I’m trying to fully understand the licensing model for MAX and Mojo, especially around hardware support and scaling. From what I’ve read, it seems like everything is free for development and even commercial production when using CPUs and NVIDIA GPUs, with no scaling limits. However, for non-NVIDIA hardware like AMD GPUs or Apple Silicon, there appears to be a limitation of up to 8 devices in production before needing to contact you or move to an enterprise agreement.
My expectation is that a modern systems programming language and runtime like Mojo, along with MAX, should allow fully free usage across all supported hardware types—including heterogeneous compute setups (e.g., mixing NVIDIA, AMD, and Apple devices)—without artificial scaling restrictions, especially for commercial workloads.
Can you clarify:
- Whether I can freely use AMD and Apple GPUs in commercial production workloads without restrictions, beyond just the 8-device limit?
- Whether heterogeneous clusters (mixing multiple GPU vendors) are fully supported under the free license, and if any scaling limits apply in that case?
- What exactly happens if I exceed the 8-device limit on non-NVIDIA hardware—does functionality stop, or is this purely a licensing/compliance requirement?
- Whether there are any technical or performance-related reasons (as opposed to purely licensing/business reasons) behind the different treatment of NVIDIA vs. non-NVIDIA hardware?
Ideally, I’m looking for a setup where I can scale freely across any combination of hardware (of heterogeneous compute clusters) I own or have access to, without needing to engage in enterprise agreements just to expand compute capacity. Is that aligned with your roadmap or philosophy going forward?
From a developer’s perspective, investing deeply in a platform and programming language only feels worthwhile if it guarantees long-term freedom to build, scale, and innovate across any hardware without the risk of future licensing barriers limiting that growth.