How to Pack 100 MW of Power into a 20 MW Data Center - Seguridad Mania.com - España y América Latina

The GenAI LLM boom and their GPUs are consuming so much power and cooling that data centers are literally running out of power. This is a tough problem to solve. Bringing in additional power is generally extremely difficult, costly, and more often than not, unfeasible. Compounding this power shortage problem are the AI datasets. LLM training requires petabytes to exabytes of stored data. Organizations are keeping all of their data for the foreseeable future. Many are keeping their data on storage systems with high density HDDs or SSDs. That storage also consumes massive amounts of power and cooling resources. Power and cooling that’s in very short supply. It leads to several very distasteful and costly choices: • Build a new data center with a lot more power and cooling. o It may require cutting very expensive deals with the power company to even get that much additional power. • Move the GenAI LLMs to a public cloud. o Very high costs o Where the data is held hostage with egress fees • Spread out over multiple co-locations. o Again, high costs o Difficult management o Much more difficult security o Slower performance There has to be a better way. There is. In this webinar, we’ll explore how to reclaim massive amounts of power and cooling from storage holding all that data. You’ll learn how Quantum’s ActiveScale Cold Storage delivers a breakthrough approach: • Up to 99% lower power draw than object or file HDD-based systems • Even much lower power draw than storage based on SSDs. • And still lower CapEx and OpEx than either • Extremely low-cost long-term storage—without the egregious cloud egress fees • While keeping that data readily accessible for LLM training or AI-machine learning Join us to see how you can radically reduce power consumption, reclaim budget, and build an AI-ready data foundation that scales for decades.