Apple unveils updated Mac Mini M6 and Mac Studio M5 Max/Ultra with AI-optimized silicon

Apple has finally refreshed two of its most popular desktop machines after months of constrained supply. The new Mac Mini ships with an M6 chip (with an M5 Pro option), while the Mac Studio gets the beefier M5 Max and M5 Ultra processors. Both lines are built around Apple’s latest silicon, which leans heavily into AI workload optimization through beefed-up Neural Engine cores and a unified memory architecture that lets the CPU, GPU, and neural cores share a single pool of high-bandwidth RAM.

Pre-orders open September 22, 2026. The Mac Mini starts at $899, which is $200 more than its predecessor’s entry price. The Mac Studio begins at $2,499.

What’s actually inside

The Mac Studio’s M5 Ultra configuration supports up to 768 GB of unified RAM, paired with a 30-core CPU and 64-core GPU.

Apple is targeting professional users in creative production, scientific computing, and AI model development. The unified memory architecture is particularly relevant for machine learning workflows, where large language models and image generators need to load massive parameter sets into memory all at once. Traditional machines with discrete GPUs hit a wall when model sizes exceed VRAM capacity. Apple’s approach sidesteps that limitation by letting the entire memory pool serve any processor that needs it.

The $200 question

The price bump on the Mac Mini is notable because Apple rarely raises base prices on its most accessible desktop. The company appears to be passing along higher component costs rather than absorbing them.

Bloomberg’s Mark Gurman and MacRumors have both reported that RAM constraints are expected to keep premium pricing and limited availability of higher-end configurations in place well into late 2026. The previous Mac Mini and Mac Studio models had already experienced extended wait times due to high demand.

The AI hardware landscape gets more crowded

Apple’s refresh lands at a moment when the AI hardware market is intensely competitive. Nvidia dominates the data center and cloud training segments with its H100 and successor GPUs. AMD has made inroads with its MI300X accelerators. On the consumer and prosumer side, companies like Qualcomm and Intel are embedding neural processing units into laptop and desktop chips.

Memory suppliers stand to benefit as well. Companies involved in high-bandwidth memory production, including SK Hynix and Samsung, are already seeing elevated demand from AI-related hardware across the industry. Apple’s commitment to configurations with up to 768 GB of unified RAM only adds pressure to an already tight supply chain.

Independent benchmarks for the M6 and M5 Ultra have not yet been published. The real test comes after September 22, when real-world benchmarks start appearing.