Apple Silicon Macs gain eGPU support for AI workloads as Apple approves TinyGPU driver

Apple Approved Drivers Restore eGPU Functionality to Apple Silicon

Apple has officially approved TinyGPU drivers developed by Tiny Corp, marking a significant shift for Apple Silicon users who have been unable to use external GPUs since the M1 processor launched in November 2020 1

1

. The approval means users can now connect AMD and Nvidia eGPUs to their Macs without disabling System Integrity Protection or relying on workarounds that previously compromised system security 1

1

.

Tiny Corp announced the development on X, stating that installation has become remarkably straightforward: "It's so easy to install now a Qwen could do it, then it can run that Qwen" . While Intel-powered Mac devices previously supported external GPUs on macOS, Apple silently removed this capability when transitioning to its proprietary processors 4

4

.

TinyGPU Driver Enables AI Acceleration on Mac Mini and Beyond

The TinyGPU driver supports macOS 12.1 or later on devices equipped with Thunderbolt/USB4 ports, enabling Mac Mini owners and other Apple Silicon users to run larger local AI models . Compatible hardware includes AMD GPUs from the RDNA3 generation onward and Nvidia cards from the Ampere series and later .

The timing proves particularly relevant given the Mac Mini's unexpected popularity for hosting OpenClaw agents. Apple Store employees have reportedly begun calling the device the "OpenClaw machine" due to its effectiveness at running AI agents 2

2

. This demand has contributed to severe shortages of high-memory Mac configurations, with delivery windows extending from six days to six weeks 1

1

. Apple has even discontinued the 512GB Unified Memory option for Mac Studio while raising the price of the 256GB model by $400 1

1

.

AI Workloads Only: No Gaming Support Included

The TinyGPU driver focuses exclusively on AI based tasks rather than traditional graphics rendering or gaming applications 1

1

. Users can run AMD workloads natively, while Nvidia GPUs require Docker Desktop to execute AI computations through NVCC . Once installed and approved, demanding models such as Qwen 2.5 27B can run effectively through the tinygrad framework .

Tiny Corp first tested an eGPU on Apple Silicon in May 2025, but official approval eliminates the need for security compromises 1

1

. The company, known for its tinybox AI accelerator and previous conflicts with AMD over driver issues that required CEO Dr. Lisa Su's intervention, currently sells the red v2 with four AMD 9070XTs for $12,000 and the green v2 Blackwell with four RTX Pro 6000 Blackwell GPUs for $65,000 1

1

.

Implications for AI Training and Inference on Mac

This development arrives as Apple has permanently discontinued the Mac Pro, removing the device from its website and redirecting the flagship model's page to the general Mac homepage . With no M4 Ultra variant planned and limited sales suggesting weak demand for high-end desktop configurations, eGPU support now provides the modular expansion path that Mac Pro previously offered .

For professionals and enthusiasts working with AI, the TinyGPU driver enables AI training and inference capabilities without requiring dedicated AI supercomputers, though some limitations apply 1

1

. While integrated GPUs on Apple Silicon M1, M2, M3, M4, and M5 processors and their Ultra and Max variants deliver strong performance, they may not meet everyone's demands for advanced AI acceleration 4

4

. External GPUs now provide the additional on-demand processing power that AI researchers and developers running models locally require 4

4

.