Apple Signs Nvidia and AMD eGPU Driver for Apple Silicon Macs — But Only for AI

In a quiet but significant move, Apple has officially approved a third-party GPU driver that enables AMD and Nvidia external GPUs to work on Apple Silicon Macs. The driver, called TinyGPU, was developed by Tiny Corp (founded by George Hotz of iPhone jailbreaking fame) and is specifically designed for AI research and compute workloads — not graphics acceleration or gaming.

This is the first time Apple has officially signed a driver enabling Nvidia or AMD GPUs on Apple Silicon since the transition from Intel. For the AI research community, it’s a game-changer: Mac users can now harness the raw compute power of external GPUs for training and running AI models without disabling macOS security protections.

What the TinyGPU Driver Does

Why This Matters for AI Research

Apple Silicon Macs have excellent unified memory and power efficiency, but their GPU compute capabilities lag behind dedicated Nvidia or AMD cards for AI workloads. The TinyGPU driver bridges this gap:

• Run larger AI models by offloading compute to powerful external GPUs
• Train models faster using Nvidia CUDA-capable or AMD ROCm-capable hardware
• Keep macOS security intact — no need to disable System Integrity Protection (SIP)
• Use existing eGPU enclosures — any Thunderbolt/USB4 enclosure with sufficient power delivery

Tiny Corp, led by George Hotz (who famously jailbroke the first iPhone and hacked the PlayStation 3), has been working on the tinygrad deep learning framework — a lightweight alternative to PyTorch and TensorFlow. The eGPU driver is a natural extension of that work, enabling tinygrad users to leverage external GPU hardware on Macs.

What It Doesn’t Do

To be clear: this is not a gaming solution. The driver provides zero graphics acceleration:

• No video output to external monitors through the eGPU
• No macOS graphics acceleration (Metal, OpenGL, etc.)
• No gaming performance improvements
• Compute-only — designed for AI/ML workloads via tinygrad

If you’re hoping to plug in an RTX 4090 and play games on your MacBook, this isn’t it. But if you’re an AI researcher who loves macOS but needs more GPU compute, this is exactly what you’ve been waiting for.

The Bigger Picture: Apple and AI

Apple’s decision to sign this driver is notable because the company has historically been hostile toward third-party GPU support. When Apple transitioned to Apple Silicon in 2020, eGPU support was dropped entirely. Nvidia drivers haven’t been available on macOS since 2019.

By signing TinyGPU, Apple is acknowledging that its own GPU hardware isn’t sufficient for serious AI compute workloads — at least not yet. It’s a pragmatic move that keeps AI researchers on macOS rather than forcing them to switch to Linux or Windows for GPU-heavy work.

Whether this leads to broader GPU support on macOS remains to be seen. But for now, it’s a crack in Apple’s walled garden that the AI community will happily exploit.