Source
The yield imperative: Rethinking success in AI infrastructure
At SEMICON Taiwan 2026, Microsoft's Rani Borkar urged the AI industry to prioritize "useful yield," which measures how efficiently infrastructure resources are converted into useful intelligence rather than focusing solely on raw capacity.
Why it matters
Increasing the efficiency of AI infrastructure can reduce costs and power consumption, making artificial intelligence more affordable and accessible to a broader range of people and organizations.
The details
Useful yield is measured by gains such as higher throughput, lower latency, and more tokens per dollar and watt. Microsoft's approach includes using a two-tier scale-up network topology in Azure Maia and cross-layer design that integrates silicon, cooling, and fleet operations. Additionally, the rise of agentic AI is creating new demands for computing infrastructure.
Get the weekly recap
The stories like this one, picked and explained — once a week, straight to your inbox.