The relentless pursuit of artificial intelligence has accidentally converted the modest data center and transformed a once common infrastructure into a critical technological bottleneck and a lucrative new border for traditional industrial power plants. This profound shift, which is largely raised by the accelerated product cycles of Nvidia and the insatiable requirements of the AI, forced a complete new architecture of the physical backbone of the digital world.
https://www.youtube.com/watch?v=yfhn6mxzkls
Andrew Obin, Senior Industrials Analyst at Bofa Securities, recently spoke quickly with CNBCs about this underestimated revolution and emphasized how companies such as GE, Honeywell, Eaton, Parker and Rockwell are now of central importance for the bird boom in the data center. His analysis deals with the question of why electrical and cooling systems, which were previously considered subsequent thoughts, have become high -quality, high -quality components and in which demanding investors should look for opportunities in this developing infrastructure trade.
In the past, data centers worked in a conventional level, typically around 10 megawatts, with the standard 12-volt power supply and the Rack-HLK and Electrical Systems being used. As Obin explains: reliability was of the greatest importance and prompted big players to choose high -ranking suppliers, but the underlying technology remained largely undifferentiated. This era treated performance and cooling as a necessary supply company and not as an integral high-tech components.
Today's landscape is dramatically different. The average data center is now around 100 megawatts, a ten -time increase in electricity requirements. This escalating power density is not just an incremental change. It requires completely new approaches for design and operation. The costs per megawatts have therefore increased, and Obin noticed: “Now an average data center is around 100 megawatts, is there? The costs per megawatts are 50 million.” This astonishing figure underlines the immense investment editions that are now necessary to build up and maintain these Hyperscale -AiFabrics.
The catalyst for this transformation is the Advanced Computing Architecture, which underpins modern AI, in particular the high-performance GPUs of companies such as Nvidia. These powerful chips create unprecedented heat values and require a significantly more sophisticated power supply. Obin emphasizes: “Suddenly you start doing something completely different. Suddenly the HLK system, the cooling system you had before, will not work. Suddenly the electrical stuff, if you go into the Kyber rack, you have to effectively tear all your electrical things out into the middle of the aden.” This is not upgrade. It is a wholesale replacement.
This new reality means that data centers of 12-volt systems change to higher DC (DC) and possibly run at 400 or even at 800 volts. Such a fundamental change in the electrical infrastructure requires special switchgear, power division units and uninterrupted power supplies (UPS), which can cope with these increased tensions and currents safely and efficiently. The former commoditized components are now tailor -made solutions.
Similarly, conventional air cooling is no longer sufficient for intensive thermal loads, which are generated by AI-optimized racks. The liquid cooling, including immersion cooling, always becomes essential, which leads to an increase in demand for advanced calves, cooling towers and highly developed liquid management systems. These are not minor adjustments, but critical technical services.
The financial effects are significant. Of these $ 50 million per megawatts center costs, Obin estimates that the electrical infrastructure is assigned around 1.9 million US dollars and the cooling of $ 1.4 million. If these highly specialized systems fail, the consequences are direct and catastrophic. “You make the mess,” he warns, “40 million dollars chips per megawatts, almost burning within 30 seconds.” This strong reality increases the importance of these industrial components from mere supply companies to mission -critical technology.
This shift has created a new class of “winners” among industrial companies. Companies such as Eaton, Schneider Electric and Vertiv, which offer switchgear, electricity distribution and UPS systems, have an unprecedented demand for their specialized electrical solutions. SPX Technologies, Daikin Industries, Trane Technologies and Johnson Controls, the leading manufacturers of cooling towers and chillers, are at the top of the thermal management revolution. Even generator manufacturers such as caterpillar, rolls-royce and cummins see increased opportunities because data centers need a robust security service for their expanded operations.
These industrial actors are no longer just suppliers; They are integral partners in the AI ecosystem, their innovations enable direct computer generation directly. “This has become a technological bottleneck,” concludes Obin, “without the plan of Nvidia not working.” The AI revolution is often framed by the lens of chips and software, depends equally on the physical infrastructure that supplies it and cools down. Understanding this mutual dependency is crucial for anyone who navigates through the future of technology and investment.