Apple Signs Tiny Corp Driver: Nvidia eGPUs Finally Work on Apple Silicon Macs
A Historic First: Apple Approves Third-Party Nvidia Driver for Arm Macs
In a move that would have seemed impossible just months ago, Apple has approved and signed a third-party driver from Tiny Corp that enables Nvidia eGPUs to work with Apple Silicon Macs — without requiring users to disable System Integrity Protection (SIP).
The Significance
Apple has historically maintained tight control over macOS hardware support, particularly for GPUs. The transition to Apple Silicon in 2020 appeared to permanently close the door on external GPU support, as Apple's own M-series chips integrated powerful GPUs directly into their SoCs.
The Tiny Corp Solution
Tiny Corp, the company behind the Tinygrad AI framework and TinyBox AI training hardware, developed the driver specifically for running LLMs on Nvidia GPUs connected to Apple Silicon Macs. Key details:
- The driver is signed by Apple — no need to disable SIP
- Requires compilation via Docker — not a simple plug-and-play experience
- Primarily designed for LLM inference workloads, not gaming
- The driver is Tiny Corp's work, not Nvidia's official driver
What This Enables
For AI developers and researchers who prefer the Mac ecosystem but need Nvidia GPU performance:
- Local LLM development with full CUDA support
- Cost-effective inference using existing Nvidia eGPU hardware
- Cross-platform workflows combining Apple's hardware with Nvidia's AI software stack
- No trade-off between Mac UX and AI capability
Why Apple Allowed It
Apple's approval likely reflects strategic considerations:
- Growing AI developer ecosystem demand
- Recognition that CUDA lock-in limits Apple's AI ambitions
- Tiny Corp's focus on LLMs rather than gaming keeps it from competing with Apple's own GPU strategy
- The Docker compilation requirement maintains a technical barrier that keeps it a developer tool, not a consumer feature
Industry Context
This comes as Tiny Corp positions itself as a bridge between different AI hardware ecosystems. Their TinyBox hardware and Tinygrad framework are designed to make AI training accessible beyond the Nvidia monopoly. The signed driver represents another step in making Apple Silicon a viable platform for serious AI work.