Artificial Intelligence Revolution from NVIDIA and Microsoft: RTX Spark Comes to Windows Computers
Quick Look
- NVIDIA and Microsoft CEOs announced the MXC infrastructure and RTX Spark architecture for artificial intelligence agents in the Windows ecosystem.
- With products like the new Surface Laptop Ultra and DGX Station, native AI processing power increases to 1 petaflop and 20 petaFLOPS.
AI-generated summary
Why It Matters
NVIDIA and Microsoft have long collaborated on operating system and hardware integration. Technologies such as GeForce and DirectX are products of this partnership.
Gathering together at Dogpatch Studios in San Francisco, NVIDIA CEO Jensen Huang and Microsoft CEO Satya Nadella shared their vision that will shape the future of personal computers. Huang reminded that Windows was at the center of NVIDIA's founding story and emphasized that the GeForce brand would not exist if it were not for Windows. The duo is now taking this long-established partnership to the next level with the “agent era”.
Speaking at the event, Microsoft Windows and Devices Senior Vice President Pavan Davuluri announced that the Microsoft Execution Containers (MXC) infrastructure, which enables artificial intelligence agents to run safely, controlled and uninterrupted in the background of Windows, is now available for general use. This structure is directly integrated into the operating system kernel; It paves the way for autonomous assistants that work in the background by keeping data locally without bypassing firewalls. Jensen Huang likens this step to the revolution that DirectX once created in the game and application world.
RTX Spark Architecture: 128 GB Unified Memory and 1 Petaflop Power
The biggest surprise of the event on the hardware side was undoubtedly RTX Spark. This new platform appeals to developers, content creators and players; It brings full-fledged artificial intelligence capability to laptops and compact desktop systems. Announced as the first representative of this platform, the new Surface Laptop Ultra is built directly on RTX Spark.
Technical Details and Hardware Capabilities
Processor and Graphics Power: NVIDIA Blackwell RTX GPU scaling up to 6,144 cores and NVIDIA Grace CPU up to 20 cores come together with a dedicated interconnect at 600 GB/s.
Artificial Intelligence Performance: Full 1 petaflop of computing power and up to 128 GB unified memory are offered in FP4 format.
Advanced Local Models: Huge language models with 125 billion parameters, such as Qwen 3.8 Flash Next, can run completely offline and unlimitedly on the device without sending data to the cloud.
For Gamers and Creators: 5th generation Tensor cores, NVFP4, AV1 hardware acceleration, DLSS 5 support and smooth AAA gaming performance over 100 FPS at 1440p resolution are provided.
While the pre-order process for laptops with RTX Spark architecture has started, the devices will be on the shelves as of October 16. While models signed by manufacturers such as Acer, ASUS, Dell, HP, Lenovo, Microsoft, MSI and Gigabyte are available to users, compact desktop solutions that will work 24/7 uninterruptedly go on sale in November.
Desktop SuperPC: NVIDIA DGX Station for Windows
The most critical move targeting the corporate segment came with the Windows-supported NVIDIA DGX Station. The DGX Station platform, which previously worked only in the Linux environment, now enters the Windows ecosystem directly with the GB300 Grace Blackwell Ultra Desktop Superchip. This step completely breaks down the walls between enterprise companies' AI researchers and their Windows-based office and management infrastructures.
This desktop supercomputer, which has 748 GB coherent memory and 20 petaFLOPS FP4 processing power, allows you to run and fine-tune frontier models with up to 1 trillion parameters in the local environment. Developers can manage core workflows directly from the Windows desktop while still accessing Linux toolsets via WSL (Windows Subsystem for Linux) when needed.
What to Watch
AI outlook — possibilities, not facts
RTX Spark-based laptops will be available starting October 16.
Very likely · Within days
Compact desktop RTX Spark solutions will be available in November.
Very likely · Within weeks
Windows-powered NVIDIA DGX Station can be managed directly from the Windows desktop for enterprise AI workloads.
Likely · Within months
Open Questions
- What are the security certifications and standards of the MXC infrastructure?
- What are the starting price ranges of RTX Spark-based devices?
- What will the licensing and pricing model be like for the Windows version of DGX Station?






