tech
Google tells employees it must double capacity every 6 months to meet AI demand
Google’s AI infrastructure chief tells staff it needs thousandfold capacity increase in 5 years.
TL;DR
- Google needs to double its AI serving capacity every six months to meet demand.
- Google aims for a thousandfold increase in AI infrastructure capacity within 4-5 years, while maintaining cost and energy efficiency.
- OpenAI is partnering to build six massive data centers, committing over $400 billion to reach nearly 7 gigawatts of capacity.
- Both companies face challenges serving their AI user bases, with OpenAI's paid users hitting usage limits.
- Nvidia's limited GPU production capacity is a major bottleneck for AI infrastructure development.
- Google plans to achieve scaling targets through physical infrastructure, efficient AI models, and custom silicon chips, including its new Ironwood TPU.
- Google's aggressive expansion reflects a calculation that the risk of underinvesting in AI infrastructure outweighs the risk of overcapacity.
- Google CEO Sundar Pichai acknowledges AI bubble talk but stresses the intense competition and demand for cloud and compute.