How AI-Enhanced Cloud Strategies Can Cut Down on Energy Consumption

We have often seen Jensen Huang mention energy efficiency as one of the core features of NVIDIA’s new products or services. This is not surprising as we know by now that one of the challenges in this rapid evolution of AI is its energy demands. 

The surge in AI technology has led to a notable spike in energy demands, especially within data centers. This spike isn't just driving up operational costs for businesses, but it's also taking a toll on the environment.

Organizations are considering and even using various methods to reduce AI’s energy demands, including cooling innovations such as evaporative and submersion cooling, and specialized hardware, such as energy-efficient AI chips. 

Developers are now focusing on more efficient algorithms and specialized neural network designs to cut back on power consumption. However, these measures alone may not be enough. 

Could we use AI itself to reduce its energy consumption and resource wastage? What if we could harness predictive analytics to anticipate workload demands and optimize energy allocation dynamically? 

This is where cloud infrastructure optimization using AI comes into play. By using AI-driven technologies to automatically scale resources in real-time, businesses can reduce the need for over-provisioning.

Zesty, a provider of cloud management solutions, leverages cloud's inherent flexibility to “predictively adjust” resources based on application needs. 

In 2019, Maxim Melamedov and Alexey Baikov launched Zesty after noticing that cloud infrastructure was lagging behind the rapid changes in business environments. By integrating machine learning, Zesty was able to solve inefficiencies in the cloud infrastructure. With the rapid growth, Zesty now serves over 300 customers globally.  

In an interview with AIwire earlier this week, Maxim Melamedov, CEO of Zesty, remarked “A lot of the operations done today on the public and private cloud infrastructure require prediction and capacity planning.”

“You have to figure out how to align and achieve the perfect balance between stability, performance, and cost,” added Melamedov. “Now in a very complex world of multiple applications, multiple data flows, multiple user journeys, it's practically impossible. And the prioritization usually goes as follows. Stability first, then you have performance and if you're clear on that part, then you figure out how to optimize for cost.”

The devastation caused by the Palisades fires has brought energy usage by the data center under the scanner. Several notable figures, including energy experts, politicians, and environmental activists, have raised concerns about the sustainability and efficiency of current practices. There are calls for urgent reforms for the adoption of greener technologies to reduce energy consumption and wastage. 

The concerns about energy usage by data centers are not unfounded. According to the International Energy Agency (IEA), data centers used 460 terawatt-hours (TWh) in 2022. This represents about two percent of global energy usage. 

To put this into perspective, the energy demand of data centers alone could power the entire state of California for more than a year. This enormous energy usage is having a major impact on the environment. It is contributing heavily to carbon emissions and resource depletion. 

Could cloud infrastructure optimization help? According to Zesty, their AI-driven platform can address this issue by automatically scaling resources to meet application demands in real time. This approach is aimed at reducing over-provisioning by using resources more efficiently, and as a result, reducing energy consumption. 

Melamedov shared that while public cloud utilization is typically at a decent level, the same can’t be said about private cloud, which can drop extremely low. This could be a result of better cloud utilization by large cloud service providers compared to the manual management required by private clouds. 

Zesty leverages an advanced AI model trained on various data sources to estimate the precise cloud resources needed for applications at any moment. By interpreting these predictions, Zesty dynamically adjusts the cloud infrastructure, automatically resizing storage volumes and managing cloud instances.

Commenting on the intricate balance between performance and energy efficiency in cloud infrastructure,  Melamedov emphasized that it's about priorities. He said, “There’s going to be a trade-off, and organizations have to focus on the areas of optimization that make the most sense for them.”

According to Melamdeov, the precision in scaling resources is crucial to maintaining both performance and energy efficiency. So instead of just providing recommendations, Zesty aims to go further by implementing predictive models to ensure that the optimizations result in a positive ROI for its customers.

Melamedov acknowledged that while cost-saving was initially Zesty's primary focus, the company has realized that optimizing cloud infrastructure also has substantial environmental benefits. He hopes to continue working on finding more sophisticated ways to manage cloud resource efficiency. This can go a long way in contributing to the well-being of our planet and its inhabitants.