By This Hour AI Desk
Amazon Web Services is attempting to answer a growing backlash against data-center construction with a pledge on secrecy: the company says it no longer uses nondisclosure agreements with government agencies involved in approving its projects.
The statement, made by AWS chief executive Matt Garman in a broader defense of the company’s data-center expansion, addresses one of the most politically sensitive parts of the debate. Opponents have argued that large facilities can arrive in communities after important decisions have already been made, leaving residents to confront the implications for water, electricity, air pollution and local public services with limited ability to scrutinize the process.
Garman’s position is that data centers bring important benefits and that several leading objections rest on an incomplete picture of their resource demands and effects. His argument reaches beyond disclosure to challenge claims about water consumption, household power bills, generator emissions and the returns communities receive.
But the account supplied does not establish how broadly the new NDA practice applies, when Amazon changed it, or whether it covers every phase of a project. It also leaves major disputes over the full environmental and grid effects of data centers unresolved. The central question is not only whether agencies may now disclose more, but whether communities will receive sufficiently complete and timely information to assess projects before approvals are settled.
A transparency pledge meets a deeper trust problem
Nondisclosure agreements have become a symbol of residents’ frustration with data-center siting. The concern described in the report is not simply that commercial information may be protected. It is that agreements involving local officials and developers can limit public awareness while permits, land arrangements and infrastructure commitments are taking shape.
Garman said Amazon has ended the use of such agreements with the government agencies it works with on projects. That is a concrete change in the company’s stated approach, if implemented as described. Government agencies are often central to decisions on permits and the infrastructure needed to support large computing facilities. Greater freedom for those agencies to discuss a proposed project could give the public more notice and allow local debate to occur earlier.
Still, the significance of the pledge depends on details not supplied in the report. It does not say whether Amazon will publish project-level information itself, whether agreements signed before the policy change remain in place, or whether contractors, utilities, landowners and other participants could still be bound by separate confidentiality provisions. Nor does it show what information agencies will choose or be able to release absent an NDA.
Those limitations matter because community opposition is frequently rooted in a broader lack of confidence, rather than in one contractual device alone. Residents may seek clear information on a facility’s projected electricity demand, water arrangements, backup-power equipment, anticipated construction impacts, tax arrangements and public commitments. Ending one form of confidentiality may improve the conditions for disclosure, but it does not by itself answer every question associated with a proposed site.
The report portrays the Amazon statement as a response to that wider skepticism. It also describes a climate in which officials and companies face organized resistance to new data centers. Garman said more than 100 moratoriums on data centers were under consideration around the United States, while New York had announced a one-year pause on permits for large facilities. Those measures point to a political obstacle for an industry seeking capacity for AI and other computing services.
Water figures capture only part of the resource debate
In making its case, Garman argued that direct water consumption by data centers accounts for 0.5% of industrial water use in the United States. He based that figure on an Amazon report and compared the sector’s direct consumption with that of other water-intensive activities. The framing is designed to counter the assertion that data centers are singularly large water users at a national scale.
Yet the scope of that calculation is critical. The figure is explicitly about direct data-center water use. The source material notes that such a measure may not include water associated with generating the electricity that powers servers, or with manufacturing the chips used inside them. Those are potentially material parts of the wider footprint. A narrow operational measure and a broader lifecycle measure can both be numerically accurate while describing different things.
Scientists cited in the supplied account have said independent study is needed to understand data centers’ water and energy use more fully. The reason given is that technology companies are not subject to federal or state requirements governing how they report this data. Without common reporting rules and independent assessment, comparisons across companies, regions and technical designs can be difficult.
That does not prove Amazon’s direct-use figure is wrong. It does mean the figure cannot settle the larger question of how much water a data center is responsible for consuming across its supply chain and power system. Local conditions further complicate the debate: a national share does not show how a particular project could affect a particular community’s water resources. The supplied material does not provide project-by-project consumption data, geographic analysis or a method for evaluating indirect use.
For local officials, the distinction is consequential. A nationwide percentage may inform discussion of the industry’s scale, but permitting decisions are often shaped by the constraints of a specific region. The relevant issue can be the terms of local availability and local demand, not merely the industry’s fraction of nationwide industrial consumption.
Electricity costs remain a contested explanation
Garman also disputed the claim that data centers are principally driving higher electricity rates. He said rates have risen in some states with substantial data-center development, but have fallen or increased more slowly in others. Where prices have climbed, he attributed the main cause to old grids that were not upgraded and expanded before demand arrived.
The argument places responsibility on longstanding infrastructure shortfalls rather than on data-center demand itself. That distinction has policy implications. If inadequate prior investment is the primary problem, the response could focus on transmission, generation and grid modernization. If new high-volume customers are a leading driver, regulators may instead face sharper questions about who pays for upgrades and whether ordinary customers bear costs connected to large facilities.
The supplied account presents a contrary finding from an independent watchdog, which reportedly identified data centers as the main factor in a 76% year-over-year price increase on the largest U.S. electrical grid. That finding and Garman’s account cannot simply be treated as interchangeable. They may rely on different geographies, periods, definitions of prices, or approaches to causation. The available record does not provide the underlying analysis needed to test those possibilities.
What can be said with confidence is narrower: there is a direct dispute over why electricity prices have risen in at least one major market. Amazon’s argument does not establish that data-center demand is immaterial, and the watchdog finding as summarized does not establish that aging infrastructure played no role. Grid costs can have multiple causes. The reporting supplied does not quantify their respective weight or explain how costs were allocated among customers.
That uncertainty makes transparency especially relevant. Communities evaluating a project need more than a general assurance that data centers are beneficial or a broad claim that the grid is old. They need a comprehensible account of the projected demand, planned infrastructure, likely costs and responsibility for paying them. The material provided does not show whether Amazon’s new approach to NDAs will yield that level of specificity.
Permitted emissions and actual generator use are different measures
Air pollution is another point of friction. Garman objected to criticism that focuses on the maximum emissions allowed by a data-center permit. He said backup generators almost never operate, describing them as idle 99.9% of the time and running roughly 10 hours a year, principally for required maintenance testing.
The report contrasts that contention with a planned Amazon data center in Texas permitted to release up to 33 million tons of carbon dioxide annually, a maximum described as exceeding that of any other U.S. power plant. The contrast is stark, but it does not amount to a direct measurement of real-world emissions. A permit’s maximum authorized emissions describes what a facility may be allowed to emit under stated conditions; it is not a record of what its generators have actually emitted.
Conversely, an assertion about typical generator run time does not establish actual emissions at every site or explain the circumstances in which generators could operate more extensively. The information supplied contains no operating logs, generator specifications, fuel-use records or emissions-monitoring results for the Texas project. It therefore cannot determine whether the permitted maximum is likely, unlikely or impossible to approach in practice.
This is not a minor technical distinction. Opponents may regard high permitted limits as evidence of risk and of the infrastructure a community is being asked to accept. Amazon may regard maximum-permit comparisons as misleading when equipment is intended for backup use. Both positions concern different measurements. A useful public evaluation would require both the permitted ceiling and transparent reporting of actual operations.
Community spending does not settle the local bargain
Amazon also presented its local investment as part of the answer to critics. Garman said the company had contributed more than $1 billion over the previous three years to U.S. communities where it has a meaningful data-center presence. The statement signals that Amazon wants the discussion to include payments and support associated with its facilities, rather than only their demands on infrastructure.
The supplied information does not break down that total by place, category, recipient or project. It does not say how the company defines a meaningful presence, how contributions compare with public costs, or how residents and local governments assess the balance. As a result, the figure establishes Amazon’s stated nationwide contribution but cannot demonstrate a net benefit for any particular host community.
That gap reflects the broader conflict. A company can point to aggregate investment and argue that it is building infrastructure essential to future computing. Residents can still seek proof that the benefits, costs and risks are fairly distributed where a site is proposed. Amazon’s claim that it has stopped using NDAs may help address a process concern, but the evidence provided does not show whether it will satisfy opponents or remedy concerns arising from earlier agreements.
This report has not been independently corroborated. It is based on a single source account of Garman’s statements and of competing claims about water, power prices and emissions. The available material supports reporting those positions and their limits, but not a final judgment on the full environmental, economic or community impact of Amazon’s data-center expansion.
For further context on this subject, see Manchester City case reaches reported verdict after years of financial-rule disputes.
Reporting notes
What is confirmed: Amazon says it changed its NDA practice and has contributed more than $1 billion to affected U.S. communities over three years.
Why this matters: The pledge addresses transparency concerns that have become central to opposition to large data-center projects.
What remains unclear: The scope and effects of the NDA policy, and the full indirect water, power-cost and emissions impacts, are not established by the supplied record. This report is based on one source and has not been independently corroborated.
