Tech • IA • Crypto
Announcing NVIDIA RTX Spark | GTC Taipei 2026 Keynote by CEO Jensen Huang
9/10TL;DR
Nvidia and Microsoft are introducing a new generation of AI-driven personal computers built around autonomous agents, marking what they describe as the most significant reinvention of the PC in four decades.
KEY POINTS
A 40-year shift in personal computing
The modern PC traces its roots to early Windows and Apple systems that standardized hardware abstraction, software layers, and developer ecosystems. This architecture helped transform computers from enterprise tools into mass-market consumer devices. The new initiative aims to replicate that scale of disruption by redefining how users interact with machines through AI.
Rise of the “agentic” PC
The next-generation PC is designed around AI agents that operate continuously, assist users autonomously, and interact through voice, vision, and natural language. These agents can analyze files, perform research, execute tasks, and adapt over time. They are expected to run both locally and via cloud-connected models, creating a hybrid computing experience.
Large language models as a core OS layer
Large language models (LLMs) are positioned as a fundamental extension of the operating system, comparable to past innovations like DirectX. These models enable multimodal input and output, including text, images, video, and sound, effectively embedding intelligence directly into the computing platform.
Introduction of RTX Spark platform
Nvidia unveiled RTX Spark, a new computing platform built on a Blackwell RTX GPU with 6,144 CUDA cores and up to one petaflop of AI performance. It integrates a custom 20-core Grace CPU developed with MediaTek, uses 128 GB unified memory, and is manufactured on a TSMC 3nm process with 70 billion transistors.
Full software compatibility and performance gains
The system is designed to run the entire Nvidia CUDA stack alongside all existing Windows applications, ensuring backward compatibility. Applications across domains such as biology, physics, graphics, and AI can run natively. Major software providers like Adobe are optimizing tools such as Photoshop and Premiere, with reported performance improvements of up to 2×.
Agents transforming workflows
Demonstrations show agents handling complex workflows, such as architectural design, by integrating tools like Rhino and Blender. The agent can generate layouts, correct errors, and produce photorealistic renders using generative AI models, significantly reducing manual effort and time.
Expansion across device categories
The initiative includes a full lineup of Windows-compatible systems: laptops, desktops, and high-performance workstations. Each system supports CUDA, AI Tensor Cores, and agent-based computing, forming a unified ecosystem across device types.
High-end workstation capabilities
A flagship workstation configuration offers 768 GB of memory, 20 petaflops of compute performance, and 8 TB/s memory bandwidth, enabling developers to run models with up to trillions of parameters locally before deploying them to the cloud.
Always-on personal AI systems
Desktop systems are envisioned as always-on hubs running agents continuously, connected to home devices, security systems, and personal data. These systems can manage daily tasks such as scheduling, travel booking, and home automation, becoming increasingly personalized over time.
Vision of AI in every home
The long-term vision suggests that households may adopt dedicated AI computers similar to home theaters or appliances. These systems would function as persistent assistants, evolving into interactive companions that handle a wide range of tasks autonomously.
CONCLUSION
The collaboration between Nvidia and Microsoft signals a shift toward AI-native computing, where autonomous agents and embedded intelligence redefine the role of the personal computer in everyday life.