AMD and Spectro Cloud Announce AI Coding System With Dell Hardware

AMD and Spectro Cloud announced the AMD Instinct Coder 355, an enterprise AI coding system designed to help companies manage usage costs and keep some workloads within their own environments.

The configuration pairs Dell PowerEdge XE9785 servers with AMD EPYC 9575F processors and Instinct MI355X GPUs. Each server node is designed for teams of up to 85 developers, with as many as 50 users working at the same time. The listed hardware includes 3 terabytes of DDR5 memory, eight PCIe Gen5 TLC U.2 solid-state drives and 12 high-efficiency power supplies.

Spectro Cloud’s PaletteAI Inference Launchpad manages the platform’s AI operations. It runs the GLM-5.3 model locally and can connect to Anthropic Claude, OpenAI GPT and Google Gemini when the task and enterprise policies allow. The system also supports standard API integration, letting organizations adjust routing rules, assess other local models and connect approved external services.

AMD estimates that the system could lower total cost of ownership by as much as 75% compared with cloud frontier models.

The platform is designed to send suitable coding work to local models, reducing reliance on metered frontier model APIs. Companies can set budgets, track usage and apply quotas to AI coding as a shared service. They can keep sensitive workloads local under their own policies and reserve external models for tasks where they judge the additional capability worth the cost.

Developers can use the platform with Visual Studio Code and Claude Code. AMD’s work on coding models also includes support for MiniMax-M3.1 on Instinct MI350 Series GPUs.