Building a Local AI Powerhouse

GEEKOM has officially launched a distributed AI cluster solution that leverages four A9 Mega mini PCs to handle enterprise-grade artificial intelligence tasks. By utilizing simple USB4 cables for connectivity, the company bypasses the need for traditional, complex data center server infrastructure.


Technical Specifications and Architecture

At the heart of each A9 Mega unit is the AMD Ryzen AI Max+ 395 processor. This chip integrates 16 Zen 5 CPU cores with Radeon 8060S graphics and unified memory within a space-efficient chassis. When configured as a four-node cluster, the system delivers a collective 512GB of RAM. The architecture is powered by Ubuntu, ROCm, and DwarfStar software, which coordinate to distribute the DeepSeek V4 Flash model across the interconnected machines.


Key hardware capabilities for the A9 Mega include:

  • Sustained Performance: 120W TDP under load.
  • Storage: Dual M.2 PCIe Gen4x4 slots with support for up to 8TB via RAID.
  • Connectivity: Wi-Fi 7, Bluetooth 5.4, dual 2.5G Ethernet, and support for four 8K displays.
  • AI Processing: 126 total TOPS per unit.

Enterprise Benefits and Privacy

This decentralized approach allows businesses to maintain strict control over sensitive data. By keeping prompts, internal documentation, source code, and credentials on local infrastructure, companies avoid the risks associated with public cloud routing. This enables the creation of private knowledge assistants capable of analyzing contracts and internal reports without exposing data to external parties.


«The arrangement allows organizations to keep prompts, documents, source code, and credentials within local infrastructure rather than routing them through a public cloud,» noted the company in its release.

Performance and Scalability

According to internal metrics shared by GEEKOM, the cluster achieved approximately 14.61 tokens per second at single concurrency, with a P95 time to first token of 0.42 seconds. The system is designed to handle long-context prompts, with technical testing reaching capacities up to 250K tokens.


The solution is highly modular. Organizations can start with a single or dual-node setup and scale up to four nodes as their computational requirements grow. While individual units can function independently, they seamlessly contribute to distributed inference when linked.


Pricing and Considerations

The GEEKOM A9 Mega is priced at $3,999 per unit (128GB RAM/2TB SSD), bringing the cost of a full four-node cluster to $16,000. It is important to note that the performance data provided is based on vendor-supplied benchmarks, and the long-term reliability of this setup for sustained, heavy-duty enterprise usage has yet to be independently verified by third-party evaluators.