The announcement of a strategic partnership between Submer and ZEDEDA marks an inflection point. The news will be game changing for any business, organisation or government that generates high volumes of data but is unable to draw real time inference from it because the datacenter is too far away.

The inability to process AI at the edge will soon be a thing of the past, as this strategic partnership sees the creation of a new joint solution architecture, combining Submer hardware and infrastructure with ZEDEDA Edge Intelligence Platform. The partnership will enable rapid edge AI infrastructure anywhere, at any scale, and operating where traditional datacenters cannot – regardless of environment and even in extreme temperatures up to +45oC/110oF.

Not only can the edge AI deployments be located anywhere they are needed, but the Submer-ZEDEDA collaboration is charting new territory: integrated edge AI appliances able to run real-time inference within days, not months. That speed from deployment to operational intelligence is a whole new ball game.

Submer & ZEDEDA, rapid field-deployable Edge AI

The first wave of AI was centralized with all data returning to hyperscale clouds for processing. But the world’s most critical operations generate enormous volumes of data far from any datacenter; it comes from factory floors, offshore platforms, telecommunications networks or remote energy sites, to name a few examples.

There is a great need to make intelligence possible locally in the local physical world, but this is a huge infrastructure challenge. You can’t simply shrink a datacenter and ship it to a mine site. You need purpose-built infrastructure that can handle the thermal loads of GPU-dense compute in environments that would normally be considered inhospitable.

Hardware alone isn’t enough and that’s why the Submer-ZEDEDA partnership exists. Submer’s liquid-cooled modular systems make it physically possible to run high-density AI inference anywhere in the world and ZEDEDA brings software-defined resilience and intelligent orchestration. This is critical because in production, high availability means expensive hardware redundancy at every site. ZEDEDA’s orchestration layer shifts that to software, detecting node failures and redistributing workloads automatically across the cluster.

ZEDEDA’s role in the partnership dramatically reduces the cost and complexity of going from one pilot site to fifty production sites. It’s the difference between artisanal deployment and industrial-scale operations.

Scale from edge to megawatt

The scalability of edge AI infrastructure can be a hurdle, but Submer’s modular, pre-validated infrastructure is designed so that an organization can deploy high-density GPU inference at a new site without redesigning the infrastructure each time. It’s repeatable, it’s tested and it scales.

The modular form factors are:

  • Pods for Compact Edge – up to 8 GPUs configured per server for on premise industrial & 5G telecom sites. Deliver real-time vision, predictive maintenance and local inference.
  • Packs – ruggedized Micro-DC. Up to 168 GPUs configured for energy, mining, ports & manufacturing environments. Deliver Industrial automation, agentic AI and multi-model inference.
  • Containers for megawatt scale – up to 800 GPUs with 10-, 20-, 40-foot configurations for sovereign AI & GPUaaS operators. Deliver large-scale inference, sovereign AI and GPU-as-a-Service.

Under the hood

Submer and ZEDEDA are leaders in their respective fields, bringing their expertise together to create a simplified architecture that is a step change in edge AI infrastructure.

Submer contributes liquid-cooled, modular infrastructure; power and thermal UPS, PDUs, chillers and power modules, as well as hardware integration with pre-validated and improved GPU utilization and server configurations.

ZEDEDA’s Edge Intelligence Platform provides auto workload redistribution orchestration to deploy and manage AI applications at scale; Software defined resilience for orchestration-level failover without hardware redundancy, lowering total cost of ownership; zero trust security for secure provisioning attestation and encryption. All ZEDEDA’s intelligence appliances are pre-integrated, validated and ready to deploy allowing the rapid distribution of edge nodes anywhere and – crucially – with centralized orchestration across them all.

The combined advantages are impressive:

  • 100kW+ per rack density thanks to Submer’s liquid cooling technology
  • Zero water consumption as Submer’s cooling eliminates water use entirely
  • 40% lower CO2 compared to traditional air-cooled facilities
  • <1.03 PUE certified power efficiency from liquid cooling

A new opportunity

AI intelligence at the edge can unlock massive data-driven opportunities across industries, organizations and governments but it is only as good as the infrastructure that powers it. Without purpose-built, deployable and sustainable compute infrastructure, none of it reaches the real world. We are proud to be working with ZEDEDA to help bring these opportunities to reality.