Swarm evidence graph that turns mixed-drone trial data into ministry-ready autonomy scorecards and approval packets.

1. Funding is flowing specifically to autonomy software rather than airframes, so a software evidence layer can attach to live budgets.
2. Expansion across European and allied markets means the same autonomy behavior must be explained to more than one buyer, increasing the value of reusable evidence.
3. Swarming pushes evaluation complexity up sharply, making manual after-action reporting too slow once 10-50 vehicles are involved.
4. Because the cluster points to a software-control bottleneck, buyers can adopt a neutral evidence graph without waiting for a new hardware winner.

Catalyst. Six Robotics' funding is explicitly aimed at autonomy software deployments across European and allied markets plus drone swarming, which means programs will need a repeatable evidence layer immediately rather than more bespoke post-mission reporting.

Swarm Evidence Graph would ingest mission logs, operator interventions, geofence events, and vehicle-health data from existing ground stations after each exercise. It would reconstruct a common mission timeline across vendors, score each sortie against agreed objectives, and auto-build approval packets with comparable metrics, failure annotations, and supporting clips for program reviews. The product runs on-prem or inside a secure enclave, so integrators keep their existing autonomy stacks while finally getting a neutral evidence layer above them. Over time, the company builds the benchmark dataset for takeover rates, comm-loss recovery, swarm cohesion, and repeatability across allied exercises, turning one debrief workflow into a cross-program data moat.

What's different. Unlike autonomy vendors, this company does not try to own mission planning or vehicle control. It becomes the neutral evidence layer across OEMs, integrators, and ministries, with a normalized event schema, failure taxonomy, and benchmark library that gets stronger every time a new exercise runs. That creates defensibility exactly where allied programs are forced to mix vendors but still need comparable proof of autonomy performance.

Jobs to be done

flowchart LR
  Buyer[Program director] --> Pain[Mixed-vendor swarm exercises create manual debrief chaos]
  Pain --> Product[Swarm Evidence Graph]
  Product --> Outcome[Faster approvals plus reusable autonomy evidence]

Executive takeaways

• The wedge is real but narrow: mixed-vendor swarm programs are expanding faster than the evidence and approval workflows around them, creating a gap above autonomy stacks and below ministry sign-off [1][3][4][6][7][10][14][17][18].
• The best initial customer is a European integrator or OEM test leader with a live 10-50-drone exercise, not a generic defense software buyer; urgency peaks when another milestone or procurement increment depends on reusable evidence [3][4][5][17][18][28].
• Budget pull exists—Six raised €12M, Helsing is pursuing a $1.2B round, and EU defense spend hit €343B in 2024—but the beachhead TAM is modest enough that expansion into broader autonomy assurance is part of the venture case [1][2][11][19].
• Adjacent incumbents are formidable but incomplete: Applied, Anduril, Shield AI, Helsing/Systematic, and Siemens/IBM all own pieces of simulation, autonomy, C2, or traceability, yet none visibly own a neutral cross-vendor evidence graph for ministry review [7][8][20][21][22][23][24][25][26][31][32][33].
• The biggest adoption risks are secure data access, country-specific airworthiness and approval rules, and buyer temptation to stretch existing logs or digital-engineering tools instead of funding a new layer [13][15][16][17][18][34][35][36][37][38][39].

Market definition

Supplier-side autonomy evidence software for allied unmanned exercises: an on-prem layer that ingests mixed-vendor logs and C2 data after each swarm event, reconstructs one timeline, scores autonomy behavior, and emits ministry-ready approval packets. It sits above the autonomy stack and below procurement sign-off, with Europe as the first geography and 10-50-drone exercises as the first wedge [1][3][4][6][7][8][10][12][14][15][28][35][36][37][38][39].

Customer and buyer

Primary daily users are flight-test leads, autonomy-evaluation managers, and program engineers at 150-800 person integrators or OEM teams preparing a ministry exercise or contract increment. The economic buyer is usually the VP Programs, program director, or GM for unmanned systems because schedule slip, range-time burn, and approval delays hit revenue and customer trust directly [3][4][5][12][17][18][28].

Buying triggers

• A new swarm exercise or pilot milestone requires comparable evidence across vendors before another budget increment or deployment approval is released. [3][4][5][17][18]
• Coalition or marketplace procurement forces a buyer to compare plug-and-play systems rather than evaluate one OEM in isolation. [6][10][14][28][29][30]
• Autonomy software changes or contested-environment questions create pressure to prove behavior without weeks of manual debrief work. [1][17][21][22][23][24][25][26]

Willingness to pay

Budget gravity is real because Europe is explicitly pushing drones, procurement reform, and autonomy programs, while Six and Helsing show investor appetite for software-defined defense. But willingness to buy a standalone evidence layer is still unproven and should be sold against avoided milestone delay, reused approval labor, and fewer bespoke post-trial reports rather than generic analytics. [1][2][11][12][19][28][29][30]

Category dynamics

Growth signal 19% y/y EU defence expenditure growth in 2024

Validation signals

• Six Robotics raised €12M specifically for autonomy software and allied deployments, showing that software-control layers now attract dedicated defense capital.
• The Norwegian Army selected and then received Valkyrie for pilot and operational testing with FFI, proving live demand for iterative swarm-autonomy workflows.
• TIE 26 gathered 40 companies with 60 systems and 40 software applications, validating a large visible interoperability ecosystem that already needs shared evidence and comparison.
• Intelic BASE and NATO vendor-pool efforts show buyers are actively trying to compare interoperable unmanned systems faster.
• UK and EU strategy documents explicitly call for digital standards, procurement reform, and faster adoption of uncrewed systems.

Regulatory & technical constraints

• Military UAS categorization and approval remain formal and updated under national rules, so reusable packets still need country-specific annexes.
• STANAG 4586 interoperability depends on vehicle-specific modules and message translation, so cross-vendor evidence normalization is not free even when systems claim openness.
• Swarm programs increasingly need to operate across open C2 and coalition architectures, which raises schema and interface scope beyond one OEM.
• Ground-station and onboard log formats remain fragmented across ULog, MAVLink, DataFlash, rosbag2, and MCAP.

Competition is intense but mostly adjacent. Applied comes closest on test and digital-engineering workflow; Anduril and Shield AI own mission autonomy; Helsing/Systematic push sovereign swarm C2; Siemens, IBM, and JAGGAER own traceability or procurement; and many programs still default to vendor logs, spreadsheets, and PowerPoint. The startup wins only if it becomes the neutral evidence graph and approval-packet layer that these systems do not own [7][8][20][21][22][23][24][25][26][31][32][33][34][35][36][37][38][39].

Why incumbents do not win by default

• Autonomy stack vendors. Anduril, Shield AI, and Helsing/Systematic already ship autonomy or swarm-control software, but integrators and ministries may resist letting a direct platform competitor own the neutral evidence corpus, failure taxonomy, and approval logic across third-party fleets.
• Digital engineering suites. Applied, Siemens, and IBM cover simulation, test, and requirements traceability, but they are not visibly optimized for post-exercise, ministry-ready evidence packets spanning multiple OEM logs and operator interventions.
• Procurement workflow platforms. JAGGAER and marketplace-style procurement motions speed sourcing and compliance, yet they do not solve the harder supplier-side task of proving autonomy behavior after live exercises.
• Existing test toolchains. Vendor-specific telemetry, ULog, MAVLink, DataFlash, and rosbag-style workflows already exist, but they create fragmented evidence rather than reusable cross-program benchmarks.

Swarm Evidence Graph sells an on-prem autonomy-evidence layer to European unmanned-systems integrators and OEM program teams running 10-50-drone exercises for NATO-aligned ministries. The acute pain is post-exercise proof, not flight control: mixed-vendor logs, operator interventions, and clips are stitched into one-off reports, which delays milestone approvals and contract increments. The MVP reconstructs a common mission timeline from existing telemetry and C2 data, scores autonomy behavior against agreed objectives, and emits a ministry-ready approval packet without replacing the current autonomy stack. This is a disciplined beachhead because it has a live buying trigger, a clear economic buyer in the program director or VP Programs, and a measurable ROI in days saved between exercise completion and approval. The company should not build autonomy, mission planning, or generic PLM first; the credible wedge is becoming the neutral evidence record across vendors that buyers do not want any single autonomy vendor to control. Research supports real demand and timing, but the initial market is modest—roughly $13.3M estimated SAM and $2.8M estimated year-3 SOM—so the venture case depends on expanding later into broader autonomy assurance, release gating, and cross-domain benchmarking. The first proof point is an accepted packet on one mixed-vendor exercise that cuts debrief time from weeks to days and converts into an annual program license before the next program milestone. The biggest disconfirming risk is that target accounts either will not share usable logs in secure environments or will only fund the workflow when bundled inside an incumbent autonomy or integration contract.

Problem

• Mixed-vendor swarm exercises leave flight-test and autonomy-evaluation teams with telemetry, C2 events, operator actions, and clips spread across incompatible tools, making it hard to prove what the autonomy software actually did.
• Manual debrief packets delay approvals, consume contractor reporting hours, and make evidence hard to reuse across contract increments, ministries, or vendor mixes.

Solution

• Deploy an on-prem evidence graph that ingests post-exercise logs and operator artifacts, reconstructs one cross-vendor timeline, and produces comparable autonomy scorecards and failure annotations.
• Map telemetry and events into ministry-ready approval packets so the same evidence can support the next milestone review, vendor comparison, and benchmark library.

Why we win

• Neutrality is the product: integrators and ministries can adopt the layer without handing the evidence record to a competing autonomy-stack vendor.
• The connector library, failure taxonomy, approval-template mappings, and cross-exercise benchmark dataset compound with each deployment and are harder to replicate than a dashboard.

Milestones

flowchart LR
  Wedge[Mixed-vendor swarm debrief pain] --> MVP[On-prem evidence graph plus first connectors]
  MVP --> Proof[Accepted ministry packet and faster review cycle]
  Proof --> Expansion[Annual program licenses plus broader autonomy-assurance workflows]

Founding team

Experiment roadmap

Risk assessment

What must be true

• At least 5 of the first 15 target accounts will grant usable telemetry, C2, and operator-artifact access under an on-prem deployment model.
• The first live deployment can cut post-exercise packet turnaround from weeks to 5 business days or less.
• At least 2 of the first 4 pilots will convert into recurring annual program licenses or equivalent embedded program budgets.
• One approval-packet template can be reused across at least 2 ministries or program offices with limited customization.
• No autonomy or digital-engineering incumbent can match neutral cross-vendor evidence well enough to block the first 3 target deals.

Open diligence questions

• Which 3 vendor or C2 combinations cover most of the first-beachhead data volume?
• Who actually owns the budget for evidence tooling inside the first target programs: integrator, OEM, or ministry office?
• How much of the first accepted packet can be generated from sanitized or unclassified data?
• How much services work is required to adapt one approval template across Norway, the UK, the Netherlands, Germany, and France?
• What evidence do buyers need that a neutral layer is safer than using bundled reports from autonomy or C2 vendors?

Model sanity

• Revenue engine. Base revenue comes from moving from 2 paying programs at Y1 exit to 8 by Q3-Q4Y3 while mature program value settles near the researched $350K-$400K range.
• Must go right. The first two pilots must convert into annual licenses and the packet template must be reusable across ministries so deployments stop looking like consulting.
• Model breaks if. If secure data access and ministry-specific tailoring stretch sales cycles while gross margin stays below about 68%, the downside case pushes cash close to the floor before scale proof.
• Next-round proof. The seed-ready proof point is 5-6 paying programs, 3 annual conversions, and one partner-sourced deployment by Q4Y2 with reusable approval packets across 2 ministries.

Scenarios

Sensitivity

flowchart LR
  TargetAccounts[Programs with live swarm milestones] --> DesignPartners[Qualified design partners]
  DesignPartners --> PaidPilots[Paid pilot packets]
  PaidPilots --> AnnualLicenses[Annual program licenses]
  AnnualLicenses --> Expansion[More exercises, connectors, and add-ons]
  Expansion --> Revenue[Revenue]
  Revenue --> GrossProfit[Gross profit]
  GrossProfit --> Cash[Cash runway]

Flags: The base case reaches 8 of roughly 38 SAM programs by year 3, so success requires winning a meaningful share of a concentrated buyer pool. · Gross margin reaches the low-70s only if connector reuse and packet-template reuse keep deployments from remaining services-heavy. · Budget-owner ambiguity may still force sales through primes or C2 partners, which could compress both independence and margin. · The researched SAM is only about $13.3M, so venture-scale upside still depends on expanding beyond post-exercise packets into broader autonomy-assurance workflows after year 3. · Cash is modeled as EBITDA; milestone billing, security hardware, or compliance capex could shift real cash timing.

• Data access and classification. Secure ranges and proprietary vendor log formats may block the product from receiving enough exercise data to generate trusted scorecards. Mitigation: Start on unclassified or coalition exercises with a narrow connector set, ship on-prem, and prioritize vendors that already export usable telemetry.
• Prime or OEM bundling. Large autonomy vendors or primes may offer their own reporting tools and try to keep buyers inside single-vendor evidence silos. Mitigation: Win where ministries or integrators must compare multiple autonomy stacks on one program and need a neutral cross-vendor record.
• Budget-cycle drag. Defense buyers may agree the debrief problem is real but delay standalone software purchases until the next milestone or procurement window. Mitigation: Sell first through milestone-driven integrator budgets and tie ROI to shorter approval cycles, reusable evidence, and fewer contractor reporting hours.