Amazon Web Services CEO Matt Garman has warned that growing opposition to data center construction across the United States could undermine the country’s position in the global artificial intelligence race, as communities push back against the electricity, water and environmental costs of the infrastructure boom.
In a more than 3,000-word blog post published Friday, Garman argued that a broad slowdown in data center construction would have consequences extending well beyond individual projects, potentially limiting the computing capacity needed to develop and deploy increasingly powerful AI systems.
“The consequences would last generations” if the United States halted the buildout, Garman wrote.
“There is urgency to this data center buildout because we aren’t the only country that sees the benefits of AI for the economy and national security, and the countries that lead in AI will shape it and get the most from it in the short and long run,” he said.
“The choices we make today will determine that outcome, and a lapse in our nation’s focus or resolve could put us behind, and potentially irreparably so.”
Garman said more than 100 data center moratoriums are being considered across the country. If enacted, he argued, they could constrain the infrastructure required to support AI development at precisely the moment when the U.S. is competing with other countries to expand computing capacity.
“As a country, we can’t afford to find ourselves in that position,” Garman wrote.
The warning reflects a widening conflict between the economic ambitions surrounding AI and the physical realities of building the infrastructure required to support it.
The technology industry has committed hundreds of billions of dollars to data centers, GPUs, power infrastructure and related equipment. But communities hosting or considering new facilities are increasingly questioning whether the economic benefits justify the demands placed on local electricity grids, water systems, roads and the environment.
For Amazon and other cloud providers, that resistance represents more than a permitting obstacle. Delays can push back the availability of computing capacity, increase construction costs and complicate long-term contracts with AI customers.
AI’s Infrastructure Problem
Garman compared the current data center expansion with the development of America’s modern transportation infrastructure after World War II.
He described the construction of highways as a transformation that reshaped the U.S. economy and way of life, arguing that data centers represent a comparable infrastructure buildout for the digital economy.
“Seventy years later, we’re in the midst of the largest infrastructure buildout since the highways, our digital infrastructure,” he wrote.
The comparison highlights an important feature of the AI economy: advances in software are ultimately constrained by physical infrastructure. AI models require enormous quantities of computing power, while that computing power requires data centers, electricity, cooling systems, semiconductor equipment, and increasingly large connections to the power grid.
Amazon is one of the companies at the center of that investment. AWS provides cloud computing infrastructure to companies developing and deploying AI systems, meaning growth in AI usage translates into demand for additional computing capacity.
Garman argued that data centers already support services ranging from online shopping and banking to stock trading, travel reservations, and entertainment. He also pointed to potential applications in healthcare, education, defense, intelligence, cybersecurity and satellite operations.
“AI will power the next generation of businesses, hospitals, schools, and other critical public services, as well as critical defense systems, intelligence analysis, cyber defense, and satellite operations,” Garman wrote.
He also argued that the industry will generate new skilled jobs associated with building and operating the infrastructure.
That broader economic argument is central to the industry’s response to local resistance. Cloud companies present data centers not simply as private facilities serving technology companies, but as foundational infrastructure for an economy that is becoming more dependent on computing.
The difficulty is that the costs are often concentrated locally while many of the benefits are distributed nationally or globally. A community may bear the burden of higher electricity demand, water consumption, construction traffic and changes to land use, while the resulting AI services and economic activity may accrue primarily to technology companies and their customers elsewhere.
That tension is helping drive the growing number of proposed moratoriums.
Amazon Offers Communities More Money
Garman acknowledged that some communities have legitimate concerns about data center development, particularly around electricity consumption, water use and pollution.
Amazon is attempting to address some of those concerns through additional investment. The company plans to invest more than $1 billion over the next five years in communities where it operates data centers, Garman said.
“This is not Amazon deciding what communities need, it is Amazon providing resources and letting communities decide,” he wrote.
The commitment represents an effort to make the local economic benefits of the infrastructure buildout more tangible. But it also highlights a major question surrounding the industry’s expansion: is community opposition primarily a matter of insufficient economic compensation, or do some communities simply not want the environmental and infrastructure costs associated with large-scale computing facilities?
The answer to this question is likely going to be demanded as AI data centers grow larger.
Modern facilities can require enormous amounts of electricity, creating pressure for new generation capacity and transmission infrastructure. In areas where grids are already constrained, data centers can compete with households and other industries for available power.
Water is another source of concern, particularly in regions where cooling systems could increase demand on scarce resources. The environmental impact can also vary substantially depending on how the electricity used by a facility is generated.
For cloud providers, therefore, securing land is only one part of the development challenge. They must also secure power, transmission capacity, water where necessary, permits and community acceptance. The result is that the physical expansion of AI can move considerably more slowly than the development of the software itself.
Garman also raised a geopolitical dimension, arguing that foreign governments may benefit from U.S. resistance to data center construction. He said there are “widespread reports of various countries intentionally seeding misinformation in the US about data centers to trick us into slowing down.”
That is a huge claim, but it also illustrates how technology companies are framing AI infrastructure as part of national competition rather than simply a commercial investment.
The United States and China are competing to develop advanced AI systems, while other countries are also attempting to build domestic computing capacity. Access to electricity, advanced chips, and data center infrastructure has consequently become part of the broader competition over AI capability.
For Amazon, maintaining rapid construction is also a commercial priority. AWS is competing with Microsoft Azure and Google Cloud to supply computing capacity to AI developers, while specialized data center operators and so-called neocloud companies are expanding to meet demand from companies that cannot secure sufficient capacity from traditional cloud providers.
That has provided an incentive for technology companies to emphasize the costs of delay. But the industry’s infrastructure push also faces a financial constraint. Data centers require huge upfront investments, and developers now need to secure long-term power and customer commitments before construction. Higher interest rates and rising equipment and construction costs can make projects less attractive if permitting or community opposition delays completion.
The debate is therefore moving beyond whether AI will transform the economy to a more immediate question of who should pay for the infrastructure required to make that transformation possible. Amazon’s $1 billion community investment may help address some local concerns, but it does not resolve the larger trade-off. The United States can build more computing capacity and potentially accelerate AI development, but doing so requires communities to absorb at least some of the associated costs.
