Today’s digital economy, commercial and industrial digitalization, and the recent explosion in artificial intelligence and machine learning (AI/ML) powered computing are driving massive growth in data centers for both data storage and computational processing. As a result, existing and new-build data centers are one of the United States’ biggest contributors to new electricity demand. For data centers looking to stay competitive, contain power costs, and streamline their carbon footprint, this puts facility energy optimization front and center more than ever.
Data center power demand is forecasted to balloon at a 10% combined annual growth rate (CAGR) through 2030, according to McKinsey analysis. Moody’s Ratings analysis is even more bullish, forecasting 28% annual growth in data center power consumption through 2028.
Said another way, Goldman Sachs predicts that data center electricity will increase a whopping 160% by the end of this decade globally. In the US alone it will more than double, from 17 GW in 2022 to 35 GW by 2030. That’s more than the entire Commonwealth of Virginia and nearly as much as the entire State of New York.
By 2030, colocation (colo) data centers are expected to comprise the largest share of this electricity demand. This drives an innovation urgency, since electricity accounts for about 70% of data center costs. This also means that data center facility energy use — and electricity use especially — are major drivers of the associated carbon emissions footprint, too. The industry is already seeing such effects play out in the real world.
In Georgia, utility Georgia Power has proposed building 1.4 GW of polluting fossil-fueled power plants to meet that state’s data center electricity needs. Hyperscaler Google, for example, has seen its carbon emissions grow by 48% just in the last five years, despite ambitious zero-emissions goals and downward emissions trends before the explosive growth of AI/ML computing. Microsoft similarly has seen a 30% increase in carbon emissions since 2020. But hyperscalers aren’t the whole story. Colo data centers are wrestling with many of the same electricity reliability, power cost, emissions, and other challenges — but often without the internal resources of the Big Tech giants.
Let's take a closer look at five interrelated trends that strike at the core of these issues that are becoming increasingly central to data center operations.
States and local jurisdictions are becoming increasingly aware of (and concerned about) data centers’ potential impact on regional electricity grids, water resources, community and landscape character, and other values. For both new construction data centers as well as “land and expand” evolution at existing data centers, this means potential permitting headwinds — unless data center owners and operators can demonstrate best practices in siting / interconnection locations, energy strategy, and climate / environment planning to address concerns.
Because of transmission capacity constraints, grid congestion, and service bottlenecks, both existing and new data centers need to better plan for power quality and reliability. The US interconnection queue of new wind and solar projects waiting to come online is larger than the entire US generation fleet of power plants today. Even if those clean energy generators could come online overnight, the transmission grid needs huge expansion to better connect supply with demand centers. Consequently, data centers now need to take into account potential onsite generation and UPS backup power options to stay online, even when the grid falters.
According to Morgan Stanley research published last month, the global data center industry is on track to emit 2.5 billion tonnes of carbon emissions through the end of this decade. That’s on par with the entire annual emissions of Russia, the world’s fourth-largest emitter of greenhouse gases. In response, procuring clean, green power is increasingly the name of the game for data centers — whether through onsite solar PV, utility green tariffs, or instruments for offsite renewable energy procurement such as power purchase agreements (PPAs) and virtual PPAs. (It’s also important not to forget upfront facility energy efficiency, which can reduce electricity consumption and associated emissions, even before data centers target clean energy for the balance of their energy demand.)
In light of the aforementioned three trends, a growing number of states have created regulations often focused on climate and emissions goals at the state level broadly, for buildings generally, and/or for data centers specifically. California’s Building Energy Savings Act (ESA), for example, directs the state’s Energy Commission to develop a strategy to improve energy efficiency and reduce greenhouse gas emissions in large commercial buildings. Meanwhile, on the opposite coast, in Virginia, several counties are considering legislation that would add approval steps to building new data centers, or stunt data center development altogether.
AI tools draw almost ten times more electricity than a typical Google search. With that electricity consumption comes heightened environmental and emissions footprints, especially when the power comes from traditional fossil-fueled power plants. From public pressure, to media scrutiny, to government programs, to corporations’ voluntary sustainability targets, to colo data center landlords wanting to remain competitive in attracting tenants, reducing emissions and other impacts is more important than ever.
These macro trends will only intensify, making a well-designed energy strategy crucial to overall success. Attracting new customers to data centers and retaining current ones hinges on controlling all-in costs, solving grid connection problems, and minimizing outsized environmental impacts. Mantis understands the challenges facing data centers. In subsequent articles, we’ll take a look at how Mantis works with operators to reduce costs and emissions.