Google on Tuesday said it plans to fund more than two gigawatts of new clean energy in Michigan to serve a large data center campus proposed outside of Detroit.
The agreement with the utility DTE Energy would be funded through the latest iteration of the hyperscaler’s “clean transition tariff,” a rate structure created in 2024 as a way for large loads to pay extra for clean power — or in this case capacity — from technologies too risky or expensive for utilities to invest in without burdening ratepayers.
It has been approved for the use of two geothermal deals in Nevada, with Fervo and with Ormat. Another deal for long-duration storage to power a Minnesota data center is pending approval by the state public utility commission.
Google said that in Michigan, it would invest in a mix of solar and short and long-duration energy storage, as well as add demand response capabilities to its data centers so they can shift power use when the grid is stressed. Specifically, the company plans to fund 1.6 gigawatts of new solar capacity, 400 megawatts of storage, and procure an additional 300 MW of clean power through other bilateral deals with energy producers in the state.
The “clean capacity acceleration agreement,” which still has to be reviewed and approved by state energy regulators, is similar to Google’s deal with Xcel Energy in Minnesota for 1.9 GW of power, including 300 MW of long-duration storage capacity courtesy of Form Energy. Both Michigan and Minnesota have laws on the books requiring utilities to be carbon-free by 2040.
Google’s proposed 1 GW data center campus in Michigan — known as “Project Cannoli” to locals in Van Buren Township — would have 350 megawatts of demand response, a spokesperson confirmed to Latitude Media. It marks the first time Google included demand response under a CTT deal.
The tech giant has been testing tools to enhance the flexibility of its data centers since 2023, when it rolled out a way to reduce a data center’s energy use at times of grid stress by shifting some compute tasks to other times and locations. Initially, the tool only targeted non-urgent tasks, such as the processing of YouTube videos. But last August, following a demonstration with the Omaha Public Power District, Google said its demand response capabilities can also target machine learning, and announced demand response agreements with utilities in Michigan and Tennessee.
That is one example of the kind of flexibility that could be beneficial for both large loads like data centers and the grid itself. For instance, flexibility can potentially allow data centers to secure a grid interconnection faster, and reduce how much new transmission and generation is needed.
But load flexibility remains niche, and skeptics argue that as AI shifts from training models to inference — the actual use of chatbots and agents — hyperscalers may be less inclined to pause activity to aid the grid. Plus, if a data center flexes down or sits idle, that could cost its owner revenue.
The latest expansion of Google’s CTT comes as hyperscalers have promised to foot the bill for all the new generation and grid infrastructure needed to serve AI data centers in order to shield other ratepayers from rising costs. Google said it will fully cover its electricity and infrastructure costs, while helping Michigan transition away from coal-fired power plants.
As part of its deal with DTE, Google also introduced a $10 million affordability fund for local organizations working on home weatherization and energy efficiency programs that can drive down monthly bills for Michigan residents.
Jennifer Granholm, who was Energy Secretary during the Biden administration and has praised the CTT before, said on LinkedIn that the DTE deal is a model for other hyperscalers: “Add new clean supply, protect ratepayers, strengthen grid reliability, invest in local communities, protect water and the environment.”
