Microsoft's Datacenter Expansion Sparks Water Usage Concerns Amid AI Boom

Microsoft is forging ahead with plans to build 15 new server farms in Wisconsin worth over $13 billion, even as internal documents reveal the company's water consumption could triple by 2030. The expansion comes as part of Microsoft's aggressive push into AI infrastructure, with the tech giant also announcing $30 billion in UK datacenter investments and $10 billion for Portugal through 2028.

Environmental Impact of AI Infrastructure

The water usage concerns are particularly acute given Microsoft's previous admission that its emissions have increased by approximately 30 percent from 2020 levels, despite pledging to become carbon-negative by 2030. The New York Times obtained internal Microsoft documents showing forecasts that annual water needs for roughly 100 of its global campuses could more than triple to 28 billion liters by 2030 compared to 2020 levels.

Microsoft has since updated its projections, claiming it now expects to use about 18 billion liters of water in 2030—still a 150 percent increase from 2020 levels. The company launched its Community-First AI Infrastructure initiative earlier this month, promising greater transparency about water usage and commitments to replenish more water than it draws.

Local Opposition and Community Impact

Local officials in Mount Pleasant, Wisconsin, approved Microsoft's plans this week despite some pushback from residents concerned about the environmental impact. The new facilities will be built northwest of Microsoft's existing AI datacenter campus, which the company bills as "the world's most powerful AI datacenter" and expects to come online in 2026.

The expansion has raised questions about the true costs of AI infrastructure development. More than 230 organizations across America signed a letter last year calling for a moratorium on datacenter construction, citing environmental and social threats. Critics argue that the current build boom represents a massive strain on local resources, particularly water supplies in drought-prone regions.

Industry-Wide Concerns

Microsoft isn't alone in its datacenter expansion. The broader tech industry is experiencing a frenzied pace of construction to support AI workloads, with similar criticisms directed at other cloud providers and datacenter operators. The Bank of England's Financial Policy Committee warned last year that the current AI investment bubble could burst, potentially slowing the datacenter construction boom.

Transparency and Accountability Issues

The lack of transparency around datacenter operations has become a growing concern. Local officials are often required to sign non-disclosure agreements with tech companies, making it difficult for communities to fully understand the environmental and economic impacts of these massive facilities. Microsoft's vice chair and president Brad Smith has promised greater local transparency, stating that people deserve to know how much water Microsoft datacenters use.

The AI Infrastructure Arms Race

Microsoft's expansion plans reflect the broader competition in AI infrastructure. The company's $30 billion UK investment and $10 billion Portugal commitment are part of a strategy to more than double its datacenter capacity across 16 European countries by 2027. This aggressive expansion is driven by the massive computational requirements of AI training and inference, which demand specialized hardware and cooling infrastructure.

The environmental costs of this AI arms race extend beyond water usage to include greenhouse gas emissions and energy consumption. As AI models become larger and more complex, the infrastructure required to train and deploy them continues to grow, creating a feedback loop of increasing resource demands.

The current trajectory suggests that unless significant technological breakthroughs in cooling efficiency or alternative energy sources emerge, the environmental impact of AI infrastructure will continue to be a major concern for communities hosting these facilities and for the planet as a whole.