Spectro Cloud and AMD Launch Instinct Coder 355 Appliance for Local-First AI Coding

SALT LAKE CITY -- Spectro Cloud and AMD announced at KubeCon + CloudNativeCon North America 2026 the AMD Instinct Coder 355, a turnkey, local-first AI coding appliance delivered by Spectro Cloud through Dell Technologies OEM Solutions.

The appliance pairs AMD Instinct MI355X GPUs, each with 288 GB of HBM3E memory and up to 8 TB/s of memory bandwidth, with Dell PowerEdge XE9785 servers and Spectro Cloud’s PaletteAI Inference Launchpad software. The software layer provides local-first inference, prepackaged and pre-optimized open-weight models, bring-your-own-model support, intelligent model routing, metering and quotas, KV cache optimization and governance, with on-premises and hosted deployment options. One pre-validated appliance supports approximately 50 concurrent developers, according to the companies’ announcement.

Policy-Based Model Routing

The appliance supports a range of open-weight models, including regional and sovereign options, so organizations can choose models that meet policy and compliance requirements. Policy-based routing sends each request to a model based on workload, capability and cost, while tasks that require a frontier model can still be routed to one. Organizations can extend the setup with the Spectro Cloud PaletteAI control plane, available with an additional subscription, to deploy and centrally manage a larger fleet of appliances.

The Coder 355 expands the AMD Instinct Coder portfolio, which launched with the Coder 325 in August 2026. “AMD Instinct Coder 325 proved that enterprises want a local-first path to AI coding, and today’s open-weight models can handle most of that work,” said Tenry Fu, co-founder and CEO of Spectro Cloud. The AMD Instinct Coder 355 can be ordered through Spectro Cloud and its reseller and distribution partners.