The ArenaInsights
Defense-Tech Talent Migration
Issue 01
The 9 most active 2026 software hirers among neo-primes & scale-ups · US + Europe

Where defense tech's software engineers really come from.

We traced the full career history of every 2026 software-engineering hire at the nine neo-primes and scale-ups hiring software engineers the fastest — 956 people, spanning the US, Germany, and Finland. The surprise isn't who's biggest. It's who over-indexes as a feeder once you adjust for size — and how many of these “commercial” engineers quietly did a defense tour first.

Companies
9 by 2026 software-hiring volume
Software hires
956
Full-career histories
955
Period
Jan – Jul 2026
956
Software hires, 9 fastest-hiring neo-primes & scale-ups
5.3×
Booz Allen Hamilton fed engineers here 5x faster than RTX — same 7 hires, a fifth of the engineering headcount
21%
Have a defense/military background on full career history (last job: 14%)
14%
Of the “commercial” hires are defense boomerangs
01

The scale of it

Anduril is hiring software engineers faster than anyone — 258 so far this year, roughly a quarter of all 956 across these nine. Two of the nine, ICEYE and Helsing, are European.

Anduril
258
Palantir
217
Scale AI
152
SpaceX
115
Applied Intuition
57
Shield AI
52
ICEYE
40
Saronic
34
Helsing
31
Genuine new external hires only — interns and internal moves removed. Software-engineering titles with a 2026 role start. Source: Crustdata career-history signal (cached), Aug 2026.
02

The flow — where each company pulls from

About four of five came from the commercial world (left, in accent). But raw counts mostly measure how big a company is, not how defense-leaning its engineers are — which is why the feeder rate in the next section matters more.

Commercial/other 581 · 61%Big Tech 165 · 17%Academia 70 · 7%Other defense startup 31 · 3%Legacy prime 62 · 6%Neo-prime 22 · 2%Military & gov 20 · 2%Anduril 256 · 27%Palantir 215 · 22%Scale AI 151 · 16%SpaceX 115 · 12%Applied Intuition 57 · 6%Shield AI 52 · 5%ICEYE 40 · 4%Saronic 34 · 4%Helsing 31 · 3%
951 of 956 hires shown in the flow (5 unresolved prior employers omitted). Node percentages are of all 956 hires. Band width = number of hires. Categories follow prior-employer taxonomy v2.0, which places federally funded research centers (MITRE, The Aerospace Corporation) and federal systems-engineering and consulting shops (Booz Allen, CACI, Leidos, Parsons, Collins) under “Legacy prime,” the same way the rate table in section 03 and the chart in section 05 treat them. An earlier version of this page showed a category split that could not be regenerated from the stored dataset; these numbers can, and the full cross-tab is downloadable. Source: Crustdata career-history signal (cached), Aug 2026.
Exact flow — prior category by destination company
Prior categoryAndurilPalantirScale AISpaceXApplied IntuitionShield AIICEYESaronicHelsingTotal
Commercial/other13413591683738312225581
Big Tech7123381611·231165
Academia8305182412·70
Other defense startup1623423·1·31
Legacy prime20159725·4·62
Neo-prime141·2162522
Military & gov664211···20
Read as: row category → column company. · = zero.

Why raw counts mislead

Amazon sent 88 engineers and Northrop 13 — but Amazon employs ~134,000 engineers and Northrop ~31,000. Bigger pond, more fish. To see who's really a strong feeder, you have to adjust for size.

03

The feeder rate — who actually “bites”

Adjusted for size and checked against a broader set of named companies, there's no single runaway feeder. Amazon — the biggest name in the raw count — lands mid-pack. IBM, the single largest engineering employer we could size, is nearly the most inert of anyone on this list.

SourceCategoryEngineers2026 hires (n)Per 10,000 engineers95% CI
ZooxCommercial/other1,740423.06.2–58.9
Snap Inc.Commercial/other2,660415.04.0–38.5
MITRELegacy prime2,840414.13.8–36.1
Epic GamesCommercial/other2,391312.52.5–36.7
Booz Allen HamiltonLegacy prime8,50278.23.3–17.0
The Aerospace CorporationLegacy prime2,48728.00.9–29.0
TeslaCommercial/other17,444126.93.6–12.0
Blue OriginCommercial/other5,90046.81.8–17.4
CACI InternationalLegacy prime4,51836.61.3–19.4
AmazonBig Tech134,131886.65.3–8.1
L3HarrisLegacy prime10,99376.42.6–13.1
Goldman SachsCommercial/other6,35146.31.7–16.1
Capital OneCommercial/other20,755125.83.0–10.1
MetaBig Tech40,879225.43.4–8.1
LeidosLegacy prime9,03544.41.2–11.3
Northrop GrummanLegacy prime31,072134.22.2–7.2
AtlassianCommercial/other7,82733.80.8–11.2
Parsons CorporationLegacy prime5,56023.60.4–13.0
MicrosoftBig Tech95,250323.42.3–4.7
Collins AerospaceLegacy prime18,12342.20.6–5.7
GoogleBig Tech86,847182.11.2–3.3
BAE SystemsLegacy prime15,53931.90.4–5.6
NvidiaBig Tech27,27451.80.6–4.3
RTX / RaytheonLegacy prime44,98871.60.6–3.2
JPMorgan ChaseCommercial/other37,52141.10.3–2.7
BoeingLegacy prime43,72720.50.1–1.7
IBMCommercial/other87,87030.30.1–1.0
Lockheed MartinLegacy prime39,66500.00.0–0.9
Rate = 2026 hires ÷ LinkedIn “Engineering” headcount, ×10,000. Covers the 28 companies with a large enough named engineering department to compare this way — 276 of 956 hires (~29%). The rest come from hundreds of individual employers too small to benchmark this way, mostly inside “Commercial/other.” The 95% CI is a Poisson interval on the observed hire count, which is what sets the precision here: Amazon’s 6.6 rests on 88 observed hires and is tight, while Zoox’s 23.0 rests on 4 and could be anywhere from 6.2 to 58.9. Amazon is counted at the parent entity (Amazon plus AWS, 88 hires); an earlier version of this page showed 87 by using the exact profile label only. Denominator: Crustdata (cached), Aug 2026.

How to read it

This is a same-methodology rate — hires per 10,000 people in a company's LinkedIn-listed engineering department — not a claim against the whole labor market. Read the n and the interval before the rank. Most rows above rest on fewer than ten observed hires, so their 95% intervals are wide and overlapping: the top of the table is not statistically separated from the middle of it. What the data does separate cleanly is the bottom — Boeing, IBM and Lockheed sit below every other row even at the top of their intervals. “The companies we've sized” is the honest scope, not every company an engineer could work for. Judged this way, Amazon (6.6 per 10,000, on 88 observed hires) sits mid-table, not at the top. The more interesting pattern isn't one company — it's a category. See below.

Two feeders too small to rank fairly

Aechelon Technology, a 30-person simulation and training contractor, sent 9 engineers into this cohort — roughly 30% of its entire engineering staff. Vannevar Labs (58 engineers) sent 4. Both would top this table by a wide margin, but at that scale a handful of hires isn't a rate, it's an anecdote — too small to compare on the same axis as Amazon or Lockheed, though notable on its own terms.

04

The boomerang — last job isn't the whole story

Judge these hires by their last job and 14% came from defense. Look at their whole careers and it's about 1 in 5 — defense tech is quietly rehiring its own alumni after Big Tech trains them up.

Defense background — last job only (n=135 of 956)
14%
Defense background — full career (n=199 of 955)
21%
Defense/military or neo-prime — full career (n=216 of 955)
23%

The move that repeats

14% of the hires whose last job was commercial or Big Tech had earlier defense or military experience — 103 of 745. They did a tour at a prime or in uniform, left for commercial tech, and have now moved into defense tech. For a recruiter, that is a population the obvious searches miss: filtering for people currently at a defense employer would surface 14% of this cohort and skip the rest. We have no conversion or offer-acceptance data on these hires and no read on why any of them moved, so this is a statement about where they came from, not about how well they respond or what motivates them.

Full-career figures verified on all 955 enriched profiles. “Defense background” includes short and early-career stints (median ~2 years) — not career defense lifers. The third bar is a union, not a sum: 216 people have defense/military or neo-prime history, and 13 of them have both. An earlier version of this page added the two shares to 24%, which double-counted that overlap. The boomerang denominator is the 745 hires whose immediate prior employer was commercial or Big Tech; an earlier version used 14% against a wider denominator that also included academia and unresolved employers. Rule-based classifier via Crustdata (cached) full work histories.
05

Legacy defense isn't one thing

“Legacy primes don't feed defense tech” is too simple. Split them by what they actually do, and a real pattern shows up: the consulting and systems-engineering shops feed real talent; most of the hardware-platform primes barely move. L3Harris is the one hardware company that breaks the rule.

MITRE n=4 / 2,840 eng
14.1
Booz Allen Hamilton n=7 / 8,502 eng
8.2
The Aerospace Corporation n=2 / 2,487 eng
8.0
CACI International n=3 / 4,518 eng
6.6
L3Harris n=7 / 10,993 eng
6.4
Leidos n=4 / 9,035 eng
4.4
Northrop Grumman n=13 / 31,072 eng
4.2
Parsons Corporation n=2 / 5,560 eng
3.6
Collins Aerospace n=4 / 18,123 eng
2.2
BAE Systems n=3 / 15,539 eng
1.9
RTX / Raytheon n=7 / 44,988 eng
1.6
Boeing n=2 / 43,727 eng
0.5
Lockheed Martin n=0 / 39,665 eng
0.0

Same hires, different pond

Booz Allen Hamilton and RTX/Raytheon each sent exactly 7 engineers into these nine companies this year. Booz Allen's engineering headcount is a fifth of RTX's (8,502 vs. 44,988) — so its engineers left for defense tech at more than 5x the rate. MITRE and The Aerospace Corporation, both federally funded research centers rather than typical contractors, show the same pattern even more sharply. Lockheed sent zero of 956; Boeing sent two.

Legacy-prime companies only, sorted by 2026 hires per 10,000 LinkedIn-listed engineers. Each label carries the observed hire count and the engineering headcount it is divided by. Same denominator and source as the table in Section 03. The ordering above the fifth row rests on single-digit hire counts and their 95% intervals overlap heavily, so read the top of this chart as a group, not a ranking.
06

Where the talent physically sits

California leads by a wide margin, but a real second cluster shows up around Washington, DC — the cleared-work corridor Big Tech doesn't have.

Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Hawaii Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota Mississippi Missouri Montana Nebraska Nevada New Hampshire New Jersey New Mexico New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island South Carolina South Dakota Tennessee Texas Utah Vermont Virginia West Virginia Wisconsin Wyoming District of Columbia District of Columbia California · 267 · 27.9%New York · 68 · 7.1%Washington, DC · 65 · 6.8%Washington · 60 · 6.3%Texas · 52 · 5.4%Virginia · 32 · 3.3%Massachusetts · 24 · 2.5%Colorado · 20 · 2.1%

Geography is a sourcing map

California carries the largest single share, but New York (68) and Washington, DC (65) are close behind — and DC sits right next to Virginia (32), the two together forming a cleared-work corridor you won't find in Big Tech. Beyond the US: Finland and Germany, driven by ICEYE and Helsing, are hubs a US-only map can't show.

Percentages are of all 956 hires (states below the top 8 and non-US hires aren't plotted, so the 8 shares don't sum to 100%). Bubble area encodes hire count. Location is the engineer's current profile location, not necessarily the role's worksite. Source: Crustdata (cached), Aug 2026.

Publication receipt

Everything needed to check this page against its own data.

Cohort IDARENA-I01-SW9-2026YTD
Cohort956 software hires at 9 companies; 955 full career histories
Hire window2026-01..2026-07
Data as of2026-08 (hire records); 2026-08-10 (engineering-headcount denominators)
Taxonomy versionprior-employer v2.0 / artifact v1.1 / discipline v1.1
Dataset hashhires fcf0b46252be · career histories 3ecb89b0682f
Build hash1565186c844f
Reproducibility115 of 160 published figures across Issues 01 and 02 regenerate from the stored data; the rest are itemised in the ledger, with what changed and why
Revised2026-08-28
Aggregate tables and claim ledger: every chart and table on this page, plus the reproducibility ledger and a de-identified row-level extract, are downloadable as spreadsheets at /data/arena-insights-data-pack-v1.zip. Person names and profile URLs are not included and are held privately for audit.
What changed in this revision: the section 02 category split and cross-tab are regenerated under a single stated taxonomy (v2.0) so that sections 02, 03 and 05 group companies the same way; the feeder-rate table and the legacy-prime chart now show the observed hire count, the denominator and a 95% interval on every row; Amazon is counted at the parent entity (88, not 87); the last-job-only defense share is 14% under the stated taxonomy rather than 13%; the third lineage bar is a union (23%) rather than a sum that double-counted 13 people as 24%; the boomerang share is stated against the denominator the sentence actually names; and two claims about recruiting conversion and candidate motivation have been removed because we hold no data on either.

Method, scope & caveats

Scope: nine companies selected for 2026 software-hiring volume (Anduril, Palantir, SpaceX, Scale AI, Applied Intuition, Shield AI, ICEYE, Saronic, Helsing) — not company size, funding, or valuation. 956 genuine new external software hires, 2026 YTD. This is a claim about software hiring at these nine firms — not about all of defense tech, and not about hardware.

Feeder rate: 2026 hires per 10,000 LinkedIn-listed engineers, benchmarked against the 28 prior employers with a large enough named engineering department to size this way (276 of 956 hires). Every row carries its observed hire count and a 95% Poisson interval, because most rows rest on fewer than ten hires and their intervals overlap. It's a same-methodology comparison across companies we've actually looked up, not an index against the full labor market — treat it as a relative ranking among peers shown, not a universal 1.0 baseline.

Career lineage: immediate-prior employer vs. full work history are reported separately. “Defense background” = ever at a legacy prime, defense contractor, or in the military. Full-career results computed on 955 enriched profiles.

Data: Crustdata career-history signal, Aug 2026 (cached). LinkedIn lags real start dates by weeks, so recent months are undercounted. Classification is rule-based and imperfect at the margins.

“The credit belongs to the man who is actually in the arena.” — Theodore Roosevelt
Who's behind this

The Arena recruits inside defense tech.

The engineers these companies win are mostly still at Amazon, Microsoft, and Meta — and about a fifth of them already did a defense tour somewhere earlier in their careers. Finding them is exactly the search we run. Building your team, or open to the jump? Let's talk.

insights.arena.vet
info@arena.vet