At GTC Taipei, NVIDIA introduced RTX Spark — a "superchip" that combines CPU and GPU on a single chip, allowing compact laptops and desktops to run large language models (LLMs) along with AI agents right on the machine, with no need for the cloud. The NVIDIA newsroom dates the announcement to 31/05/2026, while Al Jazeera reported it on 01/06/2026 (the difference is due to time zones). It is described as a step that brings data-center-class AI capability straight down to the personal desk.
Quick summary
- When: Announced at GTC Taipei — the NVIDIA newsroom lists 31/05/2026; Al Jazeera lists 01/06/2026 (difference due to time zones).
- What it is: RTX Spark — a CPU+GPU superchip for PCs to run large LLMs & AI agents right on the machine, no cloud needed.
- Performance: 1 petaflop of AI; up to 128GB of unified memory; Blackwell RTX GPU with 6,144 CUDA cores; 20-core Grace CPU.
- Capable of: LLMs up to 120 billion parameters, context up to 1 million tokens, all locally.
- Who sells it: First machines in autumn 2026 from ASUS, Dell, HP, Lenovo, Microsoft Surface, MSI (Acer, GIGABYTE later).
What is RTX Spark?
According to NVIDIA's announcement, RTX Spark is a "superchip" that combines CPU and GPU in a single design, aimed at compact laptops and desktops. The core point is that it lets you run large LLMs and AI agents right on the personal machine instead of having to send requests to the cloud. In other words, the AI processing power that previously required servers in a data center is now brought straight down to the PC on the user's desk.
NVIDIA developed RTX Spark in partnership with Microsoft (the Windows operating system) and MediaTek, in order to create a new generation of PCs that run AI directly on the device.
Specifications & performance
NVIDIA announced the key specifications of RTX Spark:
- 1 petaflop of AI performance.
- Up to 128GB of unified memory.
- Blackwell RTX architecture GPU with 6,144 CUDA cores.
- 20-core Grace CPU.
With this configuration, RTX Spark is said to be able to run LLMs up to 120 billion parameters with context up to 1 million tokens right on the machine. Beyond AI, NVIDIA says the platform can also handle 3D rendering of 90GB and above, 12K video in 4:2:2 format, 4K AI video, and AAA gaming at 1440p resolution at over 100 fps.
AI runs locally, no cloud needed
The biggest highlight of RTX Spark is that AI runs directly on the device (on-device). Instead of every query having to go through the provider's cloud, the model and data can sit right on the PC. This opens the way to using AI agents and large LLMs without depending on a network connection and without sending data outside — a significant change in both response performance and data control.
Who sells it & when?
According to NVIDIA, the first machines using RTX Spark are expected to launch in autumn 2026, coming from ASUS, Dell, HP, Lenovo, Microsoft Surface and MSI; Acer and GIGABYTE will join later. The partnership with Microsoft and MediaTek shows that NVIDIA aims to broadly popularize this AI PC line through the Windows ecosystem.
Leadership statements
Jensen Huang (CEO of NVIDIA) called this the biggest "reinvention of the PC" in 40 years, and stated: "This is going to be the new PC."
Satya Nadella (CEO of Microsoft) set out the goal of bringing intelligence to every home and every desk through Windows.
FAQ
How is RTX Spark different from using AI on the cloud?
RTX Spark lets you run large LLMs and AI agents right on the personal PC (on-device), with no need to send queries to the cloud. According to NVIDIA, it can run models up to 120 billion parameters with context up to 1 million tokens, all locally.
When will machines using RTX Spark be available to buy?
NVIDIA says the first machines are expected to launch in autumn 2026, from ASUS, Dell, HP, Lenovo, Microsoft Surface and MSI; Acer and GIGABYTE will join later.
Why does the announcement list two different dates?
RTX Spark was announced at GTC Taipei. The NVIDIA newsroom lists 31/05/2026, while Al Jazeera reported 01/06/2026 — the difference is due to time zones. This article notes both dates for accuracy.
AI runs locally — data & operational control stay in your enterprise
The "local/on-device AI" trend that RTX Spark represents aligns with how Namtech deploys internal AI on-premise: models run on the enterprise's own infrastructure, data never leaves the organization, and operational control stays in your hands.
Book a free consultationNote: This article is compiled from public sources as of 22/06/2026; specifications and the launch schedule may change. For reference only.