Amazon Data Center NDAs Are Ending, but Trust Will Take More Than a Pledge

Amazon says it has stopped using nondisclosure agreements with government agencies on data center projects, despite previously relying on such arrangements. The change makes Amazon data center NDAs a public test of whether transparency can ease growing resistance to AI infrastructure.

AWS CEO Matt Garman announced the policy inside a broader defense of data centers published on October 2. Amazon also pledged more than $1 billion for host communities during the next five years. The company framed both commitments as answers to concerns about electricity, water, pollution, jobs, and closed-door negotiations.

The timing reveals the pressure behind that response. State officials are restricting secrecy agreements, federal lawmakers are seeking disclosures, and communities are debating construction moratoriums. Amazon is promising openness after confidentiality became evidence, for many residents, that developers and officials were excluding them from decisions.

What Amazon Actually Changed

Amazon’s most consequential commitment is not the new community fund. It is the decision to stop asking government agencies for secrecy.

Garman presented the change through Amazon’s new Data Center Commitment, a collection of operating and community principles. The company says it no longer uses nondisclosure agreements with government agencies involved in its projects. It also promises earlier information sharing, public open houses, and continued dialogue as projects advance.

A nondisclosure agreement, or NDA, limits what its signatories can reveal about a negotiation. In data center development, these agreements have sometimes covered a developer’s identity, proposed land use, utility requirements, tax arrangements, or other project details.

That matters because local officials often control zoning, permits, incentives, and infrastructure decisions. Residents cannot effectively question those choices if they do not know which company is negotiating or what resources its project requires.

The Amazon commitment does not explain when the company stopped using NDAs. It also does not define which organizations qualify as government agencies or how the policy applies to outside developers and consultants.

Those omissions do not make the pledge meaningless. They show why the next stage requires documented practices, not another statement of intent.

Amazon paired its transparency pledge with a substantial community program called Built Together. The company says it will add more than $1 billion in spending over five years across education, job training, energy affordability, and water conservation.

Communities will help select local priorities, according to Amazon. Planned initiatives include free community college access, efficiency upgrades for public buildings and homes, and expanded training centers near data center sites.

Amazon estimates that its education program will connect more than 300,000 students with free degree access. It also plans to operate or develop 25 modular training centers, including three existing facilities.

The company says each center can train approximately 2,000 to 4,000 people annually. By the end of 2028, Amazon expects the network to prepare as many as 100,000 workers each year.

Another program targets efficiency improvements for more than 300 schools and community buildings, plus over 30,000 homes. Amazon says participating households could save approximately $700 annually, though actual savings will depend on local conditions and completed upgrades.

These plans create tangible benchmarks. Communities can eventually count participating students, completed building upgrades, trained workers, awarded grants, and resulting savings.

Amazon also says it contributed more than $1 billion to communities with a meaningful data center presence during the previous three years. The new commitment therefore extends a spending pattern rather than establishing community investment for the first time.

The combination is deliberate. Ending Amazon data center NDAs addresses the process used to negotiate projects. The additional funding addresses the benefits residents receive after accepting them.

Yet the two measures solve different problems. Grants can support communities, but they cannot replace public access to information before binding decisions are made.

That distinction is central to the backlash. Residents are not only asking what Amazon will contribute. They also want meaningful influence over land, power, water, tax, and environmental decisions before construction begins.

The policy change creates the possibility of that influence. Whether communities receive it will depend on what Amazon discloses, when disclosure occurs, and which parties remain outside the promise.

Why Amazon Data Center NDAs Became a Liability

Secrecy moved from a routine development practice to a political threat against the infrastructure AWS needs for AI growth.

Companies defend some confidentiality as necessary during site selection. Potential locations, commercial negotiations, grid studies, and proprietary designs can carry competitive value before a project becomes public.

The industry’s case is straightforward. Developers may need to evaluate multiple sites without causing land speculation or exposing sensitive technical information. Local agencies may also prefer confidentiality during preliminary discussions that never produce a finished proposal.

However, broad agreements can extend far beyond legitimate trade secrets. They can stop officials from identifying a developer or acknowledging that negotiations exist. They can also restrict discussion while public agencies evaluate zoning, incentives, utility capacity, and community costs.

That imbalance has become harder to defend as data centers grow in size and resource demand. A project can affect transmission infrastructure, water systems, land use, public revenue, emergency services, and residential development.

A data center explainer found that NDAs may cover tax breaks, zoning, land use, and utility consumption. It also cited researchers who contacted 31 Virginia localities with proposed or operating data centers. Twenty-five reported having confidentiality agreements.

Amazon has participated in this system. Records reported from Indiana showed that officials signed NDAs connected to a proposed Amazon campus northwest of South Bend.

The proposed project carried an announced investment of $11 billion, making the public interest unusually significant. Officials were restricted from sharing nonpublic information related to Amazon while negotiations developed.

Pennsylvania records produced another example. An agreement involving Amazon defined confidential material as information clearly marked confidential or reasonably understood that way.

Such broad wording creates uncertainty for officials. When the boundary is unclear, the safest legal response may be to disclose less, even when residents have reasonable questions.

Critics argue that this arrangement reverses public accountability. Instead of answering freely to constituents, an official must first consider obligations created during private negotiations with a corporation.

The political response is spreading. Pennsylvania Governor Josh Shapiro issued an executive order restricting data center NDAs in August 2026. Other governors and legislatures have considered similar limits.

A state policy review identified NDA restrictions alongside energy fees, wage requirements, and revisions to data center incentives. The debate has moved beyond isolated local disputes.

Federal scrutiny is increasing as well. A bipartisan House proposal would restrict data center NDAs, while lawmakers have requested information from Amazon and other large technology companies.

Amazon’s change therefore arrives before governments completely remove the company’s choice. Voluntarily ending the agreements lets Amazon present itself as responsive while influencing how transparency rules develop.

That does not necessarily make the policy cynical. Companies often change practices when political, legal, and commercial pressures align. The important question is whether the resulting standard gives communities useful information at the right time.

Garman said more than 100 data center moratoriums were under consideration across the United States. A moratorium temporarily pauses approvals or construction while officials examine regulations, infrastructure capacity, or public concerns.

His warning places the NDA decision inside a larger strategic problem. AWS needs a continuing supply of data center capacity to support cloud services and expanding AI workloads.

Every delayed site can complicate power procurement, construction schedules, customer planning, and regional capacity. A collection of local disputes can therefore become an infrastructure bottleneck for a global cloud provider.

Garman argued that widespread pauses would weaken the United States in the international AI race. That framing raises the stakes, but it can also intensify suspicion.

Residents deciding whether to approve a local project are evaluating local consequences. Invoking national competitiveness does not answer questions about a specific substation, water source, tax agreement, or generator permit.

Amazon’s retreat from NDAs implicitly recognizes that persuasion cannot begin after those decisions are nearly complete. The company now needs public legitimacy early enough to keep projects moving.

Amazon’s Pledge Meets a Trust Deficit

The primary conflict is no longer Amazon versus misinformation. It is Amazon’s promises versus communities demanding verifiable evidence.

Garman’s essay disputed four common criticisms of data centers. He rejected claims that facilities consume excessive water, broadly increase electricity prices, create severe generator pollution, and provide little community benefit.

Amazon offered figures to support its case. The company says an average Amazon data center uses less than 13,000 gallons of water daily, equivalent to about 42 US households.

It also says its data centers were seven times more water-efficient than the industry average in 2025. Amazon reported reaching 75 percent of its goal to become water positive by 2030.

Water positive means replenishing more water than the company directly consumes in its operations. Amazon says more than 65 contracted projects should return over 8 billion gallons annually to communities.

Those figures are company claims, and their scope matters. Direct facility consumption does not necessarily include water used to generate electricity or manufacture chips and equipment.

The same accounting problem affects energy and emissions. Amazon says its facilities pay for the electricity and grid infrastructure needed to serve them. It has also pledged not to shift those costs to households and small businesses.

Garman argued that rising electricity rates often reflect old grids that lacked adequate investment before new demand arrived. Amazon says its long-term utility agreements can finance transmission lines, substations, and other improvements.

That explanation can be true while leaving legitimate local concerns unresolved. A data center may pay its contracted costs, yet its arrival can still change generation plans, transmission priorities, or regional demand forecasts.

Residents need access to utility filings, rate structures, cost-allocation methods, and load projections to evaluate those effects. Corporate summaries cannot substitute for regulatory evidence.

Backup generators present another disputed area. Amazon says generators at its data centers remain idle 99.9 percent of the time and operate roughly 10 hours annually, mainly for testing.

The company also promises that new sites will use generators meeting the US Environmental Protection Agency’s Tier 4 emissions standard, or an equivalent standard elsewhere.

Critics say permitted maximums, cumulative emissions, exceptional operating periods, and associated power plants still deserve scrutiny. A generator that rarely runs may pose limited annual risk, but communities need site-specific data to confirm that assumption.

Amazon says it will publish annual energy use, energy efficiency, water use, water efficiency, and carbon-free energy percentages. That is an important commitment because comparable reporting can turn broad promises into measurable performance.

Annual company totals, however, may not answer local questions. Residents need project-level information when water systems, utility territories, and environmental conditions differ sharply between regions.

Independent reporting has highlighted the gap between corporate framing and public acceptance. The original backlash coverage noted that researchers lack mandatory, standardized reporting across states and federal agencies.

That absence encourages dueling narratives. Companies select figures showing efficiency or economic value, while opponents highlight permits, maximum demands, or worst-case impacts.

Both approaches can obscure the actual performance of a particular project. Standardized, independently reviewed disclosures would give communities a stronger basis for comparison.

Amazon’s community benefits also require context. The company cites construction jobs, operational employment, local taxes, education programs, and support for public services.

In Mississippi’s Madison County, Amazon says two sites under construction employ more than 2,300 construction workers. It expects the completed data centers to employ more than 1,700 people.

In Georgia’s Newton County, the company reports more than 1,500 construction workers and expects 400 operational jobs. These figures illustrate why local officials may support development despite public objections.

Construction employment is temporary, while operational employment continues after completion. Communities should evaluate both separately, along with wages, procurement, housing effects, and the duration of promised programs.

Tax projections also require close examination. Amazon estimates that its St. Joseph County projects will produce more than $3 billion in county taxes over their relevant term.

Large totals can be meaningful, but residents need the underlying period, incentive structure, public infrastructure obligations, and comparison with alternative land uses. Transparency must extend to these assumptions.

The new fund drew a similarly mixed response. Independent community reporting presented the investment as Amazon’s response to intensifying political and consumer resistance.

Environmental organization Stand.earth praised the NDA decision while criticizing the broader package. It argued that the community spending remained small beside Amazon’s infrastructure investment and did not resolve major environmental concerns.

That response captures Amazon’s problem. Even concessions that critics welcome are filtered through years of opaque development and disputed impact claims.

Ending NDAs can remove one reason for distrust. It cannot instantly erase the distrust that secrecy helped create.

What the Promise Does Not Settle

Amazon’s policy leaves unresolved questions about intermediaries, timing, enforcement, and the exact information communities will receive.

The most immediate uncertainty concerns scope. Amazon says it no longer uses NDAs with government agencies it works with on projects.

Data center developments often involve landowners, utilities, engineering companies, economic development consultants, construction firms, and special-purpose entities. Amazon may not directly conduct every early negotiation.

A government official could therefore receive confidential information from an intermediary rather than Amazon. The pledge needs to explain whether Amazon will prohibit contractors and development partners from requesting equivalent secrecy.

This is not a theoretical distinction. Large projects frequently use subsidiaries or third-party developers during site selection and permitting.

Residents may know a project code name long before learning its intended operator. Ending only direct Amazon NDAs would leave a significant disclosure gap.

Timing creates a second problem. Amazon promises to share plans as early as possible, but that phrase has no fixed meaning.

The useful moment for disclosure comes before public agencies make difficult-to-reverse decisions. That can include land acquisition, rezoning, tax incentives, utility commitments, water arrangements, or transmission planning.

An open house held after those decisions would increase information without creating meaningful participation. Amazon should define milestones for disclosing its identity, expected load, water strategy, land needs, and incentive requests.

The third issue is documentation. A public commitment is easier to change than a contract, regulation, or legally enforceable community agreement.

Amazon has not identified an independent auditor for its transparency promise. It also has not described a complaint process for residents who believe a local project violated the policy.

A durable standard would include written disclosure schedules and accessible project records. It would also explain which data can remain confidential and why.

Trade secrets deserve protection, but the category should be narrow. A proprietary server design differs from a project’s expected electricity demand or proposed tax treatment.

Public agencies should not have to guess which information they can disclose. Clear definitions reduce that chilling effect and protect legitimate commercial information without concealing public consequences.

The fourth issue involves historical agreements. Amazon’s statement addresses current practice but does not say whether existing NDAs will be terminated, waived, or allowed to expire.

Previously signed agreements may still restrict officials in communities where projects remain under development. Releasing those parties would provide a measurable demonstration of the policy.

Amazon could publish the number of outstanding agreements, their locations, and the steps taken to remove restrictions. It could also disclose whether any ongoing project still operates under confidentiality arranged by a partner.

The fifth issue is independent measurement. Amazon’s environmental and economic claims mostly come from company data or commissioned analysis.

That does not make them false. It means communities need comparable evidence produced under transparent methods, particularly where water supplies or utility rates are contested.

Amazon’s planned annual reporting will help if it includes consistent definitions and local detail. Independent review would make those reports more credible.

A broad national average can hide the conditions that determine local impact. Cooling requirements differ by climate, while grid emissions depend on regional generation.

Water stress also varies by watershed. A gallon consumed in a water-abundant region does not create the same risk as a gallon consumed during drought.

The debate over Amazon data center NDAs therefore points toward a larger governance challenge. Transparency is not simply the absence of a confidentiality agreement.

Useful transparency provides specific, timely, understandable information before decisions become final. It also lets outsiders test company claims against public records and measured outcomes.

Amazon has opened that door, but the company has not yet shown how far it will open. The difference will determine whether the announcement rebuilds trust or becomes another disputed promise.

Three Signals Will Show Whether Amazon’s Policy Works

The next test is not whether Amazon repeats its pledge. It is whether projects become easier to inspect before local governments commit public resources.

The first signal is the treatment of existing and indirect confidentiality agreements. Amazon should clarify whether its policy applies to subsidiaries, consultants, utilities, developers, and economic development partners acting on its behalf.

Watch for local records showing that officials can identify Amazon early in negotiations. Also watch whether Amazon releases officials from older agreements covering projects that remain active.

Evidence of broad releases would strengthen the company’s claim. Continued secrecy through third parties would weaken it, even if Amazon itself signs no new document.

The second signal is project-level environmental and utility disclosure. Amazon has committed to annual reporting on energy, water, efficiency, and carbon-free power.

The critical question is whether that reporting lets residents evaluate a proposed facility in their own community. National totals and portfolio averages cannot settle local disputes.

Useful disclosures would include expected electricity demand, peak demand, planned water sources, cooling design, backup generation, and anticipated grid upgrades. They should also distinguish operating estimates from permitted maximums.

Regulators and utilities can then compare projections with actual performance. Independent researchers can evaluate whether local rate protections and water commitments work as promised.

If Amazon publishes consistent local metrics, ending NDAs will become part of a meaningful transparency system. If disclosures remain aggregated, the policy will remove one barrier while preserving the underlying information gap.

The third signal is what happens to local opposition and moratorium proposals. Garman’s warning about more than 100 possible pauses shows that permitting risk motivated the announcement.

A decline in moratorium activity would not automatically prove that communities trust Amazon. Legislative calendars, economic conditions, elections, and project changes can also influence those decisions.

The stronger indicator would be a change in how disputes unfold. Earlier disclosure should give residents more time to question proposals, seek independent advice, and negotiate enforceable protections.

Some projects may still face rejection. Transparency does not guarantee public approval, and Amazon should not treat disclosure as a tool for manufacturing consent.

Instead, transparent negotiations should produce clearer decisions. Communities that approve a site should understand its resource requirements, expected benefits, and enforceable safeguards.

Communities that reject one should be able to identify which unresolved costs or risks drove that choice. That outcome may slow some projects while making successful projects more durable.

Amazon also needs to separate criticism from misinformation. Garman described several objections as myths and warned about foreign efforts to slow US infrastructure development.

Foreign influence can exist without invalidating local concerns. Residents asking for utility records or water studies are participating in infrastructure governance, not necessarily opposing AI itself.

Framing skeptics as misinformed can undermine the listening process Amazon says it wants to create. The company will gain more credibility by answering precise questions with reproducible evidence.

The broader cloud industry will also respond. Google, Meta, Oracle, Microsoft, and specialized data center developers face many of the same political pressures.

Amazon’s announcement raises the cost of silence for those companies. Officials can now ask why another developer still requires confidentiality when AWS says it can proceed without it.

Competitors may adopt similar pledges, narrow their agreements, or support standardized disclosures. They may also argue that Amazon’s promise leaves exceptions that remain commercially necessary.

That comparison will help define the emerging industry standard. It could also move the debate from voluntary commitments toward binding state or federal rules.

AWS customers have a reason to pay attention as well. Data center permits and power access influence where cloud capacity becomes available and how quickly providers can deploy new infrastructure.

Delayed construction can affect regional capacity planning, especially as AI workloads demand more electricity and specialized hardware. Community relations have become an operational constraint, not a peripheral public relations issue.

Developers and enterprise buyers should therefore evaluate infrastructure claims with the same care they apply to security or reliability. Capacity plans depend on regulatory durability and public acceptance.

Amazon has made a clear concession by ending its stated use of government NDAs. The decision acknowledges that secrecy now threatens the speed and legitimacy of data center development.

The remaining question is whether Amazon will turn that concession into a verifiable operating standard. Watch the contracts, project disclosures, utility records, and timing of public participation.

If those practices change, Amazon’s policy could establish a new baseline for AI infrastructure development. If they do not, communities will see the announcement as another argument for mandatory disclosure.

The burden now rests with Amazon. Ending Amazon data center NDAs removed one obstacle to trust, but only early evidence and enforceable commitments can build the rest.