Apple Approves Tiny Corp Nvidia eGPU Driver for Arm Macs, Opening Door to External GPU Computing
Nvidia GPUs Finally Work on Arm Macs — But There Is a Catch
In a move that would have seemed impossible a year ago, Apple has approved a driver that lets Nvidia eGPUs work with Arm-based Macs. The driver was built by Tiny Corp (of tinygrad fame), not Nvidia, and is specifically designed for LLM inference workloads.
What Changed
Previously, using Nvidia eGPUs on Apple Silicon Macs required disabling System Integrity Protection (SIP), Apple security feature. Now, Apple is allowing the Tiny Corp driver to be signed, meaning users can connect external Nvidia GPUs without compromising system security.
The Catch
This is not a plug-and-play solution:
- The driver is built by Tiny Corp, not Nvidia, and is community-maintained
- You need to compile it with Docker before use
- It is designed primarily for LLM workloads, not general GPU computing
- No official Nvidia support — you are on your own if something breaks
Why Apple Allowed This
Several factors may explain the decision:
- Developer demand — Machine learning developers have long wanted Nvidia GPU access on Macs
- Competitive pressure — AI developers increasingly choose Linux or Windows over macOS for GPU work
- Harmless to Apple GPU narrative — The driver is for external, user-supplied GPUs, not competing with Apple Silicon
- Security maintained — Signed driver means no SIP bypass needed
Why It Matters
This opens a new era for Mac-based AI development:
- Developers can use their Mac as a workstation with Nvidia GPU for training
- The tinygrad ecosystem gains a significant use case
- Apple may be signaling a more open approach to GPU interop
- The barrier to AI development on macOS just dropped significantly
The Tiny Corp Factor
Tiny Corp, known for the tinygrad deep learning framework, has been pushing the boundaries of GPU accessibility. This driver is another example of their philosophy: making high-performance GPU computing available on any hardware.