Used Mac prices are climbing in a season when they normally drop. ATRenew, Apple and JD.com’s resale and refurbishment partner in China, reports that second-hand Apple product prices have risen approximately 15% and are holding at levels typically seen during the autumn iPhone announcement peak — despite it being spring, according to MyHostNews.

The driver: developers and hobbyists buying used Mac Minis to run OpenClaw AI agent fleets 24/7 on dedicated local hardware.

Why Mac Minis

Apple Silicon’s performance-per-watt ratio makes it unusually well-suited for always-on inference workloads. An M2 Mac Mini draws roughly 10-15 watts at idle and handles local model inference without requiring enterprise GPUs or cloud API subscriptions. For developers running multiple OpenClaw agents around the clock, the economics are simple: buy a $400-600 used Mac Mini once versus paying per-token API costs indefinitely.

But the hardware choice also reflects a security calculation. OpenClaw agents operate with broad system access — reading messages, browsing the web, executing code, interacting with APIs. Many users prefer to isolate that access on a separate machine rather than grant it to their primary workstation. ATRenew’s Jeremy Ji, CSO of the company, described the demand pressure as comparable to “during the pandemic,” CNBC reported.

The Mac Mini has effectively become a dedicated agent box: a cheap, quiet, efficient machine whose entire purpose is running autonomous AI agents isolated from the user’s primary workstation and personal data.

The Trade-Up Cycle

ATRenew is also observing a secondary effect. Users with older M1 and M2 Mac Minis are trading them in for M4 and M5 models to get more headroom for larger local models and more concurrent agents. That trade-up cycle is pushing ATRenew to raise its buyback prices to maintain supply, which in turn elevates prices across the entire used Mac market.

The result is a feedback loop: OpenClaw adoption drives demand for used Macs, which drives up prices, which makes newer models relatively more attractive, which drives more trade-ins, which tightens supply of the mid-range units that entry-level OpenClaw users want most.

Apple’s Unintended Position in the AI Hardware Stack

The pricing data surfaces an underreported dynamic in the AI hardware landscape. Apple has not released a foundation model. Apple has not announced an agent framework. Apple has not positioned itself as an AI infrastructure company. And yet Apple Silicon is collecting rent on a meaningful slice of the local AI agent deployment market simply because its chips run inference efficiently and quietly.

Jensen Huang called OpenClaw “the operating system of agentic AI” at GTC 2026 this week and shipped NemoClaw as a free security layer to accelerate enterprise adoption. Nvidia’s strategy depends on developers buying Nvidia hardware. But for the hobbyist and small-team segment that drives OpenClaw’s viral adoption, a used Mac Mini at $450 is beating a $3,000 GPU setup or recurring cloud costs.

Constraints

The trend has limits. China’s government has already issued warnings about OpenClaw use in professional environments due to security risks and prompt injection vulnerabilities. If regulatory pressure increases or high-profile security incidents multiply, demand could shift toward managed cloud solutions or hardened enterprise deployments that don’t rely on personal hardware.

Ji told reporters he expects the elevated demand to continue “throughout the year,” driven in part by GTC-related media coverage amplifying awareness of OpenClaw among non-developer audiences.

For now, the second-hand Mac market is doing something it rarely does in spring: behaving like it’s September.

Sources: MyHostNews, CNBC