Back to News
ProductAI Understanding briefing

Johnson Controls and Armada begin production at Dublin AI infrastructure plant

Johnson Controls and Armada have launched Galleon Forge Two, a 90,000‑sq‑ft Dublin facility that will produce modular sovereign AI data‑centre units, starting with a Leviathan system destined for Norway.

4 min readRead the linked source
Source-page capture accompanying Johnson Controls and Armada begin production at Dublin AI infrastructure plant
Source referenceSource recorded
Publisher
constructionowners.com
Source type
Linked source — primary-source status has not been established.
ContextUnderstand this in 60 seconds

Key terms

Compute
The processing resources required to train and run models, often measured in FLOPS or GPU hours.

What happened

Johnson Controls and Armada announced the start of production at Galleon Forge Two, a new 90,000‑square‑foot manufacturing plant in Dublin, Ireland. The facility, Armada’s first European operation and the second Galleon Forge site, will build modular computing infrastructure for sovereign AI workloads. The first product, a Leviathan modular data‑centre unit, was engineered, integrated and tested on site and is slated for delivery to the Fossefall project in Norway. The plant employs roughly 700 engineers, technicians and skilled workers and is part of a broader rollout that includes a second U.S. plant in Arizona (expected early 2027) and a future expansion in Edmonton, Canada. Across the three sites, the partners target at least 500 MW of production capacity in 2027.

Johnson Controls and Armada publicly announced on November 11, 2026 that production has begun at Galleon Forge Two, a new 90,000‑square‑foot manufacturing facility located on a Johnson Controls site in Dublin, Ireland. The plant marks Armada’s first manufacturing footprint in Europe and the second Galleon Forge location overall.

The Dublin facility is staffed by roughly 700 engineers, technicians and other skilled workers who will assemble modular computing infrastructure designed for sovereign AI workloads. The first unit produced, a Leviathan modular data‑centre, was fully engineered, integrated and tested at the site and is scheduled for delivery to the Fossefall project in Norway. This order is part of an initial five‑unit contract totaling more than 9 MW of capacity, with full delivery expected in the first quarter of 2027.

The partnership leverages Armada’s platform, which manages power, cooling and across modular data‑centres, and Johnson Controls’ thermal‑management expertise, including coolant distribution units, chillers and zero‑water cooling technologies. These combined capabilities aim to reduce energy consumption and operating noise in high‑density AI compute environments.

Beyond Dublin, the companies plan to start continuous production at Galleon Forge One in Gilbert, Arizona, in early 2027, using up to 400,000 sq ft of space and supporting more than 500 jobs. A further expansion is slated for a Johnson Controls facility in Edmonton, Canada. Across all sites, the partners target at least 500 MW of production capacity in 2027.

Source details: constructionowners.com

Why it matters

The launch expands Europe’s domestic supply chain for high‑performance AI computing hardware, reducing reliance on imported components and aligning with growing demand for sovereign AI infrastructure. By combining Armada’s modular computing platform with Johnson Controls’ advanced thermal‑management technologies—such as zero‑water cooling, chillers and coolant distribution—the facility promises more energy‑efficient, low‑noise data‑centre solutions. This matters for AI developers and enterprises seeking to locate close to renewable energy sources, as illustrated by the planned deployments in Norway and Australia. The capacity boost also signals a strategic shift toward localized, vertically integrated AI hardware production, which could influence regional policy, investment decisions and competitive dynamics in the AI‑hardware market.

Localizing AI hardware manufacturing in Europe addresses geopolitical and supply‑chain concerns by keeping critical components within sovereign borders, which is increasingly important as governments and enterprises seek to mitigate risks associated with overseas dependencies.

The integration of advanced thermal‑management solutions reduces the power and cooling overhead of AI clusters, making modular data‑centres more viable in regions with limited cooling infrastructure or strict energy efficiency mandates.

Modular, sovereign‑AI infrastructure enables customers to retain ownership and control of their assets, a key differentiator for organizations handling sensitive data or operating under strict regulatory regimes.

The ability to co‑locate AI with renewable energy sources—demonstrated by the planned deployments in Norway and Australia—supports sustainability goals and can lower operational costs for AI workloads that are traditionally power‑intensive.

Interactive Mechanism

Interactive Mechanism: How It Actually Works

Explore the underlying technology behind this development interactively.

Model Parameter Size:8B Parameters
VRAM Required5.5 GBGPU memory footprint
Target HardwareMacBook / Single GPUDeployment tier
Privacy100% Air-GappedLocal device capability
Core takeaway: Small, quantized models (3B–8B) now run directly inside smartphones and laptops with complete data privacy, while mammoth 400B+ models remain the domain of datacenter clusters.
Interactive Concept Check+10 Points
AI Models Explained Quiz

Which component of an AI application is the machine-learning model itself?

What to watch next

Future deliveries of the Leviathan units, especially the full five‑unit order for Fossefall in early 2027, will indicate the plant’s ramp‑up speed and market acceptance. Monitoring the start of continuous production at the Arizona Galleon Forge One site and the planned Edmonton expansion will reveal how quickly the partners can scale to the targeted 500 MW capacity. Additionally, adoption of the modular systems by renewable‑energy projects (e.g., WinDC in Australia) will show whether the combined thermal‑management and sovereign‑computing approach gains traction in energy‑constrained environments.

The timeline and successful delivery of the remaining four Leviathan units to Fossefall will test the plant’s production scalability and quality control processes.

Progress at the Arizona Galleon Forge One site, including when continuous production begins, will indicate whether the partners can meet the 500 MW capacity target across multiple continents.

Adoption of the modular systems by renewable‑energy projects such as WinDC in Australia will reveal market appetite for AI‑optimized, energy‑efficient data‑centre solutions.

Potential policy developments in the EU regarding sovereign AI infrastructure and incentives for domestic manufacturing could further impact the facility’s growth trajectory.

Related guides & quizzes

Found this useful?