Evolving critical power needs
The widespread deployment of artificial intelligence (AI) imposes new demands on critical power infrastructure. This shift requires innovative solutions to handle higher power density servers, spikey and dynamic AI power loads, and integrating liquid cooling systems with low thermal inertia. Traditional power sources are becoming constrained, requiring grid-independent alternatives to meet AI's energy demands. Vertiv addresses these evolving needs with resilient and flexible solutions designed to minimize the impact of single-point failures and support distributed power protection.
Impact of AI applications on critical infrastructure
With the forecasted growth in rack densities for AI and high-performance computing (HPC) applications, Vertiv is developing solutions designed to withstand higher voltages, manage spikey AI power loads, and provide power protection for liquid cooling systems. This includes increasing power converter block sizes and enhancing power distribution circuits to meet the demands of high-density racks in AI infrastructures.
AI power challenges and emerging solutions
- Power fluctuations: Occur as graphics processing unit (GPU) clusters may shift from idle to overload in milliseconds. Operators can explore:
- Rely on large uninterruptible power supply (UPS) overload capabilities designed to support such sags
- Leveraging control algorithms on UPS in two possible scenarios:
- The user wants to avoid hitting the electrical source or local generators, which aren’t sized to withstand large power swings.
- The user wants to avoid hitting the battery to prevent accelerating its aging due to small but frequent discharge/recharge cycles.
- Thermal cycling: Rapid temperature changes in components due to varying workloads. Operators can consider:
- Adopting centralized large UPS, resilient by design and tested to increase reliability in extreme conditions and corner cases.
- Adjusting fan speed control algorithms in UPS and other power components to manage thermal cycling effectively.
- Coolant distribution units (CDUs): Must adapt to sudden increases in cooling demand as AI workloads intensify. This involves:
- Facilitating real-time communication between power products and CDUs to meet rapid cooling demands.
- Power density requirements: Confined spaces demand more processing power, pushing traditional power protection technologies to their limits. Solutions include:
- Installing large UPS optimized for high power and integrations with other elements of powertrain.
- Adopting higher voltage distribution and medium voltage solutions to handle increasing power demands.
AI-ready UPS systems
Vertiv™ large UPS systems are designed to manage AI applications efficiently. Based on our performance tests, these units can handle rapid load changes without any performance degradation, allowing continuous operation. Additionally, they protect the power source from sudden step loads and support increased rack power density by operating at higher voltages. These capabilities make Vertiv large UPS systems ideal for the demanding requirements of AI applications, providing reliable and stable power to maintain optimal performance.
Learn more
Evaluating the performance of Vertiv™ large UPS systems with AI workloads
AI workloads need UPS systems that manage GPU load changes without power disruptions. Read our technical note to discover how Vertiv’s large UPSs maintain stable output voltage and current.
Future-ready data centers
Vertiv's approach focuses on to meet the growing demands of AI and high-density computing. By integrating advanced power management algorithms and supporting new energy sources, Vertiv enables data centers for AI and other HPC applications to handle the dynamic and intensive workloads of the future.
Choose your path to high-density
Your journey to high-density can be a smooth ride. Vertiv offers advanced solutions engineered with industry leaders to power and cool the most challenging deployments. Unleash your AI Evolution with Vertiv today.
