News / Taiwan / cnyes
Microsoft Executive Urges AI Infrastructure Focus on 'Useful Yield' at SEMICON
Technology Services · Internet Software/Services · cnyes · 2026-09-02
Microsoft's Rani Borkar emphasizes that AI infrastructure success depends on maximizing 'useful yield' through cross-layer design and system efficiency.
What Happened
Redefining AI Infrastructure: Microsoft Azure's Rani Borkar argued at SEMICON that the success of AI infrastructure should not be measured solely by raw computing capacity. Instead, the industry must prioritize 'useful yield,' focusing on how efficiently resources are converted into valuable AI outcomes.
Maximizing Efficiency: As AI demands grow, simply expanding hardware is no longer sufficient. Microsoft advocates for optimizing the entire cloud stack to increase the number of tokens generated per dollar and per watt, while simultaneously reducing latency.
Cross-Layer Integration: Borkar highlighted the importance of a 'silicon-to-systems' approach, which integrates hardware, software, power, and thermal management. This cross-layer design is essential for overcoming current bottlenecks and achieving breakthrough innovations in AI performance.
Collaborative Ecosystems: Microsoft remains committed to working closely with its ecosystem partners to improve cost-effectiveness and scalability. By fostering deeper integration across the stack, the company aims to make AI infrastructure more efficient and accessible for diverse workloads.