Apple Ships AI Desktops, Targets Cloud Cost Savings for Business
Apple has commenced shipping new desktop computers, with company executives promoting them to corporate buyers as a more economical alternative to renting data centres, according to Reuters. The upgraded Mac Minis and Mac Studios, costing up to USD 20,000, are designed to handle artificial intelligence tasks locally. These machines will compete directly with new desktop systems from Nvidia and other personal computer manufacturers.
These new Mac models are intended for complex artificial intelligence applications such as code generation and intricate business operations. They aim to reduce reliance on "tokens", which represent the basic unit of artificial intelligence computing, from cloud service providers like OpenAI and Anthropic. Apple's objective is to apply its expertise in optimising battery powered devices, such as iPhones, to challenge Microsoft's established position in the corporate desktop sector. However, this presents a considerable undertaking, given Apple's approximately 4.6 per cent share of the enterprise desktop market, compared to Windows' 91.3 per cent, according to IDC's Linn Huang.
Steve Jobs, an Apple co-founder, was reportedly hesitant about enterprise computing due to end users' inability to select their preferred products. But Apple has since established a strong position in the artificial intelligence desktop segment, attributed partly to a consistent focus on power efficiency. This commitment originated when Apple introduced its initial Apple Silicon chips several years ago. The organisation integrated computing and memory components, traditionally separate in personal computers, into a unified memory architecture to extend battery life.
The close integration between computing and memory components inadvertently enhanced Macs' capability for artificial intelligence processes. This design, which Nvidia and other hardware producers have only recently begun to adopt, facilitated an increase in Mac Mini sales. This surge coincided with the rise of OpenClaw, an open source agentic artificial intelligence tool, which gained traction in markets including China. Mac Studios, initially developed for creative professionals involved in video or music production, have also received specialised artificial intelligence features in recent years, such as bespoke chip to chip networking called RDMA over Thunderbolt.
At a recent product launch event, Apple demonstrated four Mac Studios connected to operate an artificial intelligence model with a trillion parameters, a measure of computational complexity. This setup was used to identify and correct a graphics coding fault. Such operations typically necessitate a data centre, yet this array of Macs functions from a single wall outlet. Johny Srouji, Apple's chief hardware officer, stated that once a machine is on a user's desk, it has been paid for, offering value in both performance and cost. He noted there is no per token cost; the machine is simply used repeatedly.
And Microsoft is pursuing a similar market for what its Chief Executive Officer, Satya Nadella, terms the "unmetered intelligence" of on-device artificial intelligence. Mr Nadella has also indicated that Microsoft intends to incorporate many of its artificial intelligence functions into a "super app" for Windows. However, Microsoft's historical dominance in corporate computing has required supporting hardware from a wide array of vendors. This variety can create additional work for Windows devs seeking to maximise chip performance.
Microsoft stated it has been collaborating with chip partners to streamline artificial intelligence operations through its Windows Machine Learning tools. The company further indicated that features such as Remote Direct Memory Access, or RDMA, represent an active area of investment. Nvidia declined to provide a statement. However, at the introduction of its new personal computer chip a few months ago, its Chief Executive Officer, Jensen Huang, minimised any direct competition with Apple, asserting that Nvidia's focus remains on expanding the capabilities of Windows personal computers.
Nvidia's primary concentration remains within the data centre sector. Mr Srouji explained that Apple presents corporate buyers with the concept that on-device artificial intelligence models developed on its devices can be scaled across its product range. This includes its most expensive Mac Studios down to its most economical iPhones and iPads. This scalability is possible because their respective chips share fundamental design principles and architecture. He concluded that Apple offers a variety of products from which customers can choose based on individual requirements.
Apple positions its new Mac Mini and Mac Studio desktops as cost-effective alternatives to data centre rental for artificial intelligence tasks.
These upgraded Macs, priced up to USD 20,000, are engineered for local artificial intelligence processing, competing with offerings from Nvidia and other PC makers.
Apple's current market share in enterprise desktops is approximately 4.6 per cent, significantly trailing Windows' 91.3 per cent.
The Mac's ability to handle artificial intelligence tasks is partly attributed to its unified memory architecture, originally designed for power efficiency.
Johny Srouji, Apple's chief hardware officer, stated that on-device artificial intelligence removes per token costs associated with cloud computing.
Source: Reuters
