At GTC Taipei 2026, NVIDIA officially introduced NVIDIA RTX Spark, an all-new personal computer platform built on the ARM architecture. It is the result of a strategic partnership between NVIDIA, Microsoft, and MediaTek for the AI era. With the advanced GB10 chip, NVIDIA RTX Spark promises graphics and AI performance that mainstream users have never had before. It is widely seen as NVIDIA's boldest move yet to push deep into the traditional PC market, which Intel and AMD have long dominated. The first laptops and desktops running this platform are expected to launch in late 2026.
NVIDIA RTX Spark Changes How We Use Personal Computers
NVIDIA RTX Spark is the first PC platform focused squarely on running AI directly on the device. Unlike earlier Qualcomm Snapdragon chips, which stopped at battery savings, NVIDIA RTX Spark is designed to run LLMs and heavy graphics workloads. Putting an ARM CPU and a Blackwell GPU on the same chip removes the data bottleneck. The machine responds noticeably faster than a traditional configuration, with no more waiting when launching complex applications or processing heavy multimedia content. Where Windows on Arm previously struggled with app compatibility, NVIDIA RTX Spark takes advantage of a modern ARM architecture and a software ecosystem optimized for AI. Microsoft has committed to integrating development tools deeply into Windows, making it easier for manufacturers to build compatible devices.
The most distinctive feature of NVIDIA RTX Spark is its Unified Memory architecture. The CPU and GPU access the same shared memory pool, which allows LLMs of up to 200 billion parameters to run locally without the cloud. This makes creative workflows seamless, unlike conventional computers where the CPU and GPU must shuttle data back and forth between separate memory pools.
NVIDIA RTX Spark Performance and Specs with the GB10 Chip
At the heart of NVIDIA RTX Spark is the GB10 chip, built on TSMC's 3nm process, which pairs 20 ARM v9.2 CPU cores with an integrated Blackwell GPU. It delivers 31 TFLOPs of FP32 and 1000 TOPS of AI performance while drawing only 140W. That means you can own a machine with workstation-class power in a compact, cool-running device. The GB10 chip also improves data security, since information does not need to be sent elsewhere for processing. Beyond raw compute, NVIDIA RTX Spark supports fast LPDDR5x memory of up to 128 GB. Even with dozens of browser tabs open, heavy apps running, and AI tools in use, the computer keeps running smoothly. The system also offers impressive multi-display support, driving up to four 4K displays at 120Hz or an 8K display. This is a big advantage for multimedia creators, or anyone who needs a roomy workspace while running AI.
| Spec | Technical details |
| Processor | GB10 (ARM CPU + Blackwell GPU) |
| Manufactured by | TSMC 3nm process |
| System memory | Up to 128 GB LPDDR5x |
| AI performance | 1000 TOPS (NVFP4) |
| Display resolution | 4K @120Hz or 8K @120Hz |
These figures have been validated in real-world testing. NVIDIA RTX Spark accelerates AI 10 times faster than the previous solution. AAA games like Alan Wake II and Fortnite run smoothly thanks to advanced DLSS 4.5. As this technology becomes widely available, mainstream users will get realistic graphics and smart AI features that were once found only on expensive servers. The arrival of NVIDIA RTX Spark marks a new chapter, one where the line between the personal computer and the AI supercomputer is gradually blurring.
