AMD at Microsoft Build 2026: Building Local & Cloud AI with Open-Source Tools

AMD joined developers, engineers, and AI builders at Microsoft Build 2026 in San Francisco on June 2-3, hosting four hands-on workshops that brought the AMD AI ecosystem directly to attendees. Spanning cloud-scale AMD Instinct™ MI300X GPUs and local AMD Ryzen™ AI Max+ 395 (codename: "Strix Halo") systems, the workshops demonstrated a consistent message: AMD is delivering an open, scalable AI platform that meets developers wherever they build.

From Cloud to Local: AI Across Every Tier

The workshops at Microsoft Build reflected a core principle of the AMD AI strategy, giving developers the same open-source tooling and consistent experience whether they are running workloads on cloud-scale GPUs or on local hardware at their desks.

On the cloud side, attendees accessed AMD Instinct MI300X GPU instances with 192 GB of HBM3 memory, providing the memory headroom required for large model inference, video generation, and multi-step agent workloads. On the local side, Ryzen AI Max+ "Strix Halo" laptops with 128 GB of unified LPDDR5X memory, 16 Zen 5 CPU cores, and a 40-CU RDNA 3.5 integrated GPU gave attendees the ability to run meaningful AI workloads entirely on device, with no cloud dependency. For developers accustomed to cloud-only workflows, seeing comparable inference happen locally on a laptop was a consistent point of engagement throughout the event.

Workshops: Hands-On Building Across the AI Stack

The workshop sessions covered agents, generative media, local AI deployment, and AI-assisted development, each highlighting a different layer of the AMD AI ecosystem.

Building Your Personal OpenClaw Agent with Open-Source Models walked attendees through the OpenClaw framework, teaching them to build AI agents powered by open-source language models on AMD GPUs. The session featured a challenge where participants had to prompt their agent to recognize its own multimodal capabilities. The task proved deceptively difficult, with participants iterating through creative prompting strategies before succeeding. The format generated strong engagement and turned the session into a collaborative problem-solving exercise rather than a passive tutorial.

Video Generation with ComfyUI on AMD Instinct MI300X GPU gave attendees hands-on access to ComfyUI's node-based generative AI interface, powered by the MI300X instances. Participants built video generation workflows from scratch, experiencing real-time media creation on AMD hardware and exploring what is now possible with open-source generative AI tooling at scale.

Local AI Inference with Lemonade on AMD Hardware showcased Lemonade, the AMD open-source local AI server, running on Strix Halo laptops. Attendees experienced how Lemonade auto-detects AMD GPUs and NPUs and serves LLM chat, image generation, and more through an OpenAI-compatible API, all running locally and privately. For developers accustomed to cloud-hosted endpoints, seeing equivalent inference running on a laptop demonstrated the viability of local-first AI development on AMD silicon. The challenge featured a “Space Battle Agent” Game with OpenCode, which brought attendees from a blank terminal to a playable multiplayer game using OpenCode--the open-source AI coding agent. The session highlighted how AI-assisted development workflows can accelerate prototyping and creative experimentation, with all compute running on AMD hardware.

All four workshops saw strong attendance and high levels of hands-on participation from attendees across experience levels.

Demo Highlights: Open-Source AI in Action

Across the workshops, the demos showcased how AMD-powered systems are being used in real-world AI development:

• ComfyUI running real-time video generation workflows on AMD Instinct MI300X GPUs
• Lemonade serving LLM chat and image generation via an OpenAI-compatible API on Ryzen AI Max+ laptops
• OpenClaw powering agentic AI systems built with open-source language models
• OpenCode enabling AI-assisted game development entirely on AMD hardware

What AMD Exhibited at Microsoft Build 2026

The AI ecosystem is converging on agentic architectures, local-first deployment, and open tooling that gives developers the full control they require from experimentation to production. The AMD platform is built around that shift: delivering open-source software, hardware-consistent APIs, and a compute platform that scales from a developer’s desk to cloud-scale GPU infrastructure.

The level of engagement across all four workshops made one thing clear: developers want to build, not watch, and AMD’s AI hardware and software are at their most compelling when people can get their hands on them.

Ready to continue your AI journey? 

Explore AMD AI Playbooks for hands-on guidance and join the AMD AI Developer Program to access free MI300X cloud credits, AI Academy courses and other member benefits.

We look forward to seeing you at the next Advancing AI event.