Modular open-sources the full Mojo compiler under Apache 2.0 at ModCon 2026

Modular open-sourced the full Mojo compiler and tooling under the Apache 2.0 license at its ModCon 2026 conference on Aug. 18, days after the Mojo systems programming language reached its 1.0 milestone on Aug. 12.

The release, detailed in Modular’s announcement, opens the last closed piece of a language Modular has pitched as a faster, Python-compatible way to write AI infrastructure code. Mojo’s standard library has accepted outside contributions since 2024; the compiler and tooling source is now published in the company’s GitHub repository, with broader contribution access expected by year-end.

It was Modular’s first conference since its acquisition by Qualcomm, and the timing signals where the chipmaker wants Mojo to run. In a set of ModCon announcements, Modular said its Modular Cloud reached general availability serving billions of tokens per minute, and that it expanded hardware support to AWS Trainium, Google TPUs and Qualcomm’s Cloud AI 100 and Dragonfly chips, alongside existing Nvidia and AMD support.

The pitch is portability — one toolchain across rival accelerators — which puts Mojo in direct contrast with Nvidia’s CUDA lock-in. Logan Iyer, Microsoft’s corporate vice president for Windows Platform and Developer, said Microsoft is bringing Mojo to Windows; d-Matrix CEO Sid Sheth also spoke in support.

Open-sourcing a compiler is easier to announce than to sustain: an active outside contributor base, not a license change, is what determines whether Mojo becomes a genuine community project. For now the governance still runs through Modular and its new owner, Qualcomm.

Whether third-party contributions to the compiler actually open by year-end will be the first real test of how open the platform becomes.