Apple approves AMD and Nvidia eGPU drivers, enabling AI‑focused external graphics on Mac

Apple’s newly approved eGPU drivers mark the first time macOS will natively support external AMD and Nvidia graphics cards for artificial‑intelligence workloads, a development that could reshape the economics of AI research on Apple‑silicon machines. According to Tom’s Hardware, the drivers were signed by Apple after a series‑of tests that began in May 2025, eliminating the need for users to disable System Integrity Protection or employ other work‑arounds to get external GPUs (eGPUs) running on Macs. The company behind the driver, Tiny Corp, announced on X that installation “is so easy a Qwen could do it,” implying a plug‑and‑play experience that previously required deep system tweaks.

The practical impact of the driver approval is immediate for a niche but growing market of AI‑focused Mac users. Tiny Corp’s “tinybox” accelerator, which bundles four high‑end GPUs, has already been marketed to professionals who need on‑premise LLM inference or modest training capacity. The red v2 model, powered by four AMD 9070XTs, sells for $12,000, while the green v2 Blackwell, equipped with four RTX Pro 6000 Blackwell GPUs, commands $65,000 (Tom’s Hardware). With Apple’s driver now signed, owners of Thunderbolt or USB‑4 eGPUs can attach these external enclosures to MacBook Pros, Mac Studios or the newer Mac mini models and run AI software without the previous “unsupported hardware” warnings. The move is unlikely to revive Mac gaming—Tiny Corp’s driver is explicitly designed for LLM processing rather than graphics rendering—but it does give developers a legitimate path to leverage the massive unified memory and high‑bandwidth interconnects of Apple silicon for AI inference.

From a market‑share perspective, the driver approval could accelerate the adoption of Apple hardware in AI‑centric enterprises that have traditionally gravitated toward Linux‑based workstations. The recent surge in demand for Macs equipped with large unified‑memory configurations—driven by AI agents such as OpenClaw—has already strained supply chains, pushing delivery windows from six days to six weeks and prompting Apple to discontinue the 512 GB unified‑memory option on the Mac Studio (Tom’s Hardware). By enabling external GPUs, Apple may alleviate some of that pressure: users who need additional compute can augment a lower‑spec Mac with an eGPU rather than purchasing a top‑tier, high‑memory Apple system. This could broaden the appeal of mid‑range Macs for AI teams that are cost‑conscious but still require GPU acceleration.

The strategic significance extends beyond the immediate hardware ecosystem. Tiny Corp’s driver was not supplied by AMD or Nvidia; the company developed it independently, according to Tom’s Hardware. This suggests a potential shift in how GPU vendors approach Apple’s closed platform: rather than waiting for Apple to certify their own drivers, third‑party firms may take the initiative to bridge the gap, especially when the target use case—AI inference—is distinct from the traditional gaming focus that Apple has historically deprioritized. The fact that AMD’s CEO Dr. Lisa Su reportedly intervened to resolve earlier driver conflicts underscores the commercial stakes for GPU makers in gaining access to the Mac market (Tom’s Hardware).

Nevertheless, the approval comes with limitations. The drivers are optimized for AI workloads, not for the high frame‑rate, low‑latency demands of gaming, and the external GPUs will still be constrained by the bandwidth of Thunderbolt/USB‑4. Moreover, the cost of high‑end eGPU enclosures—$12,000 for the AMD‑based tinybox and $65,000 for the Nvidia‑based version—places them out of reach for most hobbyists. For enterprise users, the solution may still fall short of the performance of dedicated AI clusters or cloud services, but it does provide a viable on‑premise alternative for inference tasks that require data‑privacy or low‑latency edge processing.

In sum, Apple’s driver signing represents a modest but meaningful expansion of macOS’s compute capabilities, aligning the platform with the burgeoning AI workload demand that has already reshaped its hardware sales. By removing technical barriers to eGPU adoption, Apple is likely to see incremental growth in AI‑oriented Mac deployments, while GPU vendors may be prompted to pursue similar collaborations in the future. The development does not herald a revival of Mac gaming, but it does signal Apple’s willingness to accommodate specialized external hardware when the market narrative shifts from entertainment to enterprise AI.