[AI Prism] NVIDIA launches AI assistant that works on my PC… Local AI strategy

NVIDIA has begun targeting the next-generation PC market with local AI technology that directly runs artificial intelligence (AI) agents on personal PCs.

It presents a new PC usage environment that allows users to utilize AI assistants and create content without relying on the cloud.

◇ AI assistant running on my PC… Cost and data protection strengths

Jeff Yen, technical marketing director for NVIDIA Asia Pacific (APAC), introduced the spread of local AI agents and the PC technology that supports them at the 'RTX Spark' press conference held at the JW Marriott Hotel in Seoul on the 6th.

“This year was such an important inflection point that it could be called the year of the local agent,” said Director Yen. “AI is changing the way we interact with PCs.”

He emphasized, "Unlike existing large-scale language model (LLM)-based chatbots, AI agents can understand the user's goals, deduce the necessary tasks, and even perform actual actions using various tools. AI agents are changing PCs from simple tools to team members who work together."

In particular, personal information protection, offline use, low response delay, and cost reduction were cited as advantages of local AI. Running an AI model on a personal PC eliminates the need to transmit data to an external cloud and allows work to continue even in an environment without an Internet connection.

Director Yen explained, “If you run it locally, you don’t have to worry about costs based on token usage,” and “As long as you have the hardware, you can actually use AI with just electricity costs.” Unlike the usage-based billing of cloud AI services, local AI is intended to reduce the burden of additional costs due to repeated use.

However, not all AI tasks are processed locally. NVIDIA introduced 'Nemo Switchyard', a technology that selects an appropriate model according to the characteristics of AI tasks, and presented a hybrid method that uses a PC for general tasks and a cloud-based cutting-edge AI model for advanced inference tasks.

◇ NVIDIA expands local AI ecosystem with RTX Spark

NVIDIA introduced ‘RTX Spark’ as the hardware that supports this local AI utilization. It is a superchip that combines a Blackwell graphics processing unit (GPU) and a high-efficiency Grace central processing unit (CPU), supporting up to 128GB of integrated memory and 1 petaflop of AI computing performance.

In collaboration with Microsoft (MS), we have optimized six major AI agents for RTX Spark, and also provide playbooks and blueprints for developers to easily install and run AI agents in a Windows environment.

On this day, AI agent demonstrations were held at the site, including 'Chief of Staff', which supports schedule management, and 'Microduck', which helps with robotics development. Examples of video production using ‘Video Producer Agent’ and ‘Music Performance Reimagined’ were also presented.

Attendees directly experienced RTX Spark laptops, including the MS Surface Laptop Ultra, ASUS ProArt P16, MSI Prestige N16 Flip AI+, and Lenovo Yoga Pro 9n. Game demonstrations including ‘NBA 2K27’ and ‘Aion 2’ were also held.

RTX Spark began domestic sales through a launch event held at COEX in Seoul on October 16.