The cleanest data center regions in the U.S. are California and the Northwest, but they're also among the most expensive options, according to new research from cloud cost management company Cirrus Nexus.

The research calculated energy consumption and tracked cloud environment costs at the top three public cloud service providers during the beginning of June 2022. The report found while the financial cost of running cloud workloads is usually stable day to day, the carbon intensity of those workloads can fluctuate a lot based on the available balance of fossil fuel and renewable energy.

The wind doesn't always blow, and the sun doesn't always shine, which leaves carbon-intensive fossil fuel power to fill in the gaps. To that point, the research calls for increased investment in energy storage technology to allow power grids to take better advantage of the times when solar, wind, and water power is strong.

"There’s been a lot of investment by government agencies like the Department of Energy, but converting that into commercially viable production has been a challenge. At this point, we expect investment from the private sector to increase as the costs and efficiencies of batteries improve along with government policies like tax incentives or research grants," Cirrus Nexus CEO Chris Noble told SDxCentral.

The Cleanest and Dirtiest Data Centers in the U.S.

The report identifies California as the most sustainable data center region and pegs its workload carbon intensity fluctuations on changes in renewable energy production. During a daily analysis, solar energy peaked at 57% of the region's total energy production, and natural gas made up the difference. But the region's carbon intensity increased as wind power production fell during the week, which was offset by natural gas production in the Northwest.

California is also the most expensive location for data centers, with the price of running workloads deemed 3-10% higher in the golden state than any other U.S. data center region.

Reducing the cost of renewable energy is "about changing behavior," Noble said. "As more enterprises demand accountability for the carbon produced by their computing resources and will only put their workloads in cleaner environments, cloud service providers will start building and providing the bulk of services in cleaner regions. Demand shifts may also encourage other trends like building more local, carbon-free power production and winding down dirtier data centers."

Data centers in the Northwest are also some of the cleanest in the country, according to the research. Hydroelectric power accounted for more than 50% of energy production during Cirrus Nexus' analysis, and that rate remained steady throughout the week.

Although the Northwest is more competitively priced than California, it's still not a very popular data center location for companies with operations outside of the region. Enterprises tend to run their workloads closer to their operations for latency reasons, the report explained.

According to the research, the Midwestern U.S. "was persistently dirty" because coal and natural gas are the region's main sources of energy production. Wind power fell throughout the week, leaving natural gas production to pick up and cover the gaps.

Texas landed in the middle in terms of carbon intensity. Natural gas (43%) and wind power (24%) comprised the majority of the region's energy, and a mix of solar and nuclear made up another 15% of energy production.

The research noted that the fluctuations in renewable energy production and availability serve as another piece of the explanation for why more companies that claim a focus on sustainability and environmental, social, and governance (ESG) principles aren't moving their workloads to cleaner regions.

"'Green' data center regions don’t always stay that way, and can change at different moments in time. During this week, Texas at certain times had similar carbon intensity to that of California and the Northwest," the report reads.

"It was interesting to see how drastically data center regions could fluctuate from clean to dirty. For example, Ireland went from being middle of the pack, then the dirtiest, and ended as the third cleanest data center region, all in one week," Noble added.

What to do?

Since carbon intensity fluctuates more than the actual costs of running cloud workloads, the research recommends companies reduce the cloud carbon emissions by tracking workloads daily to optimize for fluctuations in renewable energy availability.

Tracking workloads is beneficial because "understanding your utilization is critical to optimizing your operations" because consuming less power means producing less carbon, Noble explained.

"Second, carbon intensities fluctuate drastically throughout the day and across regions, but costs from the cloud service providers are relatively stable. So, you can reduce carbon emissions by tracking to these daily patterns and taking advantage of low-carbon intensity regions without really impacting your cloud spend," he added.

In his experience, the majority of companies are still in "exploratory and learning" phases of sustainability where they're just beginning to consider the impact carbon has on their business holistically. But as sustainability conversations have evolved, more enterprises are thinking about the environmental impacts of their entire value chain in terms of scope 3 emissions, which includes carbon emissions associated with IT cloud workloads, Noble noted.

So if this isn't happening already, companies need to include IT orgs in broader sustainability conversations. "Often, IT is a blind spot for enterprises, and they aren’t grasping the magnitude of their IT carbon footprint and don’t have awareness of how they can proactively manage it to make a tangible impact on the planet," Noble said.

In addition, enterprises should determine the financial cost of carbon and apply that knowledge to their operations to make environmentally-informed business decisions, he said, adding that the research found once the cost of carbon is factored into the equation, the true price of running cloud workloads can be 2-9% higher than U.S. providers' published costs.