The backlash to America’s data-center buildout has grown loud enough that Amazon is changing the terms of the conversation. On October 2, AWS chief executive Matt Garman announced a program called “Built Together”: more than $1 billion over five years for the communities that host Amazon’s data centers, to be spent on free community college, job training, and home energy retrofits. The company is also ending a practice that had become a symbol of local distrust. It will no longer ask local governments to sign nondisclosure agreements.
The announcement was cast as an answer to what Garman called “misinformation and outright lies.” He said more than 100 proposals to ban or restrict data centers are pending across the United States. That tally explains why the company felt the need to speak in public. When the conversation about a buildout is carried on through proposed bans, the builder has to make its case to residents rather than through contracts.
The buildout that Garman is defending is the physical backbone of the AI boom. The models behind the industry’s newest products run in rows of servers that need power, water for cooling, and land close enough to fiber to be useful. Demand for that capacity has grown faster than anyone can build it, which has turned data centers from an unglamorous piece of corporate real estate into a scarce resource, and turned the towns that host them into a front line.
The money is aimed squarely at the objections that have slowed projects. Free community college and job training answer the complaint that data centers bring few lasting local jobs once construction ends. Home energy retrofits answer the complaint that the facilities strain the grid and push costs onto households. By tying the spending to the specific towns that host the servers, Amazon is trying to give neighbors a direct stake in a project that otherwise offers them little.
The NDA reversal is the more striking concession. For years, the deals that brought data centers to a town were wrapped in confidentiality, which left residents and sometimes even elected officials unsure what had been agreed to. Dropping the requirement is a bet that openness will cost less than suspicion. Amazon is also committing to publish its electricity and water use every year, and to pay for the new electricity its facilities draw so that local residents do not see their rates rise.
The stakes for Amazon are enormous. The company expects to spend about $220 billion on capital expenditures this year, a figure that now exceeds its own operating cash flow. Most of that money flows into data centers and the chips and power that fill them. A buildout on that scale cannot proceed if it has to fight a permitting fight in every town where it wants to break ground, which is what the hundred-plus pending bans threaten.
Amazon is not alone in the squeeze. Microsoft, Google, and Meta are building at a similar pace, and the same objections have followed them. Power has become the binding constraint on the whole industry. Local grids, water supplies, and the patience of nearby residents are all finite, and the companies that need them have spent the past two years learning that permits are harder to win than chips are to buy.
The objections the company faces are not hard to summarize. Residents worry about noise, water use, and the strain on local grids. Elected officials worry about the tax breaks and secrecy that have accompanied past deals. Environmental groups worry about the carbon cost of a buildout measured in gigawatts. None of these are answered by a single payment, which is why the program spreads its money across schools, training, and home efficiency rather than handing out checks.
The program also answers a structural complaint. Data centers often pay little in property tax relative to the value they create elsewhere, and the jobs they generate are mostly temporary construction work. Community college funding and energy retrofits are an attempt to leave something behind after the cranes leave, the kind of durable benefit a town can point to when the next permit comes up for a vote.
Garman’s framing places the blame on bad information rather than on the projects themselves, and Amazon does not pretend otherwise. The question the program leaves open is whether money and disclosure can change minds faster than the objections form. A billion dollars spread over five years and many towns is real, but it is also a small share of what the company will spend on the facilities themselves.
The test will come in the towns that are still undecided. For them, the offer is now on the table in writing: no secrecy, published numbers on power and water, and money for the schools and the homes around the fence line. Whether that changes a vote will be the first real measure of whether the program is a strategy or a gesture.
