Reports related to this article:
Project(s): View 5 related projects in PECWeb
Plant(s): View 2 related plants in PECWeb
Released September 09, 2025 | SUGAR LAND
en
Researched by Industrial Info Resources (Sugar Land, Texas)--AI and data center infrastructure don't come cheap, industry experts agreed at a recent conference at Rice University's Baker Institute for Public Policy.
Speaking at the "Powering the Energy Future" conference on September 3, executives from Amazon Web Services (AWS) (Seattle, Washington) and energy-focused data and analytics provider Enverus, as well as other experts, discussed the rise of artificial intelligence (AI) and the capital-intensive nature of building out data center capacity.
"AI is a capital-intensive business, and it's not only expensive to grow data center capacity ... it's expensive to maintain data center capacity because chips have life cycles," said Carson Keal, lead analyst, AI infrastructure and co-located generation strategies for Enverus, in a session titled "Data Center Development and the Future of Power Demand."
Keal said the five "big tech" companies--Alphabet (Google) (Mountain View, California), Amazon (Seattle, Washington), Apple Incorporated (Cupertino, California), Meta Platforms Incorporated (Facebook) (Menlo Park, California) and Microsoft (Redmond, Washington)--would need to spend roughly 141% of the capital they're projected to spend in 2026 and 187% in 2027, for the Electric Reliability Council of Texas (ERCOT) grid and the PJM Interconnection to meet the low end of their data center capacity projections for those years.
For more information on power demand for the two grids, see August 1, 2025, article - EIA: ERCOT, PJM Regions Hotspots for Rapid Growth in Power Demand.
IIR Energy last month noted more than 1 GW worth of new data centers were commissioned in the U.S. in the second quarter. For more information, see August 7, 2025, article - More Than 1 GW of New Data Centers Commissioned in the U.S. in Q2.
Keal also discussed third-party data center developers seeking to build new facilities, wondering if hyperscalers actually need a lot of their help: "It seems pretty clear to me that they [hyperscalers] own a lot of the infrastructure and the land necessary to build a lot of the products that they will be looking to develop next."
In the mid-2000s, social media sparked a rise in data center demand to provide cloud services for hosting and storing information, which remained all the way through 2020 or 2021, before a shift to configure data centers for AI applications and the semiconductor chips that power the technology.
Noting chips have an actual shelf life of four or five years for a lot of the needed technology, companies will need to reinvest capital just to maintain effective, existing data center capacity. Companies will be growing computing power exponentially throughout the process, "and that's because of the efficiency of new chips that you'll be installing in the existing facilities."
Industrial Info is tracking $148 billion worth of capital data center projects under construction across the U.S. Subscribers to Industrial Info's Global Market Intelligence (GMI) Industrial Manufacturing Project Database can click here for a full list.
Keal pointed to New Albany, Ohio, northeast of Columbus, as one of the hottest spots in the U.S. for new data centers. Google expects to finish a fifth-phase expansion of its New Albany campus toward the end of the current quarter, while a sixth-phase expansion could wrap up next spring. Google also is considering a seventh phase that could begin construction next summer. Subscribers can learn more from detailed reports on the fifth, sixth and proposed seventh phases.
He also mentioned Meta's nearly $10 billion, 2-GW AI data center campus in Louisiana, which would require enormous amounts of energy. Construction of the first of nine proposed buildings kicked off in December 2024. For more information, see August 27, 2025, article - Louisiana Greenlights Massive Power Projects for Meta Data Center.
Subscribers can click here for project information.
However, according to news media reports, some Louisiana residents and environmental advocates have pushed back on the project, citing its potential impacts on the state's energy grid, water use and the overall environment, in addition to fears that rate increases will come from the additional capacity. This seems to be a potential issue all across the country. Utilities have begun to request guidelines requiring large users such as data centers to pay for a minimum percentage of the energy that they're subscribed to, even if they use less.
Media reports also indicate the city council of Tucson, Arizona unanimously voted against the construction of a proposed hyperscale data center for Amazon Web Services, following widespread public opposition. For more information, see August 11, 2025, article - Tucson City Council Says 'No' to Massive Amazon Data Center.
Later in the conference session on data center development, Saurabh Chaugule, senior manager strategic business development, energy & utilities, Amazon Web Services, said AI is a "huge, mega trend" that's driving the growth in energy demand.
Chaugule said Amazon expects a capital budget of around $100 billion to more than $118 billion for this year, compared with about $84 billion in 2024, which the company has indicated is primarily driven by investments in AI infrastructure. He did not break the figures out into specific categories or domestic/global locations.
AI and data center infrastructure construction is expensive, and "it's hard to get that far into a project such that you wouldn't be able to take it to completion," and then have abandoned infrastructure. "I think the withdrawals will happen much earlier than that."
"We have a pipeline of projects that's going to happen, which are going to get deployed over a number of years," Chaugule said, as the company is matching the project timelines with the technology available to deploy, from an energy efficiency standpoint.
Some large infrastructure investment is "still on the sidelines, not getting invested," however, as stable industrial policy is important.
Another challenge facing the rapid buildout of data centers is a backlog in interconnection queues across the country. "We are helping the utilities run these interconnection studies faster so they can actually process the queues faster, because the queues are about three to seven years long," to get a new project interconnected to the grids in different parts of the U.S.
Amazon is the world's largest corporate purchaser of renewable energy, reaching about 34 GW worth of power purchase agreements in the five years from 2019 through 2024. "We are looking at variety of different power generation technologies, including gas power plants with carbon capture, geothermal, you name it," Chaugule said.
Keal said Enverus' clients developing data center infrastructure have indicated that they are looking to collocate power generation on site to minimize impacts on the power grid. "I think owning your own generation and building out your load with that generation is going to be the solution in a lot of cases."
Another session at the conference specifically discussed nuclear power as a main source of generation poised to support data center-related load growth now and in the future; for more information, see September 8, 2025, article - Conference: Nuclear is in Great Shape to Power U.S., Texas Grids.
Industrial Info Resources (IIR) is the leading provider of industrial market intelligence. Since 1983, IIR has provided comprehensive research, news and analysis on the industrial process, manufacturing and energy related industries. IIR's Global Market Intelligence (GMI) platform helps companies identify and pursue trends across multiple markets with access to real, qualified and validated plant and project opportunities. Across the world, IIR is tracking more than 200,000 current and future projects worth $17.8 trillion (USD).
Speaking at the "Powering the Energy Future" conference on September 3, executives from Amazon Web Services (AWS) (Seattle, Washington) and energy-focused data and analytics provider Enverus, as well as other experts, discussed the rise of artificial intelligence (AI) and the capital-intensive nature of building out data center capacity.
"AI is a capital-intensive business, and it's not only expensive to grow data center capacity ... it's expensive to maintain data center capacity because chips have life cycles," said Carson Keal, lead analyst, AI infrastructure and co-located generation strategies for Enverus, in a session titled "Data Center Development and the Future of Power Demand."
Keal said the five "big tech" companies--Alphabet (Google) (Mountain View, California), Amazon (Seattle, Washington), Apple Incorporated (Cupertino, California), Meta Platforms Incorporated (Facebook) (Menlo Park, California) and Microsoft (Redmond, Washington)--would need to spend roughly 141% of the capital they're projected to spend in 2026 and 187% in 2027, for the Electric Reliability Council of Texas (ERCOT) grid and the PJM Interconnection to meet the low end of their data center capacity projections for those years.
For more information on power demand for the two grids, see August 1, 2025, article - EIA: ERCOT, PJM Regions Hotspots for Rapid Growth in Power Demand.
IIR Energy last month noted more than 1 GW worth of new data centers were commissioned in the U.S. in the second quarter. For more information, see August 7, 2025, article - More Than 1 GW of New Data Centers Commissioned in the U.S. in Q2.
Keal also discussed third-party data center developers seeking to build new facilities, wondering if hyperscalers actually need a lot of their help: "It seems pretty clear to me that they [hyperscalers] own a lot of the infrastructure and the land necessary to build a lot of the products that they will be looking to develop next."
In the mid-2000s, social media sparked a rise in data center demand to provide cloud services for hosting and storing information, which remained all the way through 2020 or 2021, before a shift to configure data centers for AI applications and the semiconductor chips that power the technology.
Noting chips have an actual shelf life of four or five years for a lot of the needed technology, companies will need to reinvest capital just to maintain effective, existing data center capacity. Companies will be growing computing power exponentially throughout the process, "and that's because of the efficiency of new chips that you'll be installing in the existing facilities."
Industrial Info is tracking $148 billion worth of capital data center projects under construction across the U.S. Subscribers to Industrial Info's Global Market Intelligence (GMI) Industrial Manufacturing Project Database can click here for a full list.
Keal pointed to New Albany, Ohio, northeast of Columbus, as one of the hottest spots in the U.S. for new data centers. Google expects to finish a fifth-phase expansion of its New Albany campus toward the end of the current quarter, while a sixth-phase expansion could wrap up next spring. Google also is considering a seventh phase that could begin construction next summer. Subscribers can learn more from detailed reports on the fifth, sixth and proposed seventh phases.
He also mentioned Meta's nearly $10 billion, 2-GW AI data center campus in Louisiana, which would require enormous amounts of energy. Construction of the first of nine proposed buildings kicked off in December 2024. For more information, see August 27, 2025, article - Louisiana Greenlights Massive Power Projects for Meta Data Center.
Subscribers can click here for project information.
However, according to news media reports, some Louisiana residents and environmental advocates have pushed back on the project, citing its potential impacts on the state's energy grid, water use and the overall environment, in addition to fears that rate increases will come from the additional capacity. This seems to be a potential issue all across the country. Utilities have begun to request guidelines requiring large users such as data centers to pay for a minimum percentage of the energy that they're subscribed to, even if they use less.
Media reports also indicate the city council of Tucson, Arizona unanimously voted against the construction of a proposed hyperscale data center for Amazon Web Services, following widespread public opposition. For more information, see August 11, 2025, article - Tucson City Council Says 'No' to Massive Amazon Data Center.
Later in the conference session on data center development, Saurabh Chaugule, senior manager strategic business development, energy & utilities, Amazon Web Services, said AI is a "huge, mega trend" that's driving the growth in energy demand.
Chaugule said Amazon expects a capital budget of around $100 billion to more than $118 billion for this year, compared with about $84 billion in 2024, which the company has indicated is primarily driven by investments in AI infrastructure. He did not break the figures out into specific categories or domestic/global locations.
AI and data center infrastructure construction is expensive, and "it's hard to get that far into a project such that you wouldn't be able to take it to completion," and then have abandoned infrastructure. "I think the withdrawals will happen much earlier than that."
"We have a pipeline of projects that's going to happen, which are going to get deployed over a number of years," Chaugule said, as the company is matching the project timelines with the technology available to deploy, from an energy efficiency standpoint.
Some large infrastructure investment is "still on the sidelines, not getting invested," however, as stable industrial policy is important.
Another challenge facing the rapid buildout of data centers is a backlog in interconnection queues across the country. "We are helping the utilities run these interconnection studies faster so they can actually process the queues faster, because the queues are about three to seven years long," to get a new project interconnected to the grids in different parts of the U.S.
Amazon is the world's largest corporate purchaser of renewable energy, reaching about 34 GW worth of power purchase agreements in the five years from 2019 through 2024. "We are looking at variety of different power generation technologies, including gas power plants with carbon capture, geothermal, you name it," Chaugule said.
Keal said Enverus' clients developing data center infrastructure have indicated that they are looking to collocate power generation on site to minimize impacts on the power grid. "I think owning your own generation and building out your load with that generation is going to be the solution in a lot of cases."
Another session at the conference specifically discussed nuclear power as a main source of generation poised to support data center-related load growth now and in the future; for more information, see September 8, 2025, article - Conference: Nuclear is in Great Shape to Power U.S., Texas Grids.
Industrial Info Resources (IIR) is the leading provider of industrial market intelligence. Since 1983, IIR has provided comprehensive research, news and analysis on the industrial process, manufacturing and energy related industries. IIR's Global Market Intelligence (GMI) platform helps companies identify and pursue trends across multiple markets with access to real, qualified and validated plant and project opportunities. Across the world, IIR is tracking more than 200,000 current and future projects worth $17.8 trillion (USD).
