Spectro Cloud and AMD launch Instinct Coder 355 on Dell PowerEdge

One box, up to 50 developers at once, no round-trip to the cloud. That was the pitch Spectro Cloud and AMD opened with on November 10 at KubeCon North America 2026, announcing the AMD Instinct Coder 355, a turnkey on-prem AI coding appliance built on a Dell PowerEdge XE9785 packed with eight AMD Instinct MI355X GPUs.

The hardware is loud about itself. Each MI355X ships 288 GB of HBM3E memory with up to 8 TB/s of memory bandwidth; a node stacks eight of them alongside a pair of AMD EPYC 9575F 64-core CPUs and 3 TB of DDR5 system RAM, Spectro Cloud's launch materials say. The box runs GLM-5.3 locally through Spectro Cloud's PaletteAI Inference Launchpad, and when a task needs more it routes out to Anthropic Claude, OpenAI GPT, or Google Gemini.

"AMD Instinct Coder 325 proved enterprises want local-first AI coding paths, and open-weight models handle most work," Spectro Cloud co-founder and CEO Tenry Fu said in the announcement. "With the 355, customers choose appliances for their specific needs."

Spectro Cloud claims the appliance reduces AI coding token costs by up to 75% versus cloud frontier models and delivers up to 98% of frontier-model benchmark performance. Those figures come from the vendor; no third-party workload mix has been published. The 355 lands in the middle of a dense week for silicon news, one of 329 chips stories we've tracked in the last 90 days, and it slots alongside the growing shelf of private coding-assistant appliances.

Orders go through Spectro Cloud and reseller partners. Live demos were scheduled at Booth #125.