Riihimäki Test Centre Bridges Nordic Startups with NATO for Dual-Use Tech

Millog’s new Riihimäki test centre opens a direct lane between Nordic startups and NATO-linked defence customers, combining a low-threshold, standards-focused testing environment with HAMK’s GPU-backed AI infrastructure to accelerate dual-use technology development for both civil and military markets.

Background​

Finland has systematically built a dense regional defence innovation cluster around Riihimäki in recent years, led by the Defence Innovation Network Finland (DEFINE), local universities, and established defence contractors. The newest node in this ecosystem is Millog Oy’s test centre inside Riihimäki’s DEFINE Innovation Hub, launched in August. The centre is explicitly positioned to serve startups and small and medium-sized enterprises—particularly companies participating in NATO’s DIANA accelerator and the DEFINE accelerator—by offering network, software, and AI testing capabilities tuned to safety‑critical and defence‑grade requirements.
This move sits on top of a string of policy and industry actions: Finland’s national push to host accelerators and testbeds for dual‑use tech, integration of local academic compute resources, and closer collaboration between established maintenance specialist Millog and Häme University of Applied Sciences (HAMK). The combined platform—branded in local coverage as the DEFINE Testing Environment—will also expand to include additive manufacturing capability, signalling ambitions beyond software into hardware prototyping and rapid iteration.

---

Overview: What Millog’s Riihimäki test centre is and who it serves​

The Riihimäki facility is built to be a practical, low-barrier testing environment for technologies with both civilian and defence utility. Key characteristics include:

• A data‑centre style testing environment accessible onsite and remotely.
• Support for virtualised server/workstation environments running Linux and Windows.
• Support for Android‑based mobile solutions, containerised systems, and integration with external internet‑connected devices and sensors.
• Access to a GPU server facility for AI model development and testing, supplied via a partnership with Häme University of Applied Sciences (HAMK).
• A planned expansion to include additive manufacturing (3D printing) capabilities for hardware prototyping.
• A focus on compliance with defence standards and testing procedures relevant to safety‑critical systems.

The centre is explicitly targeted at early‑stage ventures and SMEs—those that typically struggle to access secure, standards‑compliant testbeds—and to companies inside NATO innovation pipelines (such as DIANA), as well as customers across NATO member states.

---

Facilities and technical capabilities​

Virtualised and containerised testing environment​

The centre provides virtualised environments to mirror operational IT stacks used by authorities and defence integrators. This includes:

• Virtual Linux and Windows servers and workstations for application testing and integration.
• Container orchestration and runtime testing, enabling microservices and cloud‑native applications to be validated under simulated network and security policies.
• Android emulation and device integration, allowing mobile agents, field apps, and sensor gateways to be validated in an isolated, resilient environment.

These capabilities allow developers to validate software behaviour against realistic network topologies and constrained hardware profiles before moving to field trials.

AI and GPU compute​

A critical addition is the availability of a GPU server to support AI model training and inference testing. That capability opens the door for:

• Development and evaluation of language model and agent‑based defence applications.
• Simulation and adversarial testing of AI systems in a controlled, offline setting.
• On‑premises evaluation suitable for sensitive datasets and model weights that cannot be exposed to uncontrolled cloud services.

The HAMK‑backed GPU resource is intended to lower the friction for startups that lack high-end compute yet need to test modern generative and agentic AI workloads.

Additive manufacturing and physical prototyping​

Planned expansion of the DEFINE Testing Environment to include additive manufacturing will enable an integrated software‑to‑hardware development path:

• Rapid prototyping of enclosures, mounts, and sensor housings.
• Iterative hardware validation alongside software updates.
• Shorter development loops for dual‑use devices that combine embedded systems, sensors, and AI.

This combination of virtual, AI, and physical prototyping tools is designed to shorten R&D cycles and reduce the gap between proof‑of‑concept and deployable systems.

---

Partnerships and ecosystem: Millog, HAMK, DEFINE and RAIN.global​

The facility is the product of an ecosystem approach:

• Millog brings operational knowledge from being a long‑standing strategic partner to the Finnish Defence Forces, with deep lifecycle maintenance and systems engineering expertise.
• Häme University of Applied Sciences (HAMK) supplies GPU compute and an applied research capability, creating a safe AI development environment intended for data‑sensitive work.
• DEFINE (Defence Innovation Network Finland) coordinates the Innovation Hub in Riihimäki, providing accelerator programmes and local access to defence schools, research agencies, and industrial partners.
• Integration partners such as RAIN.global (an AI/edge platform specialist) have been involved in stitching the environment together, particularly for AI deployment, data streaming, and edge scenarios.

Taken together, these partners aim to provide a bridge from startup prototypes to defence‑grade demonstrations and potential procurement conversations with end users.

---

Why this matters for NATO, DIANA and dual-use innovation​

• Accelerating cross‑domain solutions: NATO’s DIANA programme seeks to tap commercial innovation for alliance needs. Low‑barrier testbeds that can accept early‑stage technologies and help them mature into demonstrable, standards‑aligned systems are a key missing link in many innovation funnels.
• Local credibility and operational insight: Millog’s operational relationship with the Finnish Defence Forces provides domain credibility. That can help startups navigate military requirements (cyber, safety, interoperability) earlier in their development cycles.
• AI and edge readiness: Access to GPU compute and an offline AI development lab reduces a common bottleneck for dual‑use startups aiming to deploy language models, agent frameworks, or computer vision in constrained, secured environments.
• NATO interoperability and scaling: By aligning the centre’s offerings with DIANA and other NATO‑oriented accelerators, innovators can gain access to a broader network of test facilities and potential pathway funding and integration channels across member states.

This combination supports not only Finnish and Nordic defence preparedness but also the broader NATO objective to source dual‑use innovation from commercial markets and adapt it for military operations.

---

How startups and SMEs can realistically benefit​

• Lower cost of entry: Early‑stage teams often cannot afford secure test facilities. The centre promises a “low‑threshold” route to verify technical assumptions under realistic security constraints.
• Faster procurement readiness: By testing against defence compliance and safety standards early, companies can reduce time lost to re‑engineering later in procurement cycles.
• Access to domain mentors and end users: Being co‑located with DEFINE and within Riihimäki’s defence cluster makes it easier to involve military subject‑matter experts in testing campaigns.
• Remote test and demonstration: Both on‑site and remote testing pathways increase geographic accessibility for international DIANA participants and NATO allies.

---

Technical and regulatory considerations: what the centre enables—and what it does not​

The testing platform is designed to simulate safety‑critical and defence‑relevant conditions, but a few caveats are important:

• Standards alignment is necessary but not automatic. Passing lab tests does not equate to certification or procurement clearance. Companies should treat the centre as a validation and integration step—not a substitute for formal approvals.
• Data handling and classification rules apply. For AI and for any sensitive datasets, organisations must follow strict data governance and export control rules. The on‑premises GPU capability addresses some concerns, but operators must still design for separation, encryption, and auditability.
• Interoperability testing requires common baselines. Demonstrating an application to an end user still often requires interface agreements, message schemas, and security profiles to be agreed in advance.

---

Risks, challenges and unanswered questions​

While the test centre brings measurable advantages, it also raises several practical and policy‑level risks and questions that stakeholders should evaluate.

Security and attack surface​

• Any facility that connects to external devices, sensors, and remote participants widens the potential attack surface. Rigorous network segmentation, hardened endpoints, and continuous monitoring are essential to avoid the centre becoming a pivot for supply‑chain or IP theft attacks.
• AI model security is an emerging concern: model weights, training data, and inference endpoints must be protected from extraction and poisoning attacks.

Export controls and dual‑use governance​

• Technologies with both civilian and military applications often fall under national and international export control regimes. Startups must be aware that testing in a defence‑adjacent environment can trigger additional obligations.
• Cross‑border participation (e.g., companies from different NATO members) may require export licenses or intergovernmental agreements for certain categories of hardware or software.

Intellectual property and commercial risk​

• Early‑stage companies often trade IP for access to testing and mentoring. Clear, written agreements are needed to preserve ownership and avoid forced technology transfers or unfavorable licensing terms.
• The potential for procurement bias exists: incumbents or larger contractors might gain early insight into startup solutions and adapt them without equitable commercial arrangements.

Classification and secrecy​

• Defence customers may require demonstrations or tests to be conducted under classified conditions. The facility appears tailored for unclassified and controlled testing; moving to classified trials will need additional infrastructure and accreditation steps.

Scalability and demand management​

• If the centre becomes popular through DIANA and other NATO channels, demand could exceed capacity for high‑end GPU cycles or additive manufacturing slots, creating scheduling and prioritisation challenges for small teams.

Verification gaps and transparency​

• Some published claims (e.g., specific throughput numbers for GPU capacity, precise additive manufacturing hardware models, or detailed cybersecurity accreditation) may not yet be publicly verifiable. Companies should obtain concrete SLAs and facility specifications before planning critical tests.

---

Strategic analysis: implications for Finland, NATO and industry​

For Finland​

The test centre strengthens Finland’s position as a regional hub for defence innovation. Hosting a facility that couples operational maintenance expertise (Millog) with local academic compute (HAMK) demonstrates a pragmatic approach: combine lifecycle knowledge with modern AI tooling. This can attract both investment and talent to Riihimäki and the Kanta‑Häme region.

For NATO​

The centre provides a practical node that fits NATO DIANA’s model—identify commercial deep tech and accelerate it into alliance capabilities. By lowering the barrier for international startups to demonstrate to end users, the facility can shorten the funnel between selection by DIANA and operational evaluation by alliance militaries.

For Industry and Startups​

Existing defence OEMs gain a channel to discover and integrate promising dual‑use technologies earlier and with less upfront risk. For startups, the risk‑reward calculus improves: access to credible testing and potential procurement pathways offsets part of the commercialization risk—provided legal, IP and security concerns are managed.

---

Practical recommendations​

For startups and SMEs using the centre​

• Prepare a compliance package before booking:
• Document data classification, export control exposure, and required access levels.
• Sign clear IP and confidentiality agreements:
• Ensure ownership of core IP and define boundaries on derivatives or joint development.
• Design for data minimisation and on‑premises execution:
• Use encrypted datasets, ephemeral keys, and container isolation for model tests.
• Validate security posture:
• Bring penetration testing results and hardened builds where required.

For Millog, HAMK and DEFINE operators​

• Publish clear service specifications:
• GPU specs, network isolation architecture, additive manufacturing hardware lists, and SLAs.
• Institute strict onboarding and vetting:
• KYC, provenance checks, and export control screening for foreign participants.
• Provide tiered access:
• Unclassified sandbox for general use; elevated, accredited tracks for controlled or sensitive tests.
• Create a transparent prioritisation mechanism:
• Ensure fair access between local startups, DIANA participants, and established OEMs.

For policymakers and procurement authorities​

• Clarify export control pathways:
• Provide guidance for startups on how testing in national innovation hubs affects exports and licensing.
• Fund bridging grants:
• Small grants to help startups reach demonstration readiness for NATO or national procurement.
• Encourage standards alignment:
• Support development of common testing frameworks to reduce repeated rework during procurement.

---

Final assessment​

Millog’s Riihimäki test centre is a pragmatic, well‑timed addition to Europe’s growing network of dual‑use innovation infrastructure. By coupling Millog’s field‑level maintenance and systems expertise with HAMK’s GPU resources and the DEFINE ecosystem, the facility reduces a major barrier for startups: the ability to test real systems in a realistic, security‑aware environment. This matters for NATO’s DIANA objectives and for procurement pathways that historically have struggled to absorb rapid commercial innovation.
However, the benefits are conditioned on rigorous governance: cybersecurity hardening, export control compliance, transparent IP agreements, and operational SLAs. The expansion into additive manufacturing and deeper AI testing amplifies both opportunity and risk—hardware prototypes and model weights require strong custody controls.
For technology entrepreneurs aiming at defence and civil markets, the centre represents an important first‑mile accelerator for maturing capabilities into demonstrable, standards‑compatible solutions. For governments and alliance buyers, it offers a scalable approach to source and evaluate commercial innovation more quickly—if the governance and resourcing model is kept robust and transparent.

---

Conclusion
The Riihimäki test centre exemplifies a modern, ecosystem‑based model for defence innovation: co‑locate operational expertise, academic compute, and accelerator programmes to speed the conversion of commercial deep tech into deployable capabilities. Its success will depend as much on the technical stack as on the policies and processes that govern access, security, and the fair treatment of innovators. If those levers are managed well, the centre could become a replicable blueprint for NATO‑aligned innovation hubs across Europe—accelerating AI‑ready, standards‑compliant dual‑use technologies from lab demos to operational reality.

---

Source: Defence Industry Europe Finland: Millog opens Riihimäki test centre to boost NATO dual-use technology innovation