LG Group Chairman Koo Kwang-mo, right, and Microsoft CEO Satya Nadella poses for a photo at a top management meeting held at Microsoft's headquarters in Redmond, Washington, on Sept. 21.LG
LG is deepening its artificial intelligence partnerships with Microsoft and Nvidia, expanding into AI data center cooling, AI infrastructure and physical AI.
LG and Microsoft signed a partnership agreement Monday at a top management meeting held at Microsoft's headquarters in Redmond, Washington. LG Group Chairman Koo Kwang-mo and Microsoft CEO Satya Nadella attended, along with CEOs from LG affiliates including LG Electronics, LG U+ and LG CNS.
The two companies discussed cooperation in three areas: AI data center infrastructure, AI transformation (AX) and physical AI.
In infrastructure, LG will supply cooling, power and information technology infrastructure solutions as Microsoft expands its data centers worldwide in the near term. Longer term, the two plan to expand cooperation into prefabricated data centers built off-site in modules for faster construction.
In AX, LG will roll out Microsoft's AI and cloud platforms to its entire white-collar work force, including Microsoft 365 E7: Frontier Suite and the Copilot AI assistant.
In physical AI, LG plans to use Microsoft's cloud computing power to speed up AI model development for physical operations such as manufacturing, logistics and mobility. The two companies will also build a “data flywheel” that feeds factory floor data back into AI models to keep improving them.
The partnership builds on talks the two companies began at a tech showcase held at Microsoft's headquarters last December.
LG Electronics said the same day that it had received Nvidia certification for a large-capacity liquid cooling device called a coolant distribution unit (CDU). The device keeps massive server clusters in AI data centers from overheating. The certification qualifies the CDU for use in what Nvidia calls AI factories, which are large-scale data centers built specifically to train and run AI models at industrial scale.
LG's large-capacity liquid cooling devices known as coolant distribution unit (CDU) that are approved by Nvidia to be deployed on the U.S. company's DSX AI factory platformLG
LG Electronics' 2.5-megawatt CDU was approved as ready for deployment on Nvidia's DSX AI factory platform. The platform integrates the design and operation of AI factories across computing, networking, power, cooling, facilities and software. LG Electronics previously received Nvidia certification for a 600-kilowatt CDU in July and a 1-megawatt CDU in September.
The larger CDU matters because AI chips are drawing far more power than before. Nvidia's newest AI accelerators have driven exponential growth in chip performance. That has pushed the power draw of a single server rack from hundreds of kilowatts to the megawatt range. As a result, it has become more efficient to cool multiple racks together in groups rather than installing a smaller CDU at each individual rack.
Cooling a group of about 20 server racks, each with a capacity of about 120 kilowatts, requires a CDU in the 2.4-to-2.5-megawatt range.
Market researcher TrendForce has forecast that adoption of liquid cooling in AI data centers will rise from 14 percent in 2024 to about 60 percent by 2027.
LG Energy Solution, LG's battery arm, was also certified by Nvidia the same day. The certification qualifies its battery energy storage system (BESS) for use in Nvidia's DSX AI factory ecosystem. A BESS stores electricity in large battery packs and releases it quickly when needed, helping stabilize power supply to data centers.
LG Energy Solution's lithium iron phosphate battery product, called JF2 AC Link, met Nvidia's requirements with two features. AI load smoothing charges and discharges the battery to absorb sudden swings in AI server power use. An enhanced voltage ride-through function keeps power flowing steadily even when grid voltage drops or spikes sharply.
LG Energy Solution's lithium iron phosphate battery product known as JF2 AC LinkLG ENERGY SOLUTION
The battery unit plans to use the Nvidia certification to expand its North American data center energy storage business.
In May, the company signed a contract to supply 6 gigawatt-hours of energy storage system batteries to DTE Energy, a U.S. utility that runs large-scale AI data center projects for companies including Oracle. In July, it announced it would supply batteries for a solar and energy storage project run by Google and Cypress Creek Energy.
According to Bloomberg New Energy Finance, U.S. data center power consumption is projected to more than double to about 391 terawatt-hours by 2030, from 180 terawatt-hours last year.
