Microsoft joins AI cost-cutting trend by relying more on its own models
Microsoft is increasingly using its own AI models to reduce reliance on external providers like OpenAI, joining a broader industry trend of cutting costs by developing in-house AI technology.
Background
- Microsoft has historically relied on OpenAI for its AI services (e.g., Copilot, Azure OpenAI), but now it's shifting to use more of its own in-house AI models (like Phi and MAI) to cut costs and reduce dependency.
- This mirrors a broader industry trend: companies like Salesforce, Apple, and Meta are developing proprietary AI models instead of paying for third-party APIs, aiming for lower inference costs and tighter integration.
- Key context: Microsoft has invested over $13B in OpenAI and is its largest cloud provider. The move signals a strategic pivot as OpenAI's costs remain high and open-source alternatives improve.
- "Inference" is the process of running a trained AI model to generate answers; this is the main ongoing expense after a model is built.