Modular Open-Sources Mojo 1.0 to Break AI Hardware Lock-in
The company expands its platform to support Google TPUs and AWS Trainium while partnering with Microsoft for native Windows support.
Modular has fully open-sourced the Mojo 1.0 programming language, including its compiler and all associated tooling, under the Apache 2.0 license. Announced at ModCon 2026, the move aims to establish a vendor-neutral software foundation for the AI industry, reducing reliance on proprietary stacks.
Alongside the open-source transition, Modular launched Modular Cloud for production AI inference, which is already serving customers including MiniMax. The company also significantly expanded the Modular Platform's hardware compatibility, adding support for Google TPUs, AWS Trainium, and Qualcomm’s Cloud AI 100 and Dragonfly accelerators. To further broaden its reach, Modular is collaborating with the Microsoft Windows team to develop native Windows support for the language.
The Push for Heterogeneous Computing
Modular was founded on the premise that the current AI software stack is insufficient for the era of heterogeneous hardware. While Python offers high usability, it often lacks the systems-level performance required to maximize the efficiency of CPUs, GPUs, and custom accelerators. Mojo was designed specifically to bridge this gap, combining Python-like syntax with the performance of a systems language.
This announcement is the culmination of a phased opening of the ecosystem. It follows the 2024 release of the Mojo standard library and the 2025 opening of MAX kernels. Additionally, the company has updated the MAX license to remove device usage restrictions, noting that it will become source-available through an open alliance program.
Breaking the Hardware Monopoly
By open-sourcing the compiler and tooling, Modular is attempting to prevent a future where AI developers are locked into a single hardware vendor's proprietary stack. The ability to deploy across diverse accelerators—from Google's TPUs to Qualcomm's silicon—allows enterprises to move workloads based on performance and cost rather than software constraints.
Native Windows support is a critical piece of this strategy for enterprise adoption. "We're excited to see Mojo coming to Windows and the opportunities it creates for developers working across systems and AI," said Logan Iyer, CVP of Windows Platform + Developer at Microsoft.
The Path to AI Portability
Industry leaders view this shift toward open standards as essential for the growth of specialized AI silicon. Sid Sheth, Founder and CEO of d-Matrix, stated that the shared commitment between d-Matrix and Modular to heterogeneous computing is underpinned by mutual support of open standards.
As Mojo moves into the Windows ecosystem and expands its cloud footprint, the industry will be watching to see if the language can successfully displace the fragmented mix of C++ and Python currently dominating AI infrastructure. The primary remaining hurdle is the full rollout of the native Windows integration and the adoption rate of the new open-source compiler among enterprise developers.