Apple launches M5 Ultra Mac Studio with major hardware upgrades and steep price hikes
Apple's new Mac Studio equipped with the M5 Ultra chip offers doubled CPU, GPU and memory bandwidth compared with the previous generation, but its base price starts at $12,299, reflecting a significant increase amid a broader AI-driven memory shortage.
Apple announced the latest Mac Studio, now powered by the M5 Ultra chip, which combines two M5 Max chiplets into a four-chiplet package. The configuration delivers up to 12 performance cores, 24 efficiency cores, 80 GPU cores, 32 Neural Engine cores and more than 1.2 TB/s of memory bandwidth, roughly twice the specifications of the M5 Max. The compact chassis remains unchanged from earlier models, retaining a 7.7-by-7.7-inch footprint and a similar port selection, with the Ultra variant adding extra Thunderbolt 5 ports and support for up to eight external displays.
Performance testing shows the M5 Ultra leading Apple Silicon in multi-core CPU speed and GPU throughput, though gains are not fully linear; in some workloads the M5 Max matches or exceeds the Ultra, possibly due to power-limit throttling within the existing thermal design. Benchmarks indicate 30-50 % faster single-core CPU performance and 33-66 % faster GPU performance over the prior-generation M3 Ultra, while token-generation rates for local AI models rise from about 31 tokens per second on the M5 Max to just over 50 on the M5 Ultra.
Pricing has risen sharply across the Mac Studio line. The base M5 Ultra model with 96 GB of RAM and 1 TB of storage costs $12,299, up from $5,599 for a comparable M3 Ultra configuration 18 months earlier. Fully specced versions are expected to exceed $20,000, positioning the machine as a high-end workstation for developers and creators who run large language models locally and prefer on-device processing over cloud services.
Analysts note that Apple Silicon’s unified memory architecture and efficient power usage make the Mac Studio attractive for local AI workloads, offering more GPU memory than most consumer graphics cards and competitive memory bandwidth. However, the steep price increases and limited scalability of performance relative to hardware resources may restrict the Ultra model to a niche market of professionals willing to invest heavily in on-premise AI capabilities.