Apple reported record revenue for the fourth quarter, alongside a $33 billion cost linked to rising memory prices. Its results also pointed to a shift in how some businesses are adopting AI: rather than relying solely on cloud services, they are running inference on local devices, partly to manage costs and address security and data-governance concerns.
Apple said Gap Inc. runs loss-prevention AI models on Mac mini computers to avoid cloud token costs. Christie’s uses Mac Studio systems for local pricing-estimate models.
The company also cited other large deployments. American Airlines has purchased 25,000 iPads for pilots and mechanics and is deploying more than 37,000 iPhone 17 devices to flight attendants for in-flight purchases using Tap to Pay. In the Philippines, the Department of Education plans to deploy 170,000 MacBook Neo computers to teachers and staff.
Apple argues that its M-series and A-series processors can support demanding AI workloads on-device. The report says four Macs can be combined into an on-premises system for complex calculations, offering businesses an alternative to paying for ongoing cloud inference.
Apple is also setting out options for developers using its AI services. Developers whose apps have fewer than two million downloads can access Apple’s server-side language models without token charges, the company said. For developers using its Private Cloud Compute infrastructure, Apple has introduced three service tiers. It is also developing plans to charge iCloud+ customers for heavy use of server-side queries.
CEO John Ternus said Apple’s products are intended to support both Apple’s own and third-party AI applications. He also described Siri’s current system as request-based, while saying its orchestrator-driven architecture leaves room for future capabilities. Ternus added that Apple wants users to understand what access they grant to agentic tools. ilaas
Sources: Computerworld.