--- title: "AI computing power \"lifeline\" faces El Niño \"stress test\" as global data centers confront extreme high temperature challenges" type: "News" locale: "en" url: "https://longbridge.com/en/news/291104670.md" description: "The El Niño in 2026 will trigger extreme high temperatures in the Northern Hemisphere, posing a severe stress test for global AI data centers. Data shows that 79% of data center capacity faces high risks from climate disasters such as floods and wildfires, with severe weather becoming a major source of losses for the insurance industry. Climate risk is evolving from background noise to a core variable threatening the stable operation of AI infrastructure, forcing tech giants to confront operational and cost challenges" datetime: "2026-06-29T07:12:01.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/291104670.md) - [en](https://longbridge.com/en/news/291104670.md) - [zh-HK](https://longbridge.com/zh-HK/news/291104670.md) --- # AI computing power "lifeline" faces El Niño "stress test" as global data centers confront extreme high temperature challenges According to the Zhitong Finance APP, as European citizens try to escape the record-breaking heat wave, global AI infrastructure is facing a severe "stress test." In 2026, the rare high temperatures brought by the El Niño phenomenon will sweep across the Northern Hemisphere, and whether high-performance chips in AI data centers can continue to operate stably under extreme weather has become a core challenge that technology giants and infrastructure operators must confront. Data from the National Oceanic and Atmospheric Administration (NOAA) indicates that in 2026, the average spring temperature in the contiguous United States reached 13.22°C, the second highest spring temperature in 132 years of recorded meteorological history. Starting in May, the equatorial central and eastern Pacific officially entered an El Niño state, which is expected to develop into a moderate to strong El Niño event during the summer and autumn, affecting major global economies. The intersection of this extreme weather and the explosive growth of AI computing power is exposing a long-ignored systemic risk—climate risk is evolving from "background noise" to a core variable threatening AI infrastructure. ## Climate Risk Becomes the "Number One Threat" to the Insurance Industry: 79% of Data Center Capacity Exposed to High Disaster Risk Patrick McBride, head of international construction business at Zurich Insurance, revealed a key statistic in an interview: over the past three years, severe weather has become the primary cause of losses in the company's U.S. data center construction risk underwriting, currently accounting for one-third of its total losses. "Severe weather is no longer an issue that can be treated as a background risk," McBride stated. "It is one of the primary concerns for us and the owners we work with." A recent study by climate risk analysis firm First Street provides macro data support for this trend. The research shows that 79% of global data center capacity faces increased risks from acute climate disasters such as floods, extreme winds, and wildfires, which can disrupt operations, increase downtime, and raise insurance and maintenance costs. The study also found that more than half of global data centers are located in areas that have long suffered from persistent climate pressures such as extreme heat and drought. In terms of regional distribution, 89% of data center computing assets in the Asia-Pacific region are in high climate risk areas, the highest globally; the Americas have a ratio of 50%, while Europe, Africa, and the Middle East have 46%. Modeling analysis by Swiss Re further reveals the concentrated risks of U.S. data centers: over a quarter of data center capacity is located in areas that experience severe hail at least three days a year, and about 40% of the capacity is in tornado-prone areas. **Insurance Industry Under Pressure: Data Center-Related Premiums May Double to $24.2 Billion by 2030** The impact of climate risk on the data center industry is profoundly changing the underwriting logic of the insurance industry. The insurance demand for a single hyperscale data center has surged from the traditional approximately $2 billion to $20 billion per campus. Swiss Re warns that the reinsurance industry can only support a small portion of this limit at competitive rates. S&P Global Ratings estimates that the insurable value of the largest single project can reach up to $30 billion In practice, achieving a single-site insurance coverage of $10 billion means stacking more than 40 insurance companies. Even Aon expanded its data center lifecycle plan to $3.5 billion in April, and Willis could only gather about $3 billion in ultra-large-scale construction underwriting capacity. "You simply can't find enough underwriting capacity to cover $20 billion concentrated in one location," said Patricia Kwan, a senior credit analyst at S&P Global Ratings. "Insurance companies are not designed for this level of concentration." Michael LaRocca from Swiss Re predicted at this year's RIMS Riskworld conference that premiums related to data centers will rise from the current approximately $10.6 billion to nearly $24.2 billion by 2030. Aon warned in its 2026 Global Construction Insurance and Surety Market Report that clients must layer climate and disaster data on every construction project. The report issued a brief but powerful warning to anyone attempting to underestimate weather factors: "Never bet against the climate." ## Power Grid "Double Squeeze": AI Power Consumption and Air Conditioning Compete Under High Temperatures What makes 2026 special is the exponential growth in AI computing power demand coinciding with the extreme heat brought by El Niño. According to the International Energy Agency (IEA), from 2023 to 2025, the electricity consumption of AI data centers will remain constant at about 1.8% of global electricity generation. However, by 2026, this proportion will jump to between 3.7% and 4%. Goldman Sachs' forecast is even more specific: the electricity demand of U.S. data centers will rise from 31 gigawatts in 2025 to 41 gigawatts in 2026, and further increase to 66 gigawatts in 2027, nearly doubling. Under normal temperatures, cooling systems account for about 40% of energy consumption in data centers; this proportion will further increase under extreme heat—just as air conditioning usage also raises grid load. "Data centers need energy the most when the grid is least able to provide it," pointed out Mishal Thadani, co-founder and CEO of the AI software platform Rhizome. He cited Turin, Italy, as an example: the city experienced temperatures of 38 degrees Celsius in May, and the heatwave caused underground cables to suffer thermal stress, leading to repeated power outages. "Now add in the power consumption of each facility, which is equivalent to that of 100,000 households," Thadani said, noting that high temperatures and high loads simultaneously affecting the same transmission lines amplify the risk exponentially. On May 18 this year, the U.S. Department of Energy (DOE) officially issued an emergency order authorizing the largest regional grid operator, PJM Interconnection, to call on data centers and large commercial and industrial users' backup generators in extreme emergencies to alleviate grid pressure. This measure is essentially a prediction of the high temperatures that may occur in concentrated data center areas this year. In June, Loudoun County, Virginia, experienced daytime temperatures soaring to 36°C to 38°C due to the overlapping effects of a heatwave on the East Coast, while high humidity weakened the efficiency of the evaporative cooling systems widely relied upon by data centers. Meanwhile, in the Dallas-Austin corridor in Texas, nighttime lows remained at 23°C to 24°C, making the window for "natural cooling" reliant on outside cool air nearly disappear, forcing cooling compressors to operate continuously throughout the day ## Industry Self-Rescue: Liquid Cooling Technology as a Key Breakthrough In the face of operational challenges brought about by extreme weather, technology giants and infrastructure operators are shifting from passive responses to proactive design changes. NVIDIA (NVDA.US) recently detailed the 45°C full liquid cooling technology used in its next-generation Rubin platform, describing it as "one of the most significant energy efficiency breakthroughs in the history of data centers." The core of this solution is that the cooling liquid can operate at temperatures up to 45 degrees Celsius, rising to about 55 degrees Celsius after flowing through cold plates directly attached to the chips, and then dissipating heat into the atmosphere through outdoor dry coolers. This design not only reduces energy consumption but also nearly eliminates the need for water. Josh Parker, NVIDIA's head of sustainability, stated that the reference design could reduce water usage from approximately 0.6 gallons per kilowatt-hour in traditional cooling tower systems to nearly zero, achieving up to about 95% water reduction. In terms of energy consumption, industry estimates show that for every 1 degree Celsius increase in chilled water temperature, cooling energy costs can be reduced by about 4%. "We expect the 45-degree liquid cooling solution will bring data centers closer to 'operating without chillers,'" NVIDIA stated. "In many regions, chillers may only need to be activated during a few days of extreme heat each year." Hyperscale cloud service providers are also following suit. Microsoft stated that the design of its data centers can "reliably operate under a wide range of environmental conditions" through site selection, redundancy systems, and real-time monitoring. Aaron Lewis, Chief Commercial Officer of HVAC company Johnson Controls, revealed that he recently saw a European client include a "climate change factor" in the specifications for their data center for the first time, with facility designs based on projected temperature increases rather than historical averages. The president and CEO of Motivair, a high-end cooling subsidiary of Schneider Electric, previously stated, "Once the power consumption of a single chip reaches a certain level, liquid cooling is no longer optional but a necessity." ## Outlook: The "High-Temperature Pressure Test" of the AI Era Has Just Begun The summer of 2026 is seen as the first round of "stress testing" for the entire AI industry to validate its resilience. As the power consumption of single chips breaks the kilowatt level, the power demand and heat dissipation pressure of AI data centers will continue to rise. The construction cycle for global data centers (typically 2 to 3 years) is far faster than the supporting green electricity and grid infrastructure (5 to 7 years), and the power supply bottleneck issue may continue to ferment in the coming years. Goldman Sachs pointed out that about 60% of the new power demand driven by AI data centers will require new capacity, supplied by a combination of natural gas, solar, and wind energy, but permitting, transmission, and supply chain issues may delay the commissioning of natural gas plants for another 5 to 7 years. For investors, the impact of extreme weather on the data center industry is twofold: on one hand, insurance and reinsurance companies are facing severe challenges to their underwriting capacity, with the industry expected to see related premiums double to $24.2 billion by 2030; on the other hand, companies providing climate-adaptive infrastructure—from advanced cooling systems to weather risk modeling platforms—are likely to benefit, as operators are competing to reinforce their facilities in response to a threat that is no longer just theoretical As Joe Macejak, Head of U.S. Property Digital Infrastructure at Marsh Risk, stated, whether climate risk will affect the digital infrastructure revolution is no longer the question; the key lies in how the industry identifies, quantifies, and manages these risks within its tolerance levels. When AI's "Silicon River" encounters the "scorching sun" of extreme heat, this adaptive evolution concerning the lifeblood of computing power has only just begun ### Related Stocks - [EQIX.US](https://longbridge.com/en/quote/EQIX.US.md) - [IDGT.US](https://longbridge.com/en/quote/IDGT.US.md) - [AIQ.US](https://longbridge.com/en/quote/AIQ.US.md) - [BOTZ.US](https://longbridge.com/en/quote/BOTZ.US.md) - [DTCR.US](https://longbridge.com/en/quote/DTCR.US.md) - [IGPT.US](https://longbridge.com/en/quote/IGPT.US.md) - [CLOU.US](https://longbridge.com/en/quote/CLOU.US.md) - [ZURVY.US](https://longbridge.com/en/quote/ZURVY.US.md) - [SPGI.US](https://longbridge.com/en/quote/SPGI.US.md) - [AON.US](https://longbridge.com/en/quote/AON.US.md) - [WTW.US](https://longbridge.com/en/quote/WTW.US.md) - [GS.US](https://longbridge.com/en/quote/GS.US.md) - [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md) - [MSFT.US](https://longbridge.com/en/quote/MSFT.US.md) - [JCI.US](https://longbridge.com/en/quote/JCI.US.md) - [SBGSY.US](https://longbridge.com/en/quote/SBGSY.US.md) - [MMC.US](https://longbridge.com/en/quote/MMC.US.md) - [W4VR.SG](https://longbridge.com/en/quote/W4VR.SG.md) - [NVD.DE](https://longbridge.com/en/quote/NVD.DE.md) - [MRSH.US](https://longbridge.com/en/quote/MRSH.US.md) ## Related News & Research - [AIQ vs. BOTZ: Which AI ETF Could Deliver Bigger Returns in 2026?](https://longbridge.com/en/news/291727392.md) - [AIQ Secures New £2m Convertible Funding and Aligns Existing Notes to Support AI Infrastructure Push](https://longbridge.com/en/news/291763933.md) - [SK Telecom pursues construction of 15 GW AI data center](https://longbridge.com/en/news/291728658.md) - [Companies are buying AI tools. That doesn't mean they know what to do with them.](https://longbridge.com/en/news/291779553.md) - [F5 (FFIV) Could Be 13% Undervalued As Equinix Tie Up Tests Valuation](https://longbridge.com/en/news/292009571.md)