Microsoft and Amazon AWS announce Gluon, a new open source deep learning interface to simplify deep learning models
New Open Source Interface Simplifies Deep Learning Development — At Microsoft we have a vision and passion to amplify human ingenuity with the transformative power of AI.
Context & Ripple Effects
Gluon lands after Amazon had already tested the open-source playbook with its DSSTNE deep learning library, pitched as faster than Google's TensorFlow on specific AWS configurations. The difference this time is that Microsoft — whose Cognitive Toolkit backs the announcement — joins as co-author rather than target, putting a shared high-level API above rival frameworks.
The move reads as a truce at the framework layer between two cloud competitors who fight everywhere else. Within weeks of the announcement, AWS doubled down by releasing ONNX-MXNet, extending its support to the Open Neural Network Exchange format that Microsoft introduced with Facebook.
First-order effects
- Developers working in MXNet or Microsoft's Cognitive Toolkit get one simplified interface instead of learning each framework's API, lowering the cost of starting deep learning projects on either company's stack.
- Google's TensorFlow, which neither company adopted here, now competes against a two-cloud coalition offering an easier on-ramp to its closest alternatives.
Second-order effects
- AWS's follow-on ONNX-MXNet release signals that the interface play extends to model portability, pressuring framework vendors to interoperate or risk losing developers who can now move trained models between stacks.
- Microsoft's later acquisition of Lobe, a no-code visual model builder, shows the same simplification agenda pushed further down the skill ladder — from easier APIs to no code at all.
Third-order effects
- If rivals keep cooperating on interfaces while competing on infrastructure, the durable battleground becomes which cloud hosts and serves the models — a pattern visible decades later when an expanded Amazon–OpenAI deal brought OpenAI's models to AWS even as OpenAI revised its Microsoft partnership.
- Open interfaces like Gluon and ONNX push the industry toward a split where the model layer is commoditized and interchangeable, shifting pricing power from framework owners to whoever runs the compute underneath.
The trend: Cloud rivals are standardizing the model-development layer through open source while reserving their competition for infrastructure and hosted models.