Major tech companies are facing mounting pressure over the massive data centers they are building to power artificial intelligence applications, amid accusations relating to transparency, water and energy consumption, and impact on local communities. In a notable move, Amazon announced that it no longer uses non-disclosure agreements (NDAs) in its dealings with government entities when seeking permits to build new data centers.

This announcement was made by Matt Garman, CEO of Amazon Web Services (AWS), in a lengthy post in which he attempted to respond to the wave of skepticism surrounding data centers and to defend the idea that they benefit the communities around them.

Why Are Confidentiality Agreements Part of the Problem?

Non-disclosure agreements represent a key element in the growing objections to data centers, as they allow companies to strike deals with local officials away from residents' eyes. Environmental activist Erin Brockovich pointed out that the most prominent complaint she hears relates to a lack of transparency, where projects are often announced after permits are secured, and local officials sign confidentiality agreements before neighbors even know the project exists.

These objections have led to tangible results; New York State announced a one-year freeze on granting permits for large data centers. According to Garman, more than 100 decisions to halt or restrict these projects are currently under consideration across various U.S. states. He warned that enacting these measures could cost the United States its position in the AI race, with consequences extending across generations.

Four "Myths" Garman Tries to Refute

Garman attempted to refute what he described as the four misconceptions about data centers:

Water consumption: Based on Amazon's own report, he claimed that direct water consumption in data centers does not exceed 0.5% of the country's total industrial water use—far less than what golf courses or almond farming consume.
Electricity prices: He said prices rose only in some states, while declining or growing slowly in others, attributing the increase to aging power grids that were not upgraded before demand surged.
Pollution: He objected to critics' focus on the maximum permitted emissions, stressing that data center generators remain idle 99.9% of the time and operate only about 10 hours a year for maintenance purposes.
Community benefit: Here he announced the abandonment of confidentiality agreements, adding that Amazon has injected more than a billion dollars over three years into the communities that host its data centers.

Counter-Figures Weaken the Arguments

However, these assurances face conflicting data. On the water front, such statements usually ignore the indirect consumption associated with electricity generation and chip manufacturing. Researchers have emphasized the need to study these centers' consumption independently, given the absence of any federal or local requirement obliging tech companies to disclose this data.

As for electricity, an independent watchdog recently indicated that data centers were the primary reason behind a 76% price increase on the largest power grid in the United States compared to the previous year. On the pollution file, a planned Amazon data center in Texas is permitted to emit 33 million tons of carbon dioxide annually—a figure exceeding that of any power plant in the country.

A Deeper Crisis of Trust

The question remains: are these steps enough to quell the skepticism? Dario Amodei, CEO of Anthropic, believes that opposition to artificial intelligence is at its core a "crisis of trust," as people assume that governments and tech companies are always planning to exploit them. Similarly, writers have noted that data center opponents' response to companies' arguments is often a single sentence: "I don't believe them."

This issue reveals a fundamental challenge facing the rapid expansion of AI infrastructure: offering figures and promises is not enough; rather, genuine trust must be built with local communities through real transparency and independently verifiable data.

✦ بقلم فريق دروب أيديا