Thought Leadership
Apple is marketing its new Mac computers as a cheaper alternative to rented cloud computing power for enterprise AI tasks.
Apple is launching revised Mac mini and Mac Studio models this week and, according to a report by Reuters, is positioning them specifically for corporate customers looking to run AI workloads locally rather than through rented data centers. The devices are set to ship starting Tuesday.
No Price Per Token
Apple’s argument to corporate customers is that running AI hardware locally can be cheaper than recurring payments to cloud providers like Microsoft or language model vendors like OpenAI and Anthropic, which bill based on so-called tokens, the basic unit of computation in AI applications. Johny Srouji, Apple’s Chief Hardware Officer, explained:
“Once you have the machine on your desk, you paid for it. I think we offer really great value, not only on performance, but on price. There’s no cost per token. You just use the machine over and over again.”
— Johny Srouji, Apple’s Chief Hardware Officer
Hardware with a High Purchase Price
The new devices are primarily aimed at compute-intensive tasks such as writing code or complex business workflows. The purchase price is substantial: the Mac Studio with the M5 Ultra chip starts at 5,499 US dollars, while a configuration with 256 gigabytes of RAM and 16 terabytes of storage costs 18,299 US dollars. A version with 512 gigabytes of RAM, capable of processing even larger models, is set to follow at the end of October. Apple had already raised Mac prices in June, citing rising memory costs.
Four Mac Studios for a Trillion-Parameter Model
During a demonstration at an Apple facility in California, the company showed four networked Mac Studio computers that, according to Reuters, jointly ran an AI model with one trillion parameters while being powered by a single wall outlet. Apple states that the foundation for this is the unified memory architecture of its proprietary Apple Silicon chips, along with support for RDMA over Thunderbolt 5, which enables efficient networking of multiple devices for local computation.
Competition from Nvidia and Microsoft
With this push, Apple enters into direct competition with new desktop computers from Nvidia and other PC manufacturers, which are expected to take center stage at an upcoming Microsoft Windows event next month in San Francisco. Nvidia is also pursuing a similar strategy with its own products, such as DGX Spark and RTX Spark, to run AI models directly on local devices rather than in data centers.
Small Market Share in Enterprise Business
Despite the new orientation, Apple’s position in the enterprise business remains comparatively small. According to data from market research firm IDC, Apple recently held a share of around 4.6 percent of the enterprise PC market, compared to 91.3 percent for Windows devices.
Foundation: Apple’s Proprietary Chip Architecture
Apple’s foray into the market for AI-capable enterprise hardware builds on its proprietary chip architecture, Apple Silicon, introduced in 2020. Observers note that the unified memory architecture developed as part of that initiative—originally designed to improve iPhone battery life—is proving well-suited for AI workloads as well.
(Editorial Team)
