A couple weeks ago Microsoft announced something unprecedented: it would reject local tax breaks for new AI data centers, pay full electricity costs, and fund grid upgrades to prevent raising residents’ utility bills.1 This is a pivoting moment in the development of infrastructure for AI as it marks the end of a two-decade playbook where cities competed to give tech companies billions in incentives. Are we witnessing the dawn of a more responsible development in the midst of the AI era? Or is this just an isolated event to momentarily appease the masses?

The Current Model: Cities Paid Tech

The era of AI brought with it a massive investment in data centers needed to power the LLM models we use every day, we’re talking trillions and trillions of dollars. In 2025 alone, the US spent $425 Billion dollars on data centers.2 However, this investment was at the cost of the taxpayer. Between 2015 and 2023, states handed out massive tax breaks to attract data centers. Let’s take a quick look at some examples of recent years:

For example, Microsoft alone received $333 million in Washington sales tax exemptions. Virginia gave tech companies over $730 million in fiscal year 2024.3 Amazon’s Ohio facility secured $101 million in combined state and local incentives4.These are some examples of development costs billed to the taxpayer, but at least 42 states offered data center-specific tax packages, treating these projects as guaranteed economic wins.

Power to the People: Communities Said No

That era just ended. By 2025, communities across America started blocking projects at an accelerating rate. In 2023, only 2 data center projects were cancelled due to local opposition. By 2024, that number jumped to 6. And in 2025 25 projects were cancelled, what’s notable of 2025 is that 21 of those cancellations happened in the second half of the year.5

Moreover, another $64 billion worth of projects were blocked or delayed6, and while these numbers are literally drops in the grand ocean that’s been Big Tech investment in the last 10 years, it signals the early stages of a movement that is looking to have a more conscious development of AI infrastructure.

Microsoft learned this the hard way. In October 2025, it cancelled a data center in Caledonia, Wisconsin after facing community backlash7. Later, in December, a hearing to discuss the rezoning of a property owned by Microsoft in Kent County, Michigan was cancelled when hundreds of residents overwhelmed the township hall, forcing Microsoft to pause the rezoning process.8

The Hidden Trigger: Water

The reason wasn’t anti-tech sentiment or even electricity bills alone. Analysis of contested projects revealed water consumption was cited in over 40% of opposition cases, by far the single biggest driver of community resistance.9

The scale is staggering. In 2022, Google consumed 19.7M liters of water (or 5.2 billion gallons of water), in 2024, the consumption rose to 29.4M liters of water (or 7.7 billion gallons of water).

Some facilities consume over 25% of local community water supplies. Lawrence Berkeley National Laboratory projects U.S. data centers consumed 17 billion gallons directly for cooling in 2023, with projections to double or even quadruple by 2028.10

What’s the reason behind this enormous consumption of water? Well, that’s because AI workloads require massive cooling infrastructure because GPUs generate 5-10x more heat than traditional servers. Water gets pumped through cooling systems, absorbs the heat, and about 80% evaporates as steam. Unlike electricity, which can be generated through new power plants or accelerating nuclear energy production, water in drought-prone regions cannot be manufactured.

Why This Matters

I really hope that Microsoft’s January 2026 pledge isn’t just an altruism stint, I hope this turns into Big Tech recognizing that bargaining power has shifted. Tech companies need specific infrastructure access more than cities need their promised jobs, especially if we take into account that a facility consumes resources equivalent to thousands of homes but creates only a few hundred (at best) permanent positions.11 We’re starting to observe how communities realized that the economic math no longer favors blanket incentives, and we’re also starting to see how Big Tech is shifting to survival strategies for an industry that in the future should no longer externalize infrastructure costs.

The playbook going forward won’t resemble Amazon’s HQ2 competition, where cities were throwing blank checks to Amazon in hopes of being chosen as the next Big Tech hub. Electricity can be solved with grid investments and nuclear deals, but water scarcity has no technological quick fix, I just hope that we as a community can start making some noise and demand more responsible AI Infrastructure development before the damage is done and it’s too late.

As always, thanks for reading! ✌️