⚠ Current #grids won’t be able to support the explosive growth of data center energy consumption. 🔍 Investors are looking for companies creating energy-efficient AI solutions, including smarter #AI models, chips, hardware, and software.
🔑 A 2021 study conducted by Google and the University of California, Berkeley showed the company was able to drastically decrease carbon emissions of its deep learning Transformer model in just four years.
📍 Enhancing computing chip utilization rate to nearly 100% by granularizing GPUs, pooling, orchestrating, scheduling, and optimizing the utilization of all computing hardware resources across in-cloud, on-premise, and on-devices can achieve significant performance enhancement and energy saving, this is a practical must-have infrastructure now.
🖍 Building small-footprint modular data centers co-located with remote renewables projects to take advantage of excess clean energy and focusing on compute-intensive applications like AI training only.
📞 Renewable energy marketers use AI to optimize renewable operations and pricing along with a new business and financing model, which can help decarbonize data centers. 💡 Redesigning the Data Center With renewable energies and battery-powered microgrids. At least, building renewable energies with cost-effective energy storage to improve peak reliability.
📢 Solutions that have the potential to bolster resource adequacy in the short term fall into five buckets: (Reference: MEETING GROWING ELECTRICITY DEMAND WITHOUT GAS) => read more in the article.
💸 Integrating new energy sources such as geothermal or nuclear energies still has lots of challenges, but has already gotten much attention and progress.
💲 #Cooling accounts for 40% of a data center’s energy consumption. The #liquidcooling market is expected to approach $3 - 5B by 2028 driven by AI data center builds that will bring significantly higher rack power densities.
🛡 Coolant distribution units (CDUs) are the most valuable components in liquid cooling solutions. Some manufacturers are expanding capacity at double-digit times.
⚛ New materials such as synthetic diamonds and ultrapure glass may make the coolest, fastest chips.
⏩ New circuit architectures, custom #chips, or different fundamental physics underlying how hardware/chips work can leapfrog the current PPAC (Power, Performance, Area, Cost) constraints. (e.g. light replaces electrons - #Photonics #SiPhotonics #CPO)
More info here: https://lnkd.in/gdcbB6S8
PLEASE ADD IDEAS OR #STARTUPS SOLVING THIS CHALLENGE BELOW.
Hossein Fariborzi Jack Whelan Jack Yen Zohaib Dar Michail Kosak Abhishek Sastri Clara K. McLain Faett JD LAU Eigen Fu
#AIdatacenter #ClimateTech #CleanEnergy #Infrastructure
