New Study Maps the U.S. States Hosting the Most AI Clusters in 2025

• Texas has 17 GPU cluster sites across the state, more than any other region in the country.
• Tennessee operates 4 data centers but houses over 1.27 million AI chips, the highest count in America.
• The South accounts for the largest concentration of AI infrastructure in the US, with a total of 26 cluster sites across the region.

Companies invested over USD 200B in AI infrastructure last year, with much of that money going toward building data centers and buying specialized processors. A new study by mineral rights brokerage firm Texas Royalty Brokers examined AI data center distribution across the country to identify the states that host the most GPU clusters as of 2025.

The research compiled information on 214 AI data centers across the US, which were then grouped into 83 distinct cluster sites. These facilities house GPUs, which are specialized chips designed to run AI programs and handle complex calculations. The study tracked each location’s AI chip inventory, processing capability using H100 equivalents as a benchmark (H100 is a specific type of AI chip used as the industry standard for measuring computing power), and power capacity in megawatts. States were ranked solely by cluster count, with the other factors providing context about processing power and energy consumption at each location.

Here’s a look at the top 10 states with the most AI data centers and GPU clusters:

You can access the complete research findings here.

Texas has the largest AI infrastructure in America, with 17 clusters spread across the region. These facilities house more than 811K AI chips with a combined processing power equivalent to 6.6M H100 units, the highest computing capacity in the country, and six times larger than the typical state in the top 10. Texas generates 547M megawatt-hours of electricity annually (the most of any state in the nation), and given the size of these data centers, AI consumes about 15% of that total.

California comes in second with 8 data centers, making the state the main AI computing hub on the West Coast. These sites store approximately 216K processing units with 190K H100 equivalents. California has a massive power output of 216M megawatt-hours, second only to Texas, and with AI facilities consuming just 1% of it, the state’s grid could easily support even more infrastructure if needed.

Virginia comes in third place with 5 GPU cluster sites. The AI companies operating these clusters utilize around 163K AI chips with processing power equal to nearly 145K H100 units. Virginia also generates 91M megawatt-hours of net energy each year, which is well above what its current AI facilities require.

Next on the list is Illinois, hosting 5 cluster locations across the state. These facilities contain about 71K AI processing units with 7,600 H100 computing power. Illinois produces 177M MWh of electricity per year, and the AI infrastructure eats up less than 1% of it, leaving sizable capacity for future data center expansions.

Tennessee ranks fifth, showing the highest AI chip count in the country. The state’s 4 data center sites use more than 1.27M AI chips, nearly four times the average and more than any other state. These represent over 2.25M H100 equivalents, which is also the second-highest computing power nationally. Tennessee generates 77M megawatt-hours annually, and these facilities would consume about 30% of the local electricity output at full capacity.

Ohio takes sixth place with 4 GPU cluster locations, the same as Tennessee. The state has exactly half a million AI processing chips (40% above average) with a capacity of around 1.26M H100 units, the third-highest in the country. Ohio produces 133M MWh of electricity each year, and at maximum operation, the data centers here would use roughly 12% of that total.

In the seventh place is Indiana, where there are 3 cluster sites spread across the state. These facilities operate with more than 400K AI chips, equal to 135K H100 units in processing capability. Compared to neighboring Ohio, Indiana has less power at 90M megawatt-hours, which means the AI infrastructure can consume nearly half of the state’s electricity, the highest proportion in the country

Iowa holds eighth position with the same number of clusters as Indiana. The local data centers count 25K AI chips in total, with a computing power of 7,900 H100 units. This is the identical processing capacity to that of Illinois, meaning the chips stored here are much stronger. Iowa produces 69M MWh of electricity each year, leaving more than enough power for hyperscales.

Next up is Washington with 3 GPU cluster sites around the state. These locations are run with 15,400 AI chips that have 5,100 H100 computing power. Despite three different AI sites operating here, these data centers only consume under 1% of local power generation, which stands at 102M MWh of net energy per year.

New Mexico completes the top ten states with the most data centers and GPU clusters. Similar to the three states above, New Mexico also has a total of 3 AI clusters. Across them, there are 4,600 AI processing units stored, and their power reaches 4,200 H100 equivalents. New Mexico produces 39M megawatt-hours of electricity annually, and, like in Washington, the local AI computing accounts for only 1% of power usage, giving more room for potential expansions.

“Investment in AI data centers is accelerating rapidly, with projections showing over USD 500B in spending through 2030,” says Eric Winegar, Managing Partner at Texas Royalty Brokers. “The South has emerged as the preferred region for these facilities because of lower energy costs, available land, and fewer regulatory barriers compared to other states. As more clusters come online, states will need to expand their power generation capacity significantly. What we’re seeing now is just the beginning of a much larger infrastructure buildout.”