Amazon Drops NDAs in Data Center Deals as Hundreds of Moratoriums Target AI Infrastructure

Amazon has said it is ending its use of non-disclosure agreements when negotiating data center construction with local governments. The company is repeating a step taken by Microsoft, which dropped NDAs in such deals earlier this year. The stated reason is blunt: secrecy feeds public resistance to AI infrastructure. As a result of that resistance, hundreds of moratoriums have already been proposed or passed in the US, across a wide geography, from New York State to San Francisco.

Why NDAs Became Toxic

The playbook was familiar. A large operator approaches a municipality, signs an NDA, and discusses land, tax breaks, grid connection, and water supply. Officials are not allowed to tell residents which company is behind the project or what it has been promised. For the business, this keeps plans from competitors and keeps land prices from spiking. For local communities, it looks different: a decision about a huge consumer of electricity and water has been made, and no one invited them to the conversation.

By Amazon's own wording, this gap is what became the problem. When people learn about a project from a leak or from a meeting agenda item that no longer decides anything, the dispute is not about kilowatts and jobs but about trust. A moratorium becomes the simplest tool in that situation: it does not require anyone to understand the details of the project. It is enough to stop everything.

What the Numbers and Facts Show

The original TechCrunch piece gives few details, so we limit ourselves to what is confirmed:

• Amazon is dropping NDAs in data center negotiations with local governments.
• Microsoft took a similar step earlier in 2026.
• The number of proposed and enacted moratoriums on AI infrastructure runs into the hundreds.
• The moratoriums span different regions, from New York to San Francisco.
• The main driver of discontent, according to the publication, is the secrecy of negotiations.

The terms of the policy change have not been disclosed. It is unclear whether it covers negotiations already under way or only new ones, and what exceptions will remain for commercially sensitive data, such as capacity and launch timelines. These details determine whether the decision is a real change in practice or mostly a reputational gesture.

A Parallel Story: Agents and Credit Cards

The second part of the same newsletter item covers a different side of the industry. A growing number of startups are betting that consumers will agree to give AI agents access to their credit cards. An agent that not only picks a ticket or a product but also pays for it requires a different level of trust than a chatbot that gives advice. The available excerpt names no companies and gives no terms, so it is too early to judge specific products.

There is still a link between the two topics. Both AI infrastructure and payment agents come down to one question: is society ready to accept a technology when decisions around it are made behind closed doors? Data centers have already run into this at the level of municipal councils. Agents will run into it at the level of users' wallets.

Practical Takeaways for the Market

For data center operators and their contractors, the takeaway is simple: early openness is becoming part of the project budget. Building in time for public hearings and preparing clear data on energy and water consumption is cheaper than fighting a moratorium later. Municipalities should decide in advance which details of a deal are always published.

Investors and developers should track not only capacity and contracts but also local politics: the list of areas with active or pending moratoriums now directly affects site selection. Those building payment agents should draw the same lesson. Transparent spending limits, clear action logs, and simple reversal of transactions will matter more than a polished demonstration of autonomy.