Balancing Computational Power with Sustainability

Data Centers that continue to design and build according to the principles used over the last decade will sadly find themselves left behind. Due to increasing weight of racks and equipment, raised floors are no longer viable. The movement of air, as a cooling method, is also becoming outdated as the more intense compute environments demand a better solution that removes more heat. Additionally, there is a need to reuse the excess heat. Not such an immense challenge in temperate geographies, but in the tropics, this can be very challenging.

When the external ambient temperature is above the ASHRAE upper limits how to reuse heat? The challenge is often the same no matter the geography, a 24x7x365 solution is needed. District heating is often seen as the best use case; however, how much heating is needed in the summer months? Those in the UK may maintain that most of the year is the use case. Joking aside, putting heat to a process that needs constant dry heat is a use case everywhere. Drying biomass before energy conversion or drying of crops before a further process such as roasting could be alternative use cases.

Maintaining a base load of heat such as swimming pools is another area to consider. Data center heat has been relatively ‘low grade’ around 35 degrees centigrade and heat pumps are needed to make this usable for heating. By moving to a liquid cooling solution more heat can be extracted at a higher temperature, and therefore, more use can be made of this. AI can be turned on itself with enhancing designs to do more with less and be more efficient.

Optimizing Data Centre Redundancy for Energy Efficiency

Data Centers cannot be turned off when they are not being used, however, the servers within them can be consolidated, virtualized, which minimizes energy usage. The use of a multi-tier facility also can reduce or optimize the need for wholesale redundancy within both the power train and the cooling environment. Organizations need to review the needs for all of their applications to be in a 2N or N+1 environment. For example, the resilience and redundancy that is applied to trading systems cannot be the same requirement for the travel booking system. Therefore, only a small proportion of application need to be in a high availability zone with full redundancy. Other applications can be grouped together in a zone that does not require such high resilience. The infrastructure supporting these differing use cases will be different and the resultant energy consumption will be lower. Rating organizations are fully aware of this need and can support this.

AI at the Edge: Balancing Data Growth with Energy Efficiency

By moving computing to the edge, this will reduce the overall power draw of more centralized facilities. The inference instances of AI applications will never take as much energy as the training of the application in the first instance. True edge facilities will bring compute closer to the end user. The only way to reduce energy consumption is to use less of it, however, with the growth of data this is unlikely to ever happen. AI chips are becoming more efficient continually with the number of transistors on the chip increasing at a high rate. This means that they will be come more efficient in the way they work.

The Complexities of Renewable Energy for Data Centers

The use of energy falls into Scope 1 and Scope 2 emissions. Therefore, to be able to reduce the carbon content in these areas is good. However, the use of Virtual Power Agreements is tantamount to ‘greenwashing’. The only way that a reduction of carbon in the utility supply happens is if renewable power source is actually injected into the power grid that the data center is drawing power from. Other challenges are the renewable power is ‘peaky’ in its generation capability. No solar power at night and no wind power when the wind does not blow. In countries such as Malaysia, the average wind speed barely reaches the minimum viable wind speed for generation. Therefore, any renewable needs to be allied with a sufficiently large Battery Energy Storage System (BESS).

Anticipating the Next Decade of AI Innovation

It is impossible to say what technologies may arise in the next decade. If we look back in the last three years we could not have seen how big and pervasive AI would be, the industry designs are still playing catch up with this. However, we do need to look at what the next iteration will be, where is the industry heading. Although it can be difficult to predict we need to look at doing more with less. We need to continue to develop alternative cooling solutions beyond the current range of cooling liquids. The use of refrigerants may see a resurgence as rack density increases above 100kW. There will of course be an increase of the floor loadings along with the increase of power draw.

The Dual Challenge of AI-Ready Data Centers

Typically, across the industry, more and more GPUs are being procured and installed. This needs the Data Centers to be re-designed or retrofitted with additional power density, cooling needs and denser fiber paths in mind. This can be achieved. Major hyperscalers are offsetting carbon footprints by either purchasing renewable energy certificates or trading green power from a capable producer.

Mission Critical, Energy Neutral

Redundancy and reliability are one of two core pillars of mission critical infrastructure like Data Centers. These can’t be diluted. Energy consumption is not directly impacted negatively by these two pillars since semiconductors only use power that is needed for them vis-a-vis how the power is being distributed to these devices from the backend.

Edge Computing’s Energy-Efficient AI Promise

Edges are gaining more and more popularity since the computation can happen closer to the end customer location, giving an enhanced experience of the cloud services to a specific set of applications/use cases. Since the energy footprints are typically smaller it doesn’t have too much impact on energy consumption, even with AI related deployments.

Navigating Geography in Data Centre Sustainability

Ample renewable energy sources are available outside the city limits and it’s not easy to transfer or trade these in certain geographies. The compromise is made by procuring renewable energy certificates (RECs) in the source location to offset the consumption footprint in the destination location. Only non-critical loads can be served via renewable energy sources owing to its fundamental behavior of being attached to a natural cyclic phenomenon.

Emerging Technologies and Design Principles

Multiple. Redundancy can be built at semiconductor device level rather than at higher power levels. Controlled Hydrogen cell and Nuclear Fuels can be well researched for Direct Current (DC) applications. DC power use within DC environments may reduce conversion losses.

Data Centers Push Innovation to Meet Green AI Demands

Data centers are rapidly changing to handle the growing demands of AI while aiming to reduce their carbon footprint. They’re using specialized hardware like GPUs and AI accelerators to boost performance without increasing energy use. High-density servers pack more computing power into smaller spaces, cutting down on overall energy consumption.

Cooling technologies are also advancing. Many data centers are moving to liquid cooling methods, like immersion cooling, which are more efficient than traditional air cooling. Some are using AI to manage cooling systems in real-time, optimizing energy use based on current conditions. Additionally, free cooling techniques that utilize outside air help reduce reliance on energy-intensive cooling methods.

On the power management side, technologies like Dynamic Voltage and Frequency Scaling adjust processor power based on workload, saving energy during less intensive periods. Intelligent power distribution units monitor and manage energy use more effectively.

Renewable energy integration is key. Data centers are installing on-site solar panels or wind turbines and using energy storage solutions to balance supply and demand, especially when relying on intermittent renewable sources.

Risk-Based Strategies to Cut Energy Waste

Finding the right balance between redundancy for reliability and minimizing energy use is crucial. Data centers are adopting risk-based redundancy, focusing on critical services rather than a blanket approach, which cuts down unnecessary energy consumption. Virtualization allows for redundant systems without extra hardware.

Using energy-efficient equipment even in redundant setups helps reduce power usage. Modular data center designs enable scaling up as needed without overprovisioning. Operational efficiency improves with on-demand resource allocation and AI-powered predictive maintenance, which can reduce the need for redundant systems by addressing issues before they lead to failures.

Edge Computing to Reshape Data Center Energy Dynamics

Edge computing processes data closer to where it’s generated, reducing latency and the energy costs of data transmission. This shift means less load on central data centers but could increase overall energy use due to many edge devices. Designing low-power, high-performance edge hardware is essential to keep energy consumption in check.

Hybrid models that combine edge and cloud computing optimize workloads—immediate data is processed locally, while more intensive tasks are sent to central data centers. Building smaller, energy-efficient local data centers supports sustainability goals.

Navigating Complex Paths to Renewable Power Future

Renewable energy is becoming more prevalent in powering data centers, but transitioning isn’t without challenges. Solar and wind power depend on weather conditions, so consistent energy supply requires effective storage solutions. High upfront costs and regional grid limitations can also pose obstacles.

To overcome these challenges, data centers are investing in advanced energy storage and choosing locations with better access to renewable resources. Collaborating with energy providers and governments helps develop necessary infrastructure and supportive policies.

Sustainable Practices and New Tech Critical for AI-Ready Data Centers

Looking ahead, new technologies will be crucial. Advanced cooling methods like expanded immersion cooling can handle high-density AI workloads more efficiently. AI will play a bigger role in operations, from predicting demand to automating power and cooling optimization.

If we look even further, quantum computing offers the potential for massive processing power with lower energy use per calculation. Deeper integration of renewables through microgrids and innovative storage solutions will help manage energy supply. Sustainable building practices and designing for recyclability will further reduce environmental impact.

Staying ahead of regulations and engaging in industry collaborations will be important for adopting best practices and ensuring data centers remain efficient and compliant.

AI Surge Forces Data Centers to Reimagine Energy Solutions

Data Centers are going through a significant transformation to address the needs of huge energy demands poised by AI and HPC (high performance computing). This is actually bringing Data Centers to the edge to keep up the sustainability aspect of the business. The exceptional rise of Gen AI has led to huge demand in computing and storage capacities around the world. For example, generative AI models are projected to cost around approximately US$ 80 billion annually by 2028 which is more than double the annual expenditures of major cloud providers such as AWS. As the demand grows, energy consumption also increases. Goldman Sachs predicts that AI will be accountable for 18% of energy use in Data Centers by 2030, which approximately doubles down the emissions.

Data Centers are implementing innovative solutions to improve this. For example, AI enabled workload management is becoming more widespread to manage the data center’s workload more efficiently and thus managing the energy and cooling needs. Innovative cooling solutions are the other avenue. Air based cooling solutions are replaced by advanced liquid cooling systems that can reduce energy consumption by 50%. Cooling system accounts for nearly half of Data Centers energy consumption and thus these improvements are more vital.

Smart Redundancy Systems Drive Sustainable Uptime

Tolerance for down time is zero in Data Centers’ availability. Businesses trust Data Centers and its operators to provide aftmost redundancy for their business operations and data. Usually, Data Centers implement many redundancy methods such as N+1 or 2N etc. Balancing the availability of systems and sustainability is pretty tricky and cost sensitive. Data Centers are implementing energy saving strategies, especially on cooling systems such as Hot Standby Sparing (HSP) and Direct to Chip cooling etc. By implementing modular systems, optimized cooling systems and AI driven workload and redundancy management, we can balance the need for availability and the energy needs for it.

Edge Networks Cut Energy Waste in AI Era

Edge computing leads the ‘cloud offloading’ model where data is processed locally or closer to the user and reducing latency and bandwidth usage. Data centers are becoming more distributed in nature and started operating in hubs and spokes kind of models. This operational change provides lots of flexibility and control over resource utilization, especially on the energy usage. I believe this evolution is supporting the sustainability transition in data center industry.

Tech Giants Lead Renewable Push Despite Grid Challenges

Role of renewables is super critical for the journey towards sustainability. Data centers account for close to 1% of the global electricity demands and contributing to 0.3% of carbon emissions. Transitioning to renewable energy such as hydropower, wind, solar and geothermal etc. will reduce these emissions substantially. Major tech companies have committed to use 100% renewable energy for their Data centers by directly investing in in renewable energy projects. Recently Microsoft has signed a power deal to help resurrect a unit of the Three Mile Island nuclear plant in Pennsylvania. Though the challenges to move into renewable energy is also mounting up. The inherent intermittence of renewable energy is a great deterrent for Data centers which should be available all the time. This variability leads to more costly energy backup solutions which in turn affects the bottom line for the operations. Another major problem is the land usage requirement for these renewable projects.

Smart Technologies Converge to Power Green Computing Future

I believe the combination of multiple emerging technologies such as Edge computing, smart grids and advanced battery technology, advanced cooling systems and advancements in virtualization and AI will shape the future of data centers. When combined with sustainable operational practices with these emerging technologies, the data center industry can control the environmental impact and support ever growing computational demands.

Innovative Cooling Solutions Tackle Data Center Carbon Challenges

A large portion of data centers’ carbon footprint is a result of water and energy consumption. Cooling, essential for stable processing efficiency, accounts for about 40% of electricity demand. With the growth of developing technologies including Artificial Intelligence (AI), Internet of Things (IoT), and cloud computing, the already enormous energy and water footprint of data centers is set to increase at an alarming rate over the upcoming decade. Against this backdrop, technologies to reduce the resource consumption in data centers have been growing rapidly over the past decade. The efficiency of both IT hardware and cooling has been improved and there’s a shift away from small, inefficient enterprise data centers towards more efficient cloud and hyperscale data centers.

Specific to cooling, AI itself can be a means to achieve real-time, on-demand cooling. For example, Google reported using its DeepMind AI to reduce the electricity demand of their data center cooling systems by 40%. New cooling methods such as water-cooling and immersion cooling are also gaining popularity. In addition, the development of new heat dissipation materials, including cooling liquid, is also on the rise. As we move towards a circular economy model, exploring alternative water sources such as rainwater harvesting can also pave the way to greener operations. These energy and water-efficient cooling technologies not only reduce environmental impact but also relieve a significant burden in terms of operational costs in the long run.

Smart Pumps and Motors Drive Energy Efficiency

Despite the challenges of balancing redundancy, reliability, and energy efficiency in data centers, there is still substantial opportunities for energy optimization. Starting with the fundamentals, pumps in data centers should be correctly sized. This might seem obvious, but many data centers overlook this. Often, data center designers, while experts in cooling systems, may not have specialized knowledge in pumps. With a primary focus on ensuring system uptime, it’s common for pumps to be oversized. Oversized pumps consume far more energy than necessary, leading to significant waste.

Secondly, we’ve observed some common pitfalls in pump motor selection. Many data centers in Southeast Asia continue to use IE3 motors, even though more energy-efficient alternatives like IE5 motors are readily available. Our calculations indicate that an IE5 motor can deliver approximately 5% energy savings compared to an IE3 motor in a 10MW data center. This is the energy savings that can be achieved from the motor alone.

Once these basic aspects are addressed, we can further optimize energy consumption through digital technologies. Intelligent pumps and sensors can efficiently manage water flow in air or liquid cooling systems, saving energy. For example, Grundfos’ intelligent pumps optimize cooling tower operation by monitoring temperatures, adjusting pump speed and fan operation whilst ensuring smoother pump operation. Through this way, energy is used only when needed, thus lowering energy consumption.

Research shows that 85% of a pump’s lifetime cost is attributed to its daily operation. Thus, a more efficient pump translates to lower energy consumption and significant cost savings. Meanwhile, reliable and efficient pump systems ensure that data center cooling systems can operate continuously 24/7, meeting the stringent requirements for stability, reliability, and safety.

Intelligent Cooling and Renewables to Drive Future-Ready Data Centers

We believe that intelligent and modular cooling systems will continue to pave the way for energy-efficiency at data centers, by allowing for more tailored solutions that rely on needs-based cooling to reduce energy usage. Modular systems that deploy smaller cooling units across server racks in place of a more centralized cooling system, for instance, can self-regulate and customize the performance of singular units based on the requirements of their corresponding racks. Intelligent functions play a similar role, allowing for real-time monitoring and adjustment that automatically optimize functions to minimize energy use. The incorporation of renewable energy sources like solar panels or geothermal energy across operations can also enable data centers to reduce reliance on fossil fuels and thereby minimize their carbon footprint.

With a more robust approach to sustainability that considers impact across multiple touchpoints, data centers can demonstrate a strengthened commitment to the cause, which sharpens its competitive edge within the industry. Additionally, while renewable energy sources are typically associated with fluctuating supply due to their dependence on natural weather conditions, the use of intelligent data insights and analytics, coupled with grid integration and diversification efforts, can help to overcome such challenges to afford a more reliable source that can power the performance of data center hardware.