NVIDIA and Microsoft used this week's Microsoft Build 2026 conference to announce an expanded AI partnership, rolling out new Windows PCs built for local AI workloads, deeper cloud integrations, and a set of enterprise tools aimed at pushing agentic AI further into the mainstream.
The new hardware: RTX Spark and DGX Station
Two new PC lines take center stage. The RTX Spark is designed to run AI inferencing directly on the device — no cloud round-trip needed. Microsoft said the machine is tuned for Windows Copilot tasks and lighter AI agents that need low latency. The DGX Station, a larger desktop system, targets developers and researchers who train or fine-tune models locally. Both machines come with NVIDIA's latest GPU architecture and are being marketed as the first wave of AI-native Windows PCs.
Pricing and release dates weren't disclosed. The companies said more details would come closer to launch.
Cloud integrations for agentic AI
On the cloud side, the pair deepened ties between NVIDIA's AI Enterprise suite and Microsoft Azure. Developers can now deploy AI agents that span across edge devices and the cloud, using the same toolchain. Microsoft highlighted a new orchestration layer that lets agents break complex tasks into smaller steps, run some on a local RTX Spark and others in Azure, then reassemble the results. The integration is meant to reduce latency for time-sensitive applications like real-time customer service or industrial monitoring.
Enterprise tools: what's in the pipeline
For businesses, the announcement included a new set of pre-built agent templates for finance, healthcare, and logistics. These templates come with NVIDIA's NeMo framework and are accessible through Microsoft's AI Foundry. The companies also introduced a monitoring dashboard that tracks agent performance, cost, and error rates across hybrid deployments. Microsoft's Copilot Studio will get native support for these agents, allowing non-technical staff to customize them with natural language prompts.
The expanded partnership reflects a broader push by both firms to own the full stack of AI computing — from the silicon in a desktop to the giant clusters in the cloud. For Microsoft, the move strengthens its position against competitors like Google and Amazon, who are also racing to embed AI agents into their platforms. For NVIDIA, it locks in a major distribution channel for its newest hardware.
No timeline was given for when the enterprise tools will exit preview. Build attendees can try the templates in sandbox environments this week.
