Overview
- Nvidia announced RTX Spark at its Computex/GTC Taipei keynote on June 1, 2026, calling it a consumer PC system-on-chip that combines a 20‑core Arm CPU with a Blackwell GPU to bring agent‑style AI to Windows machines.
- The company published headline specs including 6,144 GPU cores, up to 128 GB of unified LPDDR5X memory, and roughly 1 petaflop of AI compute to support very large models and heavy creative or gaming workloads.
- Microsoft and major PC makers such as ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI committed to devices shipping this fall and to software work including Windows security primitives and Nvidia’s OpenShell runtime.
- Important open questions remain because Nvidia has not released independent benchmarks or formal pricing, and real‑world compatibility with legacy x86 Windows apps depends on emulation and further developer work; export rules and supply decisions could also limit global availability.
- If the specs hold, users could run private, low‑latency AI agents and large offline models on thin laptops, but early devices are likely to be premium, with tradeoffs for battery use, cost, and the unresolved task of proving broad app support.