On November 27, 2025, Lenovo (00992) officially announced the expansion of its Neptune liquid cooling technology to its entire data center product portfolio, addressing the growing computational demands and energy consumption pressures of modern data centers. Coincidentally, on the same day, the Chicago Mercantile Exchange (CME) experienced a system failure due to cooling issues, leading to a 10-hour trading halt in global derivatives markets. This incident highlights the mounting strain on traditional data center cooling systems amid the AI computing boom, with soaring power consumption pushing global digital infrastructure to the brink of "thermal runaway."
**The Cost of Cooling Failures** The CME outage was triggered when a data center in Aurora, Illinois, activated its high-temperature protection mechanism, forcing server clusters to shut down automatically. According to CME’s statement, the root cause was a mechanical failure in a chiller unit, causing temperatures to surge from 24°C to 48°C within 90 minutes—far exceeding safe thresholds for electronic components. The disruption impacted trillions of dollars in contracts: Asian trading sessions were halted, European traders shifted to less liquid over-the-counter markets, and S&P 500 futures saw volatility spike to 2.3% upon resumption, marking the largest intraday swing in 2025.
This incident underscores the fragility of financial market infrastructure and reflects a broader "thermal runaway" crisis in data centers worldwide.
**The Imbalance Between Computing Power and Cooling** With the rise of generative AI and large-scale model training, server power consumption is skyrocketing. Nvidia’s Blackwell GPU architecture now exceeds 1,000W per chip—triple the A100’s consumption five years ago—while standard server racks now demand 600kW, equivalent to a small factory’s energy use.
The core issue lies in cooling technology lagging behind computing advancements. Air cooling, with thermal conductivity just 1/20th that of water, struggles to handle chips beyond 350W, while modern AI chips have already surpassed this limit.
According to the International Energy Agency (IEA), data centers account for 1.8% of global electricity consumption, with 38% of that power dedicated to cooling—meaning nearly 0.4 units of energy are wasted for every 1 unit used for computation. Microsoft CEO Satya Nadella recently noted that while AI chip demand remains high, data centers face power and space constraints, leaving many chips idle in inventory.
**Lenovo’s Liquid Cooling Push** As air cooling nears its physical limits, liquid cooling is emerging as the industry’s next frontier. Lenovo’s Neptune technology, now extended across its data center lineup, reduces energy use by 40% compared to air cooling, achieving a power usage effectiveness (PUE) as low as 1.1—far outperforming air-cooled systems averaging 1.5+.
Neptune’s direct-to-chip cooling (DWC) uses 45°C water to absorb heat at the source, eliminating reliance on energy-intensive chillers. Kumar Mitra, Lenovo’s ISG APAC Executive Director, cited DreamWorks Animation’s 20% performance boost and 35% water savings, with adopters like the Korea Meteorological Administration leveraging the tech for high-density tasks like weather modeling.
The CME outage has accelerated industry shifts toward liquid cooling. China’s telecom giants plan to deploy liquid cooling in over 50% of new data centers, while AWS and Microsoft Azure aim for 80% adoption in AI servers by 2027.
China’s liquid cooling server market is projected to hit $3.39 billion in 2025, up 42.6% year-on-year, with a 48% CAGR through 2029. Goldman Sachs warns that by 2030, data center power demand could surge 165%, threatening tech giants’ profitability unless energy efficiency improves.
In this era, high-efficiency data center solutions are no longer optional—they’re imperative for sustainable growth.
Comments