As the AI revolution has shot onto the scene, the energy implications of feeding it have also dramatically come into view. Data centers are being constructed with a frenzy. And, in spite of the current building surge, geometric growth is on its way. The US already has about 40% of the world’s data centers. This US base is projected to increase by about 10% per year through 2030.
But data centers have had a complex history, originating with the US military in 1940 to support the very first computer’s secret calculations of projectile paths. Through time, computer centers migrated to public sector commercial use, where they now constitute the backbone of the Web and are independent economic entities in and of themselves. Northern Virginia data centers, for example, used about four times the energy of the Dallas or Silicon Valley markets.
The current and projected data center growth, however, is once again placing data centers in conflict with other social goals. While the net social gains in energy efficiency are clear, the data centers themselves offer an important focus for efficiency growth. The American Council for an Energy Efficient Economy, for example, has indicated that energy saved in other sectors are ten times that which is consumed by the IT area.
In spite of the externalities that data centers may provide in saving energy and reducing pollution in other sectors, the current growth of data center-based IT technology poses a clear environmental challenge in at least three areas:
1. water usage
2. energy consumption
3. greenhouse emissions
Essentially, most data centers are giant air conditioners which remove the heat generated by electrical devices positioned on rows of servers. And greenhouse gas emissions are a direct and indirect result of data center activities.
Data centers have been recognized as an environmental challenge for a number of years. Before there was AI, crypto mining was a strong driver of data center growth and evoked international concern about the pollution the centers created. Now that data centers consume around 3% of the global energy supply—indeed, more than the entire UK—the environmental challenge is increasing.
Data Centers are IoT
The economic importance of growing AI in the US makes the environmental challenge one of managing the growth of data centers even if they may come with adverse environmental impacts.
Data centers are, in a sense, the footprint of AI and its emergence. Furthermore, the cloud analytics supported by the data center are a key to environmental controls across many other industries.
Moreover, the current system for cooling data centers generally adopts room air conditioner technology. It can often create local hot spot problems, forcing in the room systems using lower temperatures and higher energy consumption. Data shows that the power usage effectiveness (PUE) is low.
Improving data center controllers is key
The sheer scale and density of energy usage in a data center make infrastructure management an essential feature of data center economics. Data Center Infrastructure Management (DCIM) can be designed for optimal processing even with rising rack densities. These systems integrate sensor data on heat, humidity, and other environmental elements affecting data center performance. The systems can be AI-enhanced with compelling visuals. The modern DCIM goes far beyond simple airflow management. It can improve efficiency, reduce costs, and enhance sustainability. With the growing concerns about cyber security, DCIMs can be a key element in security systems.
The improvement of advanced DCIM is an environmental and economic priority in the IT industry, especially for older, less efficient data centers. Multi-data center controllers are also a consideration. Less efficient data centers can be run from new hyper efficient ones.
It is essential that these data center controllers conform to Interoperability Protocols. Adherence to standardized interfaces helps in communicating with North Bound Operation Support Systems (OSSes).
The rapid growth of data centers provides a compelling context for new investments and upgrades of DCIM. This coupled with the pace of server and technology changes invite a conversion of virtually all DCIM to smarter AI models. Improved controllers can help solve the sustainability problem created by surging AI energy demands.
