The surge in applications of artificial intelligence (AI) and machine learning (ML) – as well as the more fundamental needs of access to the Internet and cloud computing – are transforming the economy and the electricity infrastructure that underpins it.

From large hyper-scale data center facilities to much smaller edge data facilities, the trends in information technology (IT) and telecommunications are challenging status quo solutions from incumbent electric utilities. Coming in all shapes and sizes, data centers create new pathways for decarbonization and a more decentralized energy solution ecosystem.

Today, there are more than 8,000 data centers worldwide comprising roughly 1 to 2% of global energy use. Over half of these data centers have been deployed in the U.S., consuming 2 to 4% of the nation’s energy supply.

A report released by the U.S. Department of Energy in late 2024 revealed that data center electricity load growth has tripled over the past decade and is projected to double or triple again by 2028. The report, prepared by Lawrence Berkeley National Laboratory, states that this surge in demand can be met with clean energy technologies. Yet utilities warn that they may not be able to supply enough power of any kind for these data centers. I would argue that the key for data centers is developing cleaner on-site energy assets configured into microgrids.

The worlds of IT and electricity have often mirrored each other. Aided by deregulation, IT innovation advanced first. A similar shift to deregulate energy systems is spurring on new technology and business models in the provision of electricity. The worlds of IT and electricity infrastructure are now overlapping in a myriad of ways. One could argue that microgrids, comprised of diverse distributed energy resources (DERs) could be the compelling connective tissue between these two previously somewhat siloed worlds.

Historically, most data centers have relied upon diesel generators and lead-acid batteries integrated into uninterruptible power supply (UPS) systems to provide resiliency. A move toward cleaner assets and microgrids for data centers started about 8 years ago. Recent pressure from the investment community pushing data centers toward more nimble decarbonization strategies has led the industry to devise creative, on-site clean energy solutions as an alternative to carbon credits from off-site wind and solar farms. This latter approach relying upon carbon credits failed to drive down the cost of innovative AI technologies and other novel solutions which could efficiently reduce emissions both for data centers and other commercial and industrial customers.

Enter the Microgrid

The scramble to find viable data center sites with adequate electricity supplies in the U.S. and elsewhere has spawned a variety of novel solutions. Here are just three examples:

• A proposal by Amazon to tap the existing Susquehanna nuclear power plant would have saddled utility ratepayers with up to $140 million in new costs, according to testimony filed by AEP and the utility conglomerate Exelon. The Trump administration may breathe new life into the nuclear industry, but the Federal Energy Regulatory Commission (FERC)’s two Republican board members denied Amazon’s interconnection request late last year for 960 MW of firm power for Amazon.

• A more successful approach is being pursued by Google, which has set a goal for all of its data centers to run on carbon-free electricity by 2030. It is tapping low cost, 100% baseload geothermal energy in Nevada for new data centers. The availability of geothermal energy is quite limited geographically, so this approach is not widely replicable throughout the U.S. or the world.

• A small 150-kW edge data center developed by Greensparc, powered 100% by surplus run-of-the-river hydroelectricity, was installed within the Cordova microgrid in Alaska. The microgrid is operated by a rural cooperative and has been hailed as one of the most advanced grid systems in the world. While applicable to many of Alaska’s rural and remote microgrids - as well as similar applications in Africa, the Asia Pacific and Latin America this model is less relevant to MW-scale data centers.

Large microgrids connected to traditional electricity networks have the broadest potential for newly proposed MW-scale data centers looking to meet sustainability goals. Optimal Power Solutions, which has deployed thousands of remote off-grid microgrids throughout the Asia Pacific region, is part of a new data center initiative called GreenSquareDC and is looking to deploy its first project in Perth, Australia: a 96 MW data center for Microsoft dubbed “WAi1,” which will be powered by a variety of zero carbon technologies. The project just secured over $1 billion in capital investment, some of which could finance additional green data centers. The project includes not only large solar and wind energy assets, but on-site biomass fuels for advanced turbines and green hydrogen fuel cells. The diverse resources would be managed within the context of a large, sophisticated microgrid also supported by dual utility feeds to ensure hyper reliability and resilience. 

This innovative model could be replicated throughout the U.S. Microgrids implemented by private developers could help bring clean energy assets to the table more rapidly, with utilities providing backup power. Utilities, however, do not have to be relegated to a second tier within the microgrid ecosystem. Though most microgrids in the U.S. are developed by non-utility companies, increasing numbers of utilities are working with the private sector to bundle existing distributed energy resource (DER) assets into microgrids.

Nexus Between Microgrids and VPPs

An advantage of leveraging grid-tied microgrids to serve data centers is their ability to provide other flexibility services ranging from traditional demand response (DR) - islanding would take the total load of the data center off-line to free up existing electricity supplies for other utility customers - to the more surgical actions of a virtual power plant (VPP), as described in a recent blog by Molly Podolefsky of Clarum Advisors.

Comparison of the IT and energy industries yields the ironic observation that DR has very different meanings for each. While DR is celebrated as the key to unlocking flexibility in the energy sector, when one uses that acronym in the data center world, most folks think of “disaster recovery,” which is a big, big negative. Though likely not directly the result of this coincidence, data centers have traditionally shunned any interest in supporting the grid through DR. Having been solely preoccupied with their own uptime, any process that could possibly compromise on-site reliability was simply ignored. As a result, the vast majority of data centers invested in on-site generators and UPS assets that sat unused most of the time.

With the help of AI provided by these very same data centers, DER assets can be fine-tuned to deliver just-in-time services optimized for on-site resilience and broader grid support, literally at the same time. By substituting solar, fuel cell and flexible fuel generators for a monolithic portfolio of diesel or gas-fired generators, whose participation in wholesale markets may be limited by regional and/or state-level regulations, data centers can become a solution, rather than a challenge, for capacity constraints, grid congestion and balancing.

As data centers bring their loads to the energy solution ecosystem, microgrids converging with VPPs offer the best of both worlds. Clean supply that is not disrupted by extreme weather, cyber-attacks or even terrorist malfeasance, is exactly what the data center industry is seeking. Utilities and wholesale grid operators are looking for advanced near-term solutions that provide bidirectional value.

Data centers may indeed be a problem for some. For others they are the solution, if wrapped up with microgrids that leverage AI to optimize operations, reduce emissions, and shrink costs across the board - while also delivering on the promise of a decarbonized power grid.

Coming back to the State of Alaska for a moment, hyper-scale data centers are viewed as new loads which could decrease instead of increasing energy costs on the so-called Railbelt grid, the only transmission system operating in the state. Alaska has among the highest energy costs in the nation. With planned fiber connections, it could ultimately serve markets in the Asia Pacific. More immediately, the Dunleavy Administration is courting large data center companies to come to Alaska to take advantage of the free cooling available in the Arctic, mimicking what Iceland has done to reduce energy costs.

Data centers are a global industry. The current U.S. administration’s refocusing on fossil and nuclear industries as energy infrastructure priorities will have an impact, yet data centers and other regional markets continue to march down a path of sustainability. Microgrids that can tap AI to improve operational excellence and leverage a wide array of distributed generation sources, energy storage technologies and electric and thermal loads, offer an attractive alternative to large, centralized options such as nuclear power plants or even large-scale geothermal facilities. Microgrids can be brought online quickly to solve energy shortages if the right mix of incentives and regulations is in place. The U.S. would be wise to pursue multiple options for meeting the surge in electricity demand for new data centers. I’m betting on microgrids and VPPs to emerge as the best option in most cases.