup?

DoD SkillBridge for pre-ETS transitioning military, prime-to-startup laterals inside the 24-month break-in-service window, and ARL-UT and FFRDC alumni networks. Combine those with GitHub-to-clearance inference (defense-adjacent OSS, .mil commit domains) and outbound sourcing that can read employer-history signals rather than relying on candidates to self-tag. That is the motion that works when 99%+ of the pool is invisible to inbound.

How does Anduril's $5B round change the market for everyone else?

Anduril captured roughly half of US defense tech VC in H1 2026 and is hiring aggressively, which resets the comp floor for the entire sector. Mid-stage Series A to B startups are effectively competing for the residual cleared engineer pool at Anduril-set benchmarks. The workable response is to compete on program scope, mission specificity, and equity rather than base comp, and to source from channels (SkillBridge, ARL-UT alumni, ClearanceJobs) where Anduril's brand is less dominant than in general tech recruiting.