June 22, 2026, (Inside AI) — Microsoft and NVIDIA have revealed RTX Spark, a new system-on-chip designed to power what they call the most capable thin-and-light Windows PCs ever built. The chip was announced at NVIDIA GTC and packs 1 petaflop of AI performance, up to 6,144 Blackwell RTX cores, 20 Arm-based CPU cores, and up to 128GB of unified memory into a single silicon package.
The architecture merges GPU, NPU, and CPU functions to handle AI-native computing locally. RTX Spark targets developers, creators, and power users who need to run advanced AI workloads, graphically demanding creative applications, and agent-based tasks without cloud reliance.
Microsoft collaborated directly with NVIDIA to optimize Windows for the chip’s heterogeneous design. This includes workload profile scheduling that distributes tasks across all 20 cores more efficiently. The Microsoft Power and Thermal Framework also maximizes performance-per-watt under sustained loads.
Unified Memory Unlocks On-Device AI
The unified memory architecture marks a leap for local AI. Systems with 128GB allow developers to load large language models entirely on-device. They can run complex rendering projects and build agentic workflows that previously required cloud compute.
Microsoft confirmed that NVIDIA will bring NVIDIA OpenShell to Windows. Hermes Agent and OpenClaw will integrate new Windows security and containment primitives for safe, local agent execution. This move could reshape how developers approach sensitive AI tasks.
Read: Apple's Core AI Lets iPhones Run 70B Language Models Locally
At launch, the app ecosystem includes Blender, DaVinci Resolve, Photoshop, Premiere, CapCut, GitHub Copilot, Claude Code, Cursor, ComfyUI, and CUDA-accelerated PyTorch. Riot Games confirmed League of Legends and VALORANT will run on RTX Spark. PUBG: Battlegrounds, Alan Wake 2, and Naraka: Bladepoint also confirmed compatibility.
From Laptops to Enterprise Supercomputers
RTX Spark-powered Copilot+ PCs will launch this fall from Microsoft Surface, ASUS, Dell, HP, Lenovo, and MSI. Microsoft also outlined a longer-term roadmap scaling Windows from RTX Spark laptops up to the NVIDIA DGX Station for Windows. That system will use the GB300 Grace Blackwell Ultra Desktop Superchip, which the company frames as putting a trillion-parameter AI supercomputer on every enterprise desk.
Industry observers note the push toward local AI compute challenges the cloud-centric model. By embedding powerful AI hardware directly into client devices, Microsoft and NVIDIA could reduce latency, improve privacy, and lower operational costs for developers. However, questions remain about battery life, thermal constraints, and real-world performance of such a dense chip in thin-and-light form factors.
Read: OpenAI and Broadcom Unveil Jalapeño: Custom AI Chip for LLM Inference at Gigawatt Scale in the U.S.
The RTX Spark announcement comes as competitors like Apple and Qualcomm also advance their own Arm-based AI silicon. Microsoft’s deep integration with NVIDIA’s software stack—including CUDA and AI frameworks—may give it an edge in developer adoption. Yet, the true test will be whether the fall hardware delivers on the promised petaflop without throttling.
