AI factories: Turning data into intelligence at scale

Robert HK Demann, Executive VP and CEO, spoke about AI Factories: Turning Data into Intelligence at Scale. AI is also about the infrastructure. When we think of steel, we think of Jamshedpur. For oil refineries, we think of Jamnagar. AI factory is about combining everything in a meaningful manner. 

Siemens data and AI technology is already delivering strong business impact. It can boost performance by 30% for higher task-specific productivity, through Industrial Copilot. There can be 35% reduction in operational costs with AI-based technology for San Juan de Dios hospital in Spain. It can also accelerate development by 40%. There is up to 20% reduction in product time-to-market at Siemens' first digital-native factory in Jiangsu.

Finally, data and AI can also enhance resilience. There is up to 67% reduction in equipment breakdown with AI-driven predictive maintenance at Tata Steel, Netherlands. Up to 55% energy savings is possible with AI-powered cooling in data centers at Bank of Montreal.

India's data center growth trajectory is building global AI and data powerhouse. India has large UPI customer base, large customer base in healthcare, etc. India's capacity can grow from 1.6GW today to 8.3GW by 2030 as per McKinsey. 

AI computing and modelling are driving foundational infrastructure shifts. We are increasing in scale, from MW to GW. We are increasing 10x in density. We have moved from years to months in innovation cycles. We are seeing air to liquid cooling, and AC to DC power, with tech getting disrupted. 

All compute power goes into the rack. We are now moving from power needed from 2BHK to power needed for 30 floors. We need to get that right. Digital twins come into the game, as we need to get it right, if we need to build equipment. 

So, what is an AI factory? It has sustainability and regulatory requirements. We have energy, air-cooled facilities to take care of poor heat reuse, and cloud services dominate. We have liquid-cooled sustainable heat recovery. We have grid partitioning for sustainability. The AI services dominate, and there is AI-driven operations for efficiency. We have direct current power distribution for high density. Direct chip cooling is currently the norm in hyperscalers. 

We have the silicon roadmap that drives infrastructure. Hyperscalers and chip providers are racing to densify the rack power, and increase silicon power dissipation at the chip level. This includes the rack power roadmap by players, and silicon roadmap. The mix of processor types drives most of the data center design disruption.

AI factory of the future
We now come to the AI factory of the future. The building will be a mix of densities. Density drives the technology choice. We need to do this best and effectively. AI changes the data center environment that is density driven. At scale, AI infrastructure should be orchestrated and optimized, and not just managed.

AI is an evolution of the industry, and Siemens is ready. PUE was used to measure efficiency. In future, it will be measured in tokens. If you can save energy, you have an advantage. Collection of the ecosystem partners needs to be very successful.