Nvidia Launches AI PC ‘RTX Spark’ in South Korea, Vows to Accelerate Agent AI Era

At an Nvidia launch event for its RTX Spark laptop at Seoul’s COEX on October 16, a user entered the prompt: “Clip only the parts of the video where Nvidia CEO Jensen Huang appears.” A connected monitor then generated a short edited video almost instantly. An Nvidia representative said the company had concentrated its technological capabilities on enabling artificial intelligence to directly operate PC software and carry out personalized tasks.

Nvidia said it officially launched the Surface RTX Spark Laptop Ultra, developed jointly with Microsoft, and the Surface RTX Spark DevBox mini desktop worldwide on October 16. The release came four months after CEO Jensen Huang unveiled the products at Computex 2026 in Taiwan in June, saying he would usher in the era of agent AI.

RTX Spark is designed to run AI models on a PC without relying on cloud servers. Information that cannot be exposed externally can be handled by on-device AI, while high-performance tasks can be processed by external AI. Running AI models locally also keeps personal information from being transmitted to services such as OpenAI or Gemini. It also eliminates token costs required to run models externally.

The device also highlights stronger interoperability between AI models and software through the Model Context Protocol, or MCP. MCP is a technology that allows AI language models to connect to external data and tools in a secure, standardized way.

That enables a single AI model to manage multiple programs at once, including Adobe Photoshop, Google email and calendars. A company representative said different AI models can exchange outputs with one another in real time, and results can be shared easily with other devices such as smartphones and tablets.

Nvidia also built in a broad range of features aimed at creators and developers. Users can edit 8K-class video without connecting the power supply. The machine can handle 3D rendering workloads of up to 90 gigabytes. It also includes functions to remove noise from video and emphasize a specific person’s voice.

The system can convert low-resolution video into ultra-high-definition 60-frames-per-second footage. It can also handle coding, testing and prototyping on a single device. Nvidia said it improved graphics performance as well to support immersive play for the latest high-resolution games.

Nvidia said the performance is supported by top-end components. RTX Spark includes 128 gigabytes of low-power LPDDR RAM, four times the capacity of today’s highest-spec laptops. The company said that enables 1 quadrillion mathematical operations per second, allowing large language models and AI agents to run continuously. It also uses the Grace Blackwell Spark CPU, developed in collaboration with MediaTek, to process a range of high-computing tasks quickly.

RTX Spark will be available from October 16 through the official websites and major retail channels of Asus, Dell, HP, Lenovo and Microsoft. Huang called RTX Spark a new generation of Windows PC for creators, developers and gamers, adding that it was designed with AI agents in mind.

Price may prove a hurdle. The laptop starts at $2,599, with pricing varying by region and country.

Won Jong-hwan, Hankyung.com reporter won0403@hankyung.com