•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
AI fever and data centers are driving an unprecedented global race for infrastructure. But behind the expensive GPU arrays, language-model processing supercomputers, and billion-dollar AI ambitions, there is another “monster” quietly consuming electricity at a massive scale: cooling and air conditioning systems.
For many years, HVAC (heating, ventilation and air conditioning) was treated as an ancillary building system. For modern data centers, however, cooling is a life-or-death requirement that determines whether the operation can function at all.
Industry studies cited in the article say cooling systems can account for 40–60% of a data center’s total electricity consumption, depending on scale and server density. As AI increases computational density, heat generation rises sharply. GPU clusters used for AI can produce far more heat than traditional servers, requiring continuous cooling 24/7.
In practical terms, more AI can mean more electricity devoted to cooling.
The article says the competitive pressure is pushing major technology companies into an HVAC optimization race. It notes that Google has used AI to coordinate cooling in data centers, claiming about a 40% reduction in energy used for cooling at some facilities. It also states that Microsoft, Amazon, and Meta are investing in liquid cooling technologies, optimizing airflow and using intelligent temperature control.
The challenge is no longer simply “cooling,” but reducing every kilowatt of electricity consumed.
The pressure is especially strong in Asia, where data-center construction is accelerating alongside the AI boom. The article says Singapore has tightened data center licensing due to concerns about electricity consumption. It also points to Malaysia, Indonesia, and Vietnam as emerging hubs for digital-infrastructure investment, driving large increases in demand for electricity and cooling.
In Vietnam, the issue is described as more sensitive because of a hot and humid climate nearly year-round. The article states that in commercial buildings, HVAC accounts for roughly 30–50% of total electricity consumption, while in data centers—where temperature and humidity must be maintained continuously—that share can reach up to 60%.
As a result, the article concludes that larger data centers translate into larger cooling electricity bills.
Against a backdrop of rising power prices, tightening Net Zero commitments, and continued growth in data-processing demand, the article says HVAC is shifting from an operating expense to a strategic infrastructure priority.
It adds that investors increasingly treat cooling as a long-term financial issue rather than just electromechanical equipment. An optimized HVAC system can reduce operating costs over a facility’s life cycle, particularly for sites that run continuously such as airports, hospitals, and data centers.
The article links this trend to rapid growth in the smart HVAC market. It cites that Schneider Electric launched the Altivar HVAC inverter line, claiming it can save more than 30% of energy for HVAC systems and improve integration with building-management systems (BMS).
While the article frames this as one product in a crowded market, it says the launch reflects a broader shift: technology and energy companies increasingly see cooling optimization as a key front in energy savings.
The article describes an underlying paradox of the AI era: the future of technology depends not only on semiconductors or algorithms, but also on cooling capability. It argues that every AI system ultimately faces a basic physical limit—heat—and that the silent determinant of operational efficiency may be the cooling system behind the chips.
Premium gym chains are entering a “golden era” that is ending or already in decline, as rising operating costs collide with shifting consumer preferences toward more flexible, community-based ways to exercise. Long-term memberships are shrinking, margins are pressured by higher rents and facility expenses, and competition from smaller, more personalized…
