Artificial intelligence (AI) companies are in intense search for powerful hardware, which has led to a shortage of products on the market. A recent report, dated August 30, 2026, indicates that OpenAI is purchasing tens of thousands of Apple Mac mini and Mac Studio devices. Furthermore, the new NVIDIA RTX Spark chips, even before their launch scheduled for next fall, are already sold out, directly impacting the PC sector. This scenario takes place in the midst of “chipflation” that has kept ordinary consumers away from the technology market.

Labs and other AI-focused organizations are showing strong interest in anything with robust processing power, including Apple Macs and upcoming NVIDIA RTX Spark computers.

Apple’s Mac mini and Mac Studio models saw strong sales at the start of the year. Its unified memory architecture and low-latency Thunderbolt 5 feature, which optimizes direct connections between PCs without the standard network layer, have established themselves as ideal solutions for managing AI agents on devices.

As per a new report from The Information, OpenAI alone has acquired tens of thousands of Mac mini and Mac Studio in recent months, likely for reinforcement learning work and AI agent processing. Anthropic, another company in the sector, would also be using these Apple devices through Amazon AWS.

Apple recently introduced the Mac mini M6 and the updated Mac Studio Ultra M5. These new versions are expected to stimulate additional demand from cutting-edge AI laboratories.

Meanwhile, according to Taiwan’s Economic Daily newspaper, the initial batch of the NVIDIA RTX Spark superchip has already been completely sold. Major manufacturers such as Asus, MSI, Dell, HP, Lenovo and Microsoft have guaranteed all systems equipped with the N1x variant.

NVIDIA’s RTX Spark platform (model N1x) uses TSMC’s 3nm technology and includes the following features:

• A 20-core Grace processor.
• A Blackwell RTX 5070 GPU, with 6,144 CUDA cores, delivering AI performance of up to 1 PFLOP FP4.
• Up to 128 GB of LPDDR5X-based unified memory.
• NVLink-C2C bandwidth of approximately 600 GB/s between CPU and GPU.
• Full support for NVIDIA software such as CUDA, TensorRT, DLSS, Reflex, G-SYNC, and RTX ray tracing.

The 20-core Grace processor present in the N1x variant is made up of 10 ARM Cortex-X925 cores and 10 ARM Cortex-A725 cores.

The N1 variant features a 12-core CPU, made up of 8 ARM Cortex-X925 cores and 4 ARM Cortex-A725 cores. This version also has a GeForce RTX 5050 GPU and up to 64 GB of unified memory.

Computers with NVIDIA RTX Spark are expected to be launched in the autumn of the northern hemisphere. The N1x systems are estimated to cost around NT$110,000, which is equivalent to approximately US$3,400.

According to the report, both ASUS and MSI have already exhausted their initial quotas and are now requesting a new shipment from NVIDIA.

Hu Shubin, co-CEO of ASUS, stated that customer orders for the new RTX Spark series products have exceeded expectations, a notable feat given the high costs of these systems.

Customer perception is that these devices are not common PCs, but rather rare and complete models, with all the power of NVIDIA graphics cards in a notebook format, aimed at specific AI uses.