Lenovo Group (992) investor relations material

Key challenges in scaling AI infrastructure

• Constraints have shifted from compute to a combination of land, power, shell, and deployment, requiring comprehensive execution and coordination across all elements.

• Speed and efficiency in deploying infrastructure are critical, with metrics like tokens per dollar and tokens per watt guiding performance and ROI.

• Delays in any phase of planning, manufacturing, or deployment can significantly impact investment returns.

• Access to flexible capital and early planning are essential for rapid infrastructure activation and business outcomes.

Optimization and operational excellence

• Optimization must begin at the initial design phase, not after deployment, to ensure long-term efficiency.

• Early engagement includes evaluating site readiness, power, thermal footprint, and operational targets.

• Continuous optimization and maintenance are required post-deployment as workloads and business needs evolve.

• Repeatable architecture and holistic planning for power, cooling, and lifecycle management are vital for scalable operations.

• Strong partnerships and integrated services reduce execution risk and support ongoing operational excellence.

Lessons and future directions

• Principles and innovations from gigawatt-scale AI factories will influence enterprises of all sizes.

• Liquid cooling and advanced facility design are becoming standard for high-density, efficient infrastructure.

• The focus is shifting from just technology to how environments are operated and optimized over time.

• Optimization is a guiding principle throughout the AI lifecycle, not just a final phase.