MSI released a new mini computer “EdgeMesa N AI+” at Computex 2026, equipped with the RTX Spark platform officially launched by NVIDIA this week. This is a palm-sized machine, but it is packed with a 20-core Arm CPU, Blackwell architecture GPU and 128GB of unified memory. The product positioning is very clear: bring cloud AI inference to the desktop.
MSI launches EdgeMesa N AI+ mini PC powered by NVIDIA RTX Spark chip
What is RTX Spark?
RTX Spark is a new Arm architecture super chip released by NVIDIA at Computex 2026. It aims to enable Windows PCs to have local AI inference capabilities. In terms of specifications, this chip is equipped with 20 Arm CPU cores, Blackwell architecture GPU (6,144 CUDA cores), and 128GB LPDDR5X unified memory with a bandwidth of 300 GB/s. The CPU and GPU are interconnected through NVLink C2C. NVIDIA states that this architecture is sufficient to run a language model with 120 billion parameters locally, and the context length can be extended to one million tokens.
In other words, large language models that used to need to be connected to a cloud server to run can now be run on a mini computer. NVIDIA positions this platform as the foundation of “Agentic AI OS” and hopes to transform Windows from a traditional operating system to an environment that can run AI Agents.
EdgeMesa N AI+ Specifications and Design
MSI’s EdgeMesa N AI+ is one of the first mini PCs to feature RTX Spark, along with HP OmniDesk Mini PC and ASUS ProArt Mini PC. All three machines target the same market: developers, data scientists and content creators who need local AI computing power.
The I/O configuration of EdgeMesa N AI+ is quite generous:
- network: 10GbE Ethernet network to meet the high bandwidth requirements for AI model training and large data set transmission
- show output: 1 set of HDMI + 3 sets of USB 20Gbps Type-C, driving up to four screens at the same time
- memory: Using a unified memory architecture (128GB LPDDR5X), the GPU and CPU share the memory pool, and there is no need to move data between the two when processing large models.
Although the body is small, MSI emphasizes the introduction of advanced cooling technology and heat flow management design to ensure stability and low noise even under long-term high-load operation. This is a fundamental requirement for AI workflows that require continuous runs of inference or training.
Who needs this machine
EdgeMesa N AI+ application scenarios include: large-scale language model deployment and inference, generative AI content creation, edge AI computing, smart analysis, etc. The industry aspect covers medical care, retail, financial services, robotics, automation control and smart cities.
There are two most direct usage scenarios for RTX Spark mini computers: first, AI developers can run model tests locally without burning cloud fees for each inference; second, enterprises that need to process sensitive data can keep AI calculations locally to avoid data outflow.
128GB of unified memory is key. The GPU memory of traditional PCs usually only ranges from 8 to 24GB. To run slightly larger models, it must be quantized or sliced. RTX Spark’s 128GB memory pool allows developers to directly load large models with full accuracy, a capability that was previously only available at the workstation level or even at the server level.
Time to market and market competition
MSI has not yet announced the price and exact release date of EdgeMesa N AI+. It is estimated that RTX Spark related products will not be launched until August or September this year at the earliest.
The direct competitor of the RTX Spark mini computer is actually the Apple Mac Studio. The M-series chips installed in Mac Studio also use a unified memory architecture, which has a solid foundation in the creator and developer market. NVIDIA is entering the market this time by stacking the CUDA ecosystem and complete AI software. The calculation is: since the AI development tool chain is almost all based on CUDA, the best choice for running AI locally should be NVIDIA hardware, not Apple.
Asus, HP, and MSI launched RTX Spark mini computers at the same time, which means that this is not a test of the water by a single manufacturer, but a collective bet by the entire PC industry on the new market of “local AI”. Want to check it out on site EdgeMesa N AI+ Friends can go to the MSI Computex booth to see actual MSI RTX Spark mini consoles and laptops.
Computex 2026 Exhibition Information
- Exhibition time: June 2 to June 5, 2026
- Exhibition location: Taipei Nangang Exhibition Hall, Hall 1, 4th Floor
- MSI booth：L0118
Source: KOCPC Chinese