Apple hardware chief says new Macs have no cost per token

Apple Inc. is positioning its newest Macs as a cost-effective alternative to cloud-based AI, arguing that local hardware eliminates the recurring per-token charges that define current AI infrastructure economics. Apple hardware chief Johny Srouji told Reuters, "There's no cost per token. You're just using the machine again and again." This pitch directly challenges the revenue model driving the recent surge in AI spending.

To demonstrate the feasibility of this approach, Apple showed four Mac Studios running a trillion-parameter AI model as a cluster. Reuters reported that this setup could operate from a single wall outlet. In Apple's framing, businesses pay a one-time hardware cost to run models locally, avoiding continuous payments to cloud providers or AI companies for usage. However, the upfront capital expenditure is significant. The M5 Ultra Mac Studio starts at $5,499, while a configuration with 256GB of memory and 16TB of storage costs $18,299. A 512GB configuration capable of handling larger models is scheduled to arrive in late October. Apple previously raised Mac prices in June, citing surging memory and storage costs.

Nvidia Corp. maintains a different view of the AI economy. CEO Jensen Huang stated in Nvidia's latest earnings release, "Its tokens are productive and profitable. Now, compute is revenue." This perspective aligns with Nvidia's financial performance, where Data Center revenue reached $89 billion last quarter, a 117% increase from the year-earlier period, driven by heavy spending on computing infrastructure by cloud providers and AI companies.

Apple's strategy suggests that some inference workloads may shift from rented data-center capacity to local machines. Nvidia is pursuing a similar shift with its DGX Spark and RTX Spark products, which are designed to run capable AI models locally. This makes desktop AI a new battleground between the two companies. Despite Apple's push, prediction-market traders remain skeptical of an immediate shift in market dominance. Polymarket data indicates a 78% chance that Nvidia will end 2026 as the world's largest company, compared with 15% for Apple, based on about $7.5 million in trading volume.

The enterprise market remains heavily dominated by competitors. According to IDC data cited by Reuters, Apple held just 4.6% of the enterprise PC market, while Windows held 91.3%. Furthermore, Apple's own most advanced cloud model reportedly runs on Nvidia GPUs through Google Cloud, highlighting the continued reliance on Nvidia's hardware even within Apple's ecosystem.