ASUS, MSI Exhaust First NVIDIA RTX Spark
ASUS and MSI have sold their initial RTX Spark allocations and are asking NVIDIA for more supply ahead of the fall launch, a report says.
ASUS and MSI have already exhausted their first batch of NVIDIA's upcoming RTX Spark systems and are asking for more supply. This news comes from Taiwan's Economic Daily, which reports the initial batch of the superchip is sold out to several major PC makers.
Demand from channel customers for the new RTX Spark series products has been better than expected, according to ASUS co-CEO Hu Shubin. He noted this is a significant achievement given the high prices of these systems. Customers appear to be treating them not as standard PCs but as scarce products with a full NVIDIA software stack integrated into a notebook form factor.
NVIDIA RTX Spark Platform Specifications
The report outlines the specifications for two variants of the new NVIDIA RTX Spark platform, which leverages TSMC's 3nm technology. The higher-end N1x system is reportedly priced around NT$110,000, or roughly $3,400.
| Variant | CPU Configuration | GPU | Unified Memory | Key Features |
|---|---|---|---|---|
| N1x | 20-core Grace (10x ARM Cortex-X925, 10x ARM Cortex-A725) | Blackwell RTX 5070 (6,144 CUDA cores) | Up to 128GB LPDDR5X | ~1 PFLOP FP4 AI performance, ~600 GB/s NVLink-C2C bandwidth, full NVIDIA stack (CUDA, TensorRT, DLSS, etc.) |
| N1 | 12-core (8x ARM Cortex-X925, 4x ARM Cortex-A725) | GeForce RTX 5050 | Up to 64GB | Full NVIDIA software stack support |
Broader AI Hardware Demand
This surge for NVIDIA's new platform fits a wider pattern of intense demand for hardware capable of AI workloads. A separate report from The Information suggests OpenAI alone has purchased tens of thousands of Apple Mac mini and Mac Studio devices in recent months. The report states these are presumably for reinforcement learning and agentic workloads.
Apple's devices have reportedly been popular due to their unified memory architecture and new low-latency Thunderbolt 5 feature. This technology bypasses the standard TCP/IP networking stack, enabling fast, low-latency PC-to-PC connections suitable for on-device agentic orchestration. Even rival AI lab Anthropic is renting these Apple devices via Amazon AWS.
Apple recently unveiled new M6 Mac mini and M5 Ultra Mac Studio models. The source report states these refreshes are also expected to drive another wave of demand from AI labs.
The source, Wccftech, frames this against a backdrop where the average consumer is being priced out of the market due to memory 'chipflation.' In contrast, machines suited for AI are selling rapidly. NVIDIA's RTX Spark-based PCs are expected to debut this fall.
