Specifications
Model
Aetina MegaEdge AIP-FR68S
Positioning
AI agent edge workstation (award-winning flagship)
Accelerator Options
Qualcomm Cloud AI 100 Ultra cards and/or NVIDIA RTX PRO Workstation GPUs
AI Performance per Card
870 TOPS (Qualcomm Cloud AI 100 Ultra)
Max Accelerators
Up to 3x Qualcomm Cloud AI 100 Ultra cards via PCIe Gen5 switch
Interconnect
PCIe Gen5 high-speed switch
Power Supply
Built-in 2000 W PSU
Model Scale
On-premises LLMs up to 300 billion parameters
Generational Gain
100% performance leap over the previous Dragonwing on-prem solution
Inference Capabilities
Real-time fine-tuning, multimodal analysis, automated instruction generation
Disaggregated Serving
Qualcomm Cloud AI 100 SDK multi-card disaggregated LLM serving
Software Suite
Qualcomm AI Inference Suite; Qualcomm Dragonwing AI on-prem solution
Agent Workloads
Voice agents, image generation and retrieval-augmented generation (RAG)
Serviceability
Patented screwless mechanism for rapid field service
Fleet Management
Out-of-band (OOB) management for distributed deployments
Data Residency
All workloads processed locally for privacy and compliance
Ecosystem
Demonstrated with Mobagel C-Suites AI Builder and the Genie CXO solution
Availability
Pre-order; deliveries began early Q2 2026
Overview
The Aetina AIP-FR68S is an edge AI workstation aimed at enterprises that want generative and agentic AI running inside their own building. Its design centre is a PCIe Gen5 switch that can drive up to three Qualcomm Cloud AI 100 Ultra cards at 870 TOPS each, giving roughly 2,600 TOPS of accelerator capacity inside one enclosure — while also accepting NVIDIA RTX PRO Workstation GPUs, which makes it one of very few platforms that can mix those two accelerator families.
The 2000 W internal power supply is there to keep that density stable in a space-constrained edge location rather than a data-centre rack. On the software side the Qualcomm Cloud AI 100 SDK supports multi-card and disaggregated LLM serving, and the Dragonwing AI on-prem stack addresses models up to 300 billion parameters — a 100% performance improvement over the previous generation, according to Aetina. Real-time fine-tuning, multimodal analysis and automated instruction generation all run on the box.
Deployment economics are addressed explicitly: a patented screwless mechanism speeds field service, and out-of-band management lets distributed appliances be reached even when the host operating system is unavailable. Aetina demonstrated the platform with Mobagel's C-Suites AI Builder running the Genie CXO agent for executive reporting, with every workload processed locally so data never leaves the premises.
Key Benefits
Up to three 870 TOPS Cloud AI 100 Ultra cards behind a PCIe Gen5 switch delivers rack-class inference in an edge-footprint appliance. Support for both Qualcomm and NVIDIA RTX PRO accelerators avoids locking a deployment to one ecosystem. OOB management plus a screwless service design lowers the cost of operating many appliances in the field.
Applications
On-premises and sovereign LLM inference, agentic AI for enterprise workflows, RAG and document intelligence inside regulated environments, private multimodal analysis, and edge fine-tuning where data residency rules prohibit public cloud use.
Request a Quote — AETINA AIP-FR68S — EDGE AI AGENT WORKSTATION WITH QUALCOMM CLOUD AI 100 ULTRA
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →