Newsletter Subscribe
Enter your email address below and subscribe to our newsletter
[forminator_form id="25163"]

phoronix+1opensourceforuservethehomeQualcomm has open-sourced Modular's entire AI software platform and the Mojo programming language at ModCon 2026, the first conference held by Modular as a Qualcomm company. The announcement, made on August 18 in San Francisco, places the full Mojo compiler and the broader Modular stack under the Apache 2.0 license with LLVM exceptions.phoronix+1
The move marks a turning point for a platform designed to challenge NVIDIA CUDA's dominance in AI software infrastructure. Modular's stack enables AI inference to run across hardware from Qualcomm Snapdragon chips, AMD Advanced Micro Devices, Inc. GPUs, NVIDIA GPUs, Apple silicon, AWS Trainium, and Google TPUs — all under a single framework.opensourceforu
Qualcomm completed its approximately $3.9 billion acquisition of Modular in late July, weeks after announcing the deal on June 24. The acquisition was designed to strengthen Qualcomm's AI software capabilities across devices, edge infrastructure, and data centers.prnewswire+2
Rather than locking down the technology, Qualcomm chose the open-source path. The code is now available on GitHub, a decision that ServeTheHome called notable "given that Qualcomm spent billions on Modular".servethehome
AMD's presence at ModCon was a highlight. The company appeared on stage to signal its support for the Modular platform, a move that ServeTheHome described as notable given that AMD has substantial AI GPU revenue and might otherwise view Qualcomm as a competitor. The platform already supports AMD Ryzen AI Halo systems and AMD GPUs at scale.servethehome
Modular has also demonstrated the platform running on a Chinese Houmo 24GB M.2 AI accelerator, illustrating the breadth of hardware the framework can target.servethehome
The Mojo programming language, created by Chris Lattner of LLVM and Swift fame, reached its 1.0 milestone on August 12. At ModCon, the team followed through on long-standing promises to open-source the full compiler. The language is designed for diverse hardware — from CPUs and GPUs to AI accelerators — and supports interoperability with Python. Modular is also working to bring Mojo to Windows alongside existing macOS and Linux support.theregister+1
GV's managing partner Dave Munichiello described Modular as enabling "any AI model to run on any chip," calling it an alternative to CUDA and a key example of opportunity in the AI application layer.startuphub