logo

Tonomia Completes Development of TonoForge™ with Strong Engineering Support from MiTAC Computing and Close Assistance from AMD - Sovereign Containerized AI Factory Enters Commercial Release

Jun 5, 2026
Belgium

BATTICE, BELGIUM - June 5, 2026 - Tonomia SRL (Belgium) today announced the completion of development and the commercial release of TonoForge™, a fully containerized, behind-the-meter AI factory for sovereign and industrial deployment. The program was carried out by Tonomia with strong engineering support from MiTAC Computing Technology Corporation (Taiwan) on the server platform, and with close assistance from AMD (United States) on accelerator and software enablement. The AMD Instinct™ MI355X GPUs and AMD EPYC™ CPUs power the platform. The first production unit has been built, tested and commissioned at Tonomia's Battice site, Belgium, with GPU capacity entering revenue service in June 2026.


Today's release follows the strategic collaboration announced by AMD and Tonomia at GITEX Global 2025 in Dubai (15 October 2025), which set out an architecture for distributed AI factories combining AMD high-performance compute engines with Tonomia's modular sovereign infrastructure. TonoForge™ is the first production embodiment of that vision - developed by Tonomia, with strong engineering support from MiTAC on the server design and rack integration, and close assistance from AMD on GPU and CPU integration, AMD ROCm™ software enablement and performance validation. Its arrival on schedule, eight months later, demonstrates that the GITEX vision has moved from announcement to commissioned hardware earning revenue.

Tonomia and MiTAC are also exploring joint production of TonoForge™ in the United States to serve North American demand, building on the strategic partnership MOU signed by the two companies on 7 October 2025 in San Francisco during the Belgian Economic Mission to the U.S. West Coast. The co-production framework under discussion would localize final assembly and qualification in the U.S., leveraging MiTAC's North American manufacturing footprint and rack-scale integration capabilities in tandem with Tonomia Inc., with first units targeted to support the ~1.5 GW U.S. potential.

TonoForge™ packages a complete AI compute environment - accelerators, networking, power conditioning, liquid cooling, fire suppression and orchestration - inside a 20-ft ISO container. Engineered to operate close to renewable and cogeneration energy sources, with on-site heat recovery, the unit deploys ~5× faster, at ~5× lower €/MW, and in ~5× less footprint than the closest modular data-centre competitor. It targets brownfield industrial sites, behind-the-meter power agreements, and sovereign customers requiring on-soil data residency and full operational control.


THE TONOFORGE™ 20 PLATFORM - DEVELOPED BY TONOMIA

TonoForge™ 20 was developed by Tonomia and is built around the AMD Instinct™ MI355X GPU on the AMD CDNA™ 4 architecture, providing 288 GB of HBM3E memory and 8 TB/s of bandwidth per GPU. Each AMD InstinctTM  MI355X Platform provides 2.3 TB of HBM3E and up to 80 PFLOPS of peak FP4 performance, with native FP4/FP6/FP8/FP16/FP64 support - covering inference, fine-tuning and frontier-class training on a single sovereign footprint. MiTAC provides strong engineering support on the server platform – the G4826Z5 AI liquid-cooled server with the OAM UBB 2.0 baseboard, the 5th-generation AMD EPYC™ 9755/ 9575F host nodes, and system-level qualification. Tonomia engineered the container, power architecture, liquid cooling, fire suppression and controls, together with the TonoFabric™ orchestration layer on AMD ROCm™ 7.0, K3s, Ray and vLLM. AMD provided close assistance on accelerator integration, ROCm™ software enablement and performance validation.

 

TONOFORGE™ 20

LIQUID-COOLED CONTAINERIZED AI FACTORY - KEY SPECIFICATIONS

GPU

AMD Instinct™ MI355X (CDNA 4) - 288 GB HBM3E, 8 TB/s bandwidth per GPU

CPU

Dual AMD EPYC™ 9755/ 9575F processors,

Node

. 8× MI355X on OAM UBB 2.0 - 2.3 TB HBM3E, up to 80 PFLOPS FP4 per node

. 12 PCIe Gen 5 x16 slots

. 24 6TB RDIMM DDR5-6400 memory for extreme data processing

Engineering partner & system

MiTAC Computing – MiTAC G4826Z5 AI liquid-cooled server

Fabric / orchestration

400 Gbps RoCE per GPU; TonoFabric™ on K3s, Ray, vLLM, AMD ROCm™ 7.0

Container / cooling

20-ft ISO, direct liquid-to-chip + RDHx, PUE ≤ 1.15, heat recovery ready

Cost & speed

~5× faster to deploy and ~5× lower €/MW than traditional data centres

 

PARTNER STATEMENTS

. Tonomia - "At GITEX 2025 we said we would place compute at the heart of energy. Eight months later, TonoForge™ is entering revenue. Developing it in-house, with MiTAC's strong engineering support and AMD's close assistance, proves European industrial sites can host frontier-class compute without exporting the workload or the value." - Dr. Mustapha Belhabib, Founder & CEO, Tonomia.

. MiTAC - "TonoForge exemplifies the key values of MiTAC’s AMD Instinct™platform: high-density, liquid-cooled, field-deployable, and engineered with uncompromising serviceability. We are proud to provide engineering support to Tonomia as the platform enters commercial release, and we are actively exploring joint production of TonoForge™ in the United States to serve the North American market."- Rick Hwang, President of MiTAC Computing Technology Corp.

. AMD - “Countries and enterprises around the world are looking to maximize the upside of sovereign AI investments by bringing advanced compute closer to their data, energy and users,” said Keith Strier, senior vice president, Global AI Markets, AMD. “AMD is pleased to closely assist Tonomia on TonoForge and advance AMD computing at the edge, with high-performance platforms deployed where power is generated and data is created - extending the collaboration first announced at GITEX 2025.”


FROM GITEX 2025 TO BATTICE 2026 - PROGRAMME MILESTONES

TonoForge™ 20 is generally available for order from May 2026, with first customer deployments scheduled for Q3 2026. Capacity is offered through outright purchase, multi-year GPUaaS offtake, and co-located behind-the-meter deployment at Tonomia-operated sites. Tonomia

The architecture announced at GITEX 2025 has now progressed through every gate of an industrial product program. Detailed engineering by Tonomia, the MiTAC engineering-support agreement, and the AMD close-assistance framework were closed in Q4 2025. Container fabrication, power and cooling commissioning, blackstart procedures, FAT/SAT acceptance, and workload validation on AMD ROCm™ 7.0 were completed at Battice in Q1–Q2 2026, combining Tonomia's green-infrastructure engineering, MiTAC's server-platform and rack integration expertise, and AMD's accelerator and software enablement into a single sovereign product line.


AVAILABILITY & PIPELINE

TonoForge™ 20 is generally available for order from May 2026, with first customer deployments scheduled for Q3 2026. Capacity is offered through outright purchase, multi-year GPUaaS offtake, and co-located behind-the-meter deployment at Tonomia-operated sites. Tonomia has 80 MW of behind-the-meter capacity secured across Belgium and Germany, with a further 400 MW identified in Europe and a longer-horizon pipeline of approximately 1.5 GW in the United States - a platform potential exceeding 2 GW across the next siting cycle, to be served in part by the planned U.S. co-production with MiTAC.

 

ABOUT THE PARTIES & MEDIA CONTACTS

. Tonomia is a Belgian sovereign AI infrastructure company (Battice/Herve), with Tonomia Inc. in the USA. It designs and operates distributed AI factories sited close to renewable energy with on-site heat recovery, deploying ~5× faster, at ~5× lower cost, and in ~5× less footprint than the closest modular data-centre competitor.  

. MiTAC Computing designs and manufactures G4826Z5 AMD EPYC™ and AMD Instinct™ server platforms, and provides engineering support to infrastructure partners building industrial AI platforms. 

. AMD is a global leader in high-performance and adaptive computing, delivering CPU, GPU and accelerator technologies for data centre, AI and high-performance computing workloads. Its AMD EPYC™ processors and AMD Instinct™ accelerators power AI factories and the world’s leading enterprise, cloud and scientific computing platforms.

Talk to Our Experts

Let's discuss your needs, from servers to clusters.

Contact Us