BOXX launches workstations featuring NVIDIA Blackwell GPUs and high-core AMD processors
BOXX has introduced new AI and deep learning workstations, powered by NVIDIA and AMD Ryzen Threadripper PRO 9000 Series processors. These systems are designed to accelerate AI model development, training, and inference, offering security and cost control advantages over cloud solutions. The product line includes DGX Spark, an AI desktop supercomputer, and NVIDIA RTX-powered workstations, emphasizing local processing for AI workflows.
Key Takeaways
- New APEXX T4 PRO-X platform supports two flagship NVIDIA RTX PRO 6000 Blackwell or four Max-Q Blackwell GPUs via a 2050W power supply.
- Configurations feature up to 96 CPU cores using AMD Ryzen Threadripper PRO 9000 Series processors based on the Zen 5 architecture.
- DGX Spark desktop supercomputer provides up to 1 petaFLOP of AI performance for prototyping and fine-tuning large models at the desk.
- Integrated software stack includes NVIDIA AI Enterprise, NVIDIA AI Workbench, and NGC to facilitate scaling from local hardware to cloud.
Why It Matters
Local AI training hardware is becoming a critical strategic alternative to public cloud instances for organizations prioritizing data residency and cost predictability. By offering on-premise access to Blackwell-generation compute, BOXX addresses the latency and security concerns inherent in shared environments. For the streaming industry, this enables faster iteration of generative video models and real-time analytics without recurring egress fees or cloud deployment queues. This launch signals a shift toward 'departmental' AI infrastructure where specialized teams manage dedicated compute. Watch for adoption rates of desktop supercomputers vs. cloud GPU rentals as Blackwell availability stabilizes across OEM channels.
Additional Context
The launch comes as high-performance hardware prices face significant volatility. Per VideoCardz and Tom's Hardware, June 2026 reporting indicates the NVIDIA RTX PRO 6000 Blackwell has seen a retail price surge to approximately $13,250—a 55% increase over its March 2025 launch price of $8,565—driven by global memory shortages and high demand for AI accelerators. This pricing pressure highlights the 'extreme ROI' argument made by workstation manufacturers like BOXX, as fixed-cost hardware becomes a hedge against rising cloud processing rates. While BOXX targets the professional workstation market, NVIDIA is simultaneously moving into the personal AI space. Per TechNewsWorld, June 2026 reports show NVIDIA and Microsoft introducing 'RTX Spark' laptops and PCs designed to run 120-billion-parameter models locally. This dual-track approach—desktop supercomputers for developers and agent-capable PCs for creators—aims to cement NVIDIA’s 80% market lead by making local AI processing ubiquitous across different hardware tiers. AMD is also intensifying its competition in this segment. The Ryzen Threadripper PRO 9000 Series (codenamed 'Shimada Peak') launched in mid-2025 has become a staple for multi-GPU workstations due to its 128 PCIe Gen 5 lanes and support for up to 2TB of DDR5-6400 memory. Per Research and Markets, the generative AI video creation sector is projected to grow to nearly $1 billion by 2030, fueling sustained demand for these high-core-count systems that can handle the massive data preprocessing required for multimodal AI workflows.
Read full article at boxx.com