Varjo’s Sharp End
Helsinki-based Varjo has quietly become a linchpin of high‑fidelity virtual and mixed reality (XR) training across NATO, with deployments spanning more than a hundred simulation programs and dedicated “secure manufacturing” and compliance stacks designed to satisfy defence acquisition requirements while keeping the product largely commercial off‑the‑shelf (COTS). XR’s shift from novelty to mainstream training infrastructure coincides with display, sensor, and GPU advances (notably on NVIDIA’s stack), while the science of “human‑eye resolution” sets a measurable bar for headset fidelity. This article dissects Varjo’s technology architecture, its defence‑market entry playbook, and its significance for the Nordic dual‑use innovation ecosystem.
“Extended reality” (XR) is an umbrella that includes virtual reality (VR), augmented reality (AR), and mixed reality (MR), blending or replacing the physical world with rendered content to varying degrees. NVIDIA frames XR within a broader spatial‑computing workflow powered by RTX GPUs, streaming (CloudXR), and digital‑twin toolchains (Omniverse), which is materially relevant because defence simulators depend on predictable, low‑latency rendering at scale. These definitions are not academic: procurement, safety cases, and certification often hinge on whether a system is VR‑only, camera passthrough MR, or hybrid, each with different security, integration, and human‑factors implications.
Optics, displays, and the “eye‑limiting resolution”
Varjo’s early headsets pursued “human‑eye resolution” using its two‑display‑per‑eye “Bionic Display” approach, a high‑density focus panel blended over a wider‑FOV peripheral panel supported by eye tracking and dynamic foveated rendering to prioritize pixels where the user looks. The developer documentation explains that Varjo devices submit four views instead of two to feed the focus/peripheral architecture and that foveation is standard in Varjo’s Unity/Unreal SDKs, reducing GPU load without perceptual loss.
The scientific ceiling matters: a Nature Communications study measured the resolution limit of foveal achromatic vision at ~94 pixels per degree (ppd), with lower limits for certain chromatic patterns, offering a quantitative benchmark for when “more pixels” stop yielding visible benefits. This contextualizes why Varjo pushes ppd and optical clarity, and why software correction (e.g., digital lens compensation) can further mitigate lens aberrations that otherwise squander panel resolution.
Varjo’s fourth‑generation XR‑4 Series moved to a high‑resolution single display per eye (dual 4K×4K, ~51 ppd, ~120×105° FOV) while retaining eye tracking for foveation, a sign that native panel density has risen enough to simplify the optical stack without abandoning performance budgets. NVIDIA’s technical outreach positions XR‑4 as part of an ecosystem with Omniverse and RTX‑class GPUs to enable photorealistic MR and ray‑traced scenes in training.
Sensors, pass‑through MR, and depth
Training‑grade MR depends on low‑latency, high‑fidelity video pass‑through to blend real cockpits and controls with synthetic outside‑the‑window scenes, plus depth sensing (e.g., LiDAR) for occlusion and world‑locking. Varjo’s XR‑3/XR‑4 families combine stereo RGB pass‑through and depth awareness to enable accurate occlusions and interaction with physical instruments, a requirement for realistic aviator training workflows.
Tracking, ergonomics, and human factors
Inside‑out tracking, automatic IPD, and 200 Hz eye tracking (on earlier XR‑3/VR‑3) reduce simulator sickness and sustain multi‑hour use; integrated hand tracking (Ultraleap) supports natural control evaluation and cockpit interactions without external markers. The performance/comfort trade-offs are relevant in defence where training blocks may be long and repeated; headset thermal management and mass also factor into readiness metrics.
Software integration and simulator stacks
Varjo’s headsets plug into common defence simulation stacks (e.g., Lockheed Martin Prepar3D, VBS Blue IG, FlightSafety Vital) and engines (Unreal/Unity), which lowers integration cost and accelerates program adoption compared with bespoke systems. NVIDIA’s XR SDKs (VRWorks, CloudXR) and Omniverse microservices are increasingly used in digital twin and synthetic environment pipelines, which defence primes and integrators leverage when building full‑mission trainers.
Why militaries care
Modern warfare is multi‑domain, sensor‑saturated, and transparency‑prone, making traditional live training both expensive and constrained by safety, OPSEC, and environmental limits. The U.S. Department of the Air Force reported substantial efficiency improvements in XR‑enabled training pipelines (e.g., faster solo, earlier certification), highlighting XR’s role in moving high‑consequence training closer to mission time. At I/ITSEC, the world’s largest modelling and training event, XR’s footprint has grown from “next big thing” showcases in 2019 to a pervasive baseline by 2024–2025, an adoption curve mirrored by Varjo’s partner presence.
From pull signal to “default choice”
Varjo’s VR‑1/VR‑2/XR‑1 (2019) launches involved automotive and architecture partners, with an early defence signal via BISim and Saab; a road‑testing collaboration with Volvo validated pass‑through MR for real‑world tasks. That early “pro‑grade optics + enterprise integration” focus differentiated Varjo from mass‑market headsets tuned for entertainment price points.
By 2025, Varjo positions itself as the default provider for high‑end military simulation across NATO and partners, citing selection in 120+ simulation programs and operations in 40+ countries, with HQs in Helsinki, Arlington (VA), and Munich. The U.S. Army’s RVCT (Air) program integrated a special XR‑3 Focal Edition variant to deliver mixed‑reality training for Apache, Chinook, and Black Hawk crews, a visible endorsement for MR‑first training. Parallel reporting in Finland notes Varjo’s growing U.S. defence ties and FAA/EASA approvals enabling more formal training credit for VR‑based flight training.
“The DoD is a process, not a customer”
Varjo’s internal narrative (reflected in the Aalto forum deck) is pragmatic: selling to DoD means navigating a process (FARs, FOCI), and winning often entails selling through system integrators and primes, while staying involved across the program lifecycle. U.S. acquisition basics reinforce that contracting officers (KOs) hold binding authority; programs move from market research and requirements to solicitation, evaluation, award, and admin, hence why non‑traditional vendors lean on primes/VARS for speed.
To preserve COTS performance while meeting defence constraints, Varjo invested in a secure manufacturing line in Finland (2024) and obtained Certificate‑to‑Field (CtF) from the U.S. Air Force (2025) for software, paired with TAA‑compliant hardware, ISO 27001, and no‑radio/offline licensing variants. This approach enables defence simulators using Varjo headsets to secure ATOs (authorities to operate) while the headset itself remains COTS, preserving velocity of commercial innovation cycles.
An AFWERX Super Goggle Challenge award in 2024 signalled alignment with next‑gen pilot training goals (digital design of high‑performance goggles for advanced synthetic environments) supporting Varjo’s long‑term positioning inside USAF innovation channels. Media across defence training circles echoed the selection, underscoring Varjo’s credibility in U.S. procurement experiments.
Varjo typically ships through defence primes and system integrators, with VARs bundling compute and support; Varjo’s own solution engineers remain embedded across requirements, integration, accreditation, and sustainment. This channel strategy mitigates barriers like FOCI sensitivity and accelerates program‑level acceptance compared to a small OEM bidding prime themselves.
Technical deep‑dive
Varjo’s SDKs implement dynamic foveated rendering driven by high‑rate eye tracking; on XR‑3/VR‑3, this spans the blended focus/peripheral displays, while XR‑4’s higher native resolution reduces the need for dual panels but retains foveation to keep frame budgets stable. Foveation is pivotal in cockpit trainers where complex out‑the‑window scenes, sensor overlays, and symbology must update at low persistence and high refresh with minimal latency.
Mixed reality depends on low‑jitter camera synchronization, depth sensing, and calibration against physical cockpit geometry; LiDAR‑aided occlusion prevents “virtual” controls from drawing over real switches and gauges, sustaining proprioceptive trust. NVIDIA positions RTX pipelines (and, increasingly, AI models) as the backbone for real‑time denoising, reconstruction, and streaming, which matters when simulators are distributed or cloud‑rendered.
The ~94 ppd foveal achromatic limit from Nature Communications suggests the industry still has headroom, but not infinite, beyond 50–70 ppd; more pixels help until the visual system saturates, especially in high‑contrast tasks like reading small cockpit labels. Chromatic resolution limits being lower implies headset pipelines must manage chromatic aberration and colour fringes to avoid perceived blur, hence software corrections like “digital lens” approaches.
Defensive training stacks frequently employ Unreal/Unity for content and Prepar3D/VBS/FlightSafety Vital for IGs; Varjo’s compatibility breadth reduces rip‑and‑replace costs during XR retrofits of legacy simulators. NVIDIA Omniverse’s role in building and streaming digital twins into headsets, sometimes via edge/cloud, extends MR beyond single‑seat trainers to multi‑node, collectively simulated environments.
Business strategy
Varjo’s defence thesis is that the advantage of COTS isn’t just price; it’s performance: consumer‑paced display and GPU cycles outstrip bespoke mil‑spec product refreshes, yielding superior fidelity if the supplier can satisfy security and governance demands. The “catch,” as Varjo acknowledges, is requirement creep, over time, defence constraints tug COTS toward bespoke, so the company pre‑invested in secure manufacturing and CtF‑grade software to preserve its release cadence.
Funding models also matter. Defence‑native R&D may slow product performance and velocity; Nordic “new‑defence” firms increasingly use private capital to build “ready‑to‑procure” products and then engage with flexible contracting vehicles, precisely the pattern Varjo describes.
Channels and case signals
The RVCT (Air) selection demonstrated MR’s portability and realism for rotorcraft crews, with a no‑radio variant and TAA‑compliant Finnish manufacturing satisfying security postures. The AFWERX Super Goggle contract indicates USAF interest in co‑designing next‑generation goggles for advanced synthetic training, aligning Varjo with future procurement pathways. European and U.S. aviation authorities (EASA/FAA) approvals for VR‑based training workflows further legitimize headset‑centric simulators for accredited training credit.
The Nordic angle
Varjo’s trajectory exemplifies a Nordic dual‑use pattern: deep academic roots (Aalto), rapid iteration in commercial verticals (automotive, AEC), and then disciplined entry into defence via compliance, not bespoke hardware. As Finland’s NATO membership normalized security collaborations, doors opened for U.S. defence programs and joint certifications, amplifying Nordic presence in high‑end training tech. The result: a Nordic supplier delivering secure COTS to allied militaries at scale, an industrial policy outcome that marries economic competitiveness with alliance readiness.
Risks and open questions
- Supply chain sovereignty vs. cost: Tightening export controls and component provenance checks could complicate COTS sourcing; Varjo’s secure manufacturing in Finland mitigates some risk but not upstream silicon dependencies.
- Accreditation drag: Program‑by‑program ATOs and evolving cyber baselines can slow deployments; keeping headsets COTS while securing surrounding systems is a constant balancing act.
- Optics vs. perception ceilings: As ppd climbs toward the eye’s limits, further gains may shift to temporal stability, colour accuracy, HDR, and comfort, rather than sheer pixel density.
- Streaming and AI dependencies: If cloud/edge rendering becomes the norm (CloudXR/Omniverse), bandwidth, sovereignty, and ML model assurance will become acquisition concerns, not just IT issues.
What’s next
- XR‑4 “Secure” variants and software updates inheriting CtF learnings into broader accreditation templates for primes and government labs.
- USAF/USN program of record adoptions leveraging AFWERX outputs into production contracts; watch contracting vehicles that favour “production‑ready prototypes.”.
- NVIDIA’s XR AI/Omniverse integration into defence digital twins, especially multi‑ship or multi‑domain training with ray‑traced fidelity at scale.
- Standards and ppd benchmarks informed by the Nature study, potentially guiding future headset procurement specs for cockpit legibility and symbology.
Varjo: the short version
- Market footprint: Selected in 120+ NATO/partner simulation programs; presence in 40+ countries; HQ in Helsinki/Arlington/Munich.
- Products: XR‑4 series (MR with high‑res pass‑through, LiDAR), eye‑tracking and foveation; wide ecosystem compatibility (Unreal/Unity, Prepar3D, VBS, Vital).
- Security: Secure manufacturing (Finland), TAA compliance, ISO 27001, no‑radio variants, offline licenses; CtF for software.
- U.S. pipeline: RVCT (Air) variant adoption; AFWERX Super Goggle award; growing FAA/EASA training approvals ecosystem.
Sources
- Aalto Dual‑Use Technology Forum deck (Varjo CTO Pyry Haulos)
- Varjo corporate release: AFWERX Super Goggle Challenge (Nov 7, 2024). [halldale.com]
- Air Force Technology coverage of AFWERX award (Nov 2024). [varjo.com]
- Halldale defense training coverage of AFWERX award (Nov 2024). [acq.osd.mil]
- Military Embedded Systems: U.S. Army RVCT (Air) Varjo variant (Aug 29, 2023). [yle.fi]
- Yle News (Aug 21, 2024): U.S. defense contracts and FAA training approvals context. [developer.varjo.com]
- Varjo Developer Docs: Human‑eye resolution, foveated rendering, and multi‑view pipeline. [almalence.com]
- Nature Communications (Oct 27, 2025): Ashraf et al. “Resolution limit of the eye — how many pixels can we see?” (94 ppd foveal achromatic). [oxbridgesi…gapore.org]
- MIXED Reality News (Nov 27, 2023): XR‑4 Series announcement/specs overview. [2026-03 Du…logy Forum | PDF]
- ST Engineering product book reprint: XR‑3/VR‑3 specifications and ecosystem integrations. [imotions.com]
- NVIDIA: XR Solutions overview; Developer XR page; NVIDIA Blog “What is XR?”; GTC trend reporting (CloudXR/Omniverse, AI). [sportsvideo.org], [blogs.nvidia.com], [analyticsinsight.net], [developer.nvidia.com]
- OUSD (A&S) contracting basics slide deck (FARs/KO authority etc.). [militaryembedded.com]
- Digital Engineering (2017): Early Varjo strategy for pro‑grade headsets. [stengg.com]
- Almalence Digital Lens: computational correction for optical aberrations in near‑eye displays (Varjo Aero demo context). [digitaleng…ing247.com]
Photo: Varjo – Varjo has been selected for an award of a contract through the AFWERX Challenge Commercial Services Offering (CSO) process. The contract is focused on the United States Air Force’s (USAF) AFWERX Super Goggle Challenge, to address a pressing challenge in the Department of the Air Force.
