AI megaprojects come with hidden costs

While Tesla’s stock has seen a sharp decline in recent weeks, Elon Musk has quietly been working to expand his influence in a different area of tech: AI supercomputing. Recent reports have revealed that Musk’s AI startup, xAI, has been quietly building a new data center in Atlanta. The facility will house around 12,000 Nvidia graphics processing units (GPUs), crucial chips that power AI computation.

While the Atlanta data center is modest compared to Musk’s supercomputer project in Memphis, nicknamed “Colossus,” which boasts 100,000 GPUs, its construction is part of a broader trend in the tech world that has been advancing largely under the radar. 

Last December, Meta announced its plans to build a $10 billion data center in Richland Parish, Louisiana. In February, OpenAI began constructing a facility in Abilene, Texas, while scoping out potential sites in 16 other states, as they seek to invest $500 billion in AI infrastructure over the next four years under the Trump Administration’s Stargate project. While megaprojects like Colossus and Stargate have captured national attention—especially as Musk trumpeted his goal of creating “the most powerful” AI supercomputer in the world—this broader wave of data center construction has largely proceeded under the radar, with minimal public oversight, particularly from the communities that will host these centers. 

This silent expansion underscores an important issue: We cannot allow Big Tech to have unchecked control over the development of AI infrastructure, especially when there is a lack of transparency and limited public scrutiny.

The problem with unchecked AI expansion is multifaceted. While tech companies promise their data centers will rejuvenate local economies with jobs and tax revenue, it is not entirely clear that the risks outweigh the rewards. 

First, the AI industry is an environmental crisis waiting to happen. Data centers typically consume an immense amount of resources, particularly water. Liquid-based cooling systems are essential to prevent overheating in the massive networks of servers that power these centers, and this process requires vast quantities of water. The Colossus project in Memphis alone is estimated to consume one million gallons of water a day, according to public utilities companies in the city. 

And it’s not just Memphis—with an estimated 50% of the world’s population projected to live in water-stressed areas by 2025, the growing demand for water in AI data centers presents a real threat to local communities. In Memphis, activists are already sounding the alarm about the impact on local water supplies, especially in a region with a history of arsenic contamination in drinking water. xAI has reached an agreement with the City of Memphis to build a recycled wastewater facility for the data center’s cooling needs, aimed at alleviating some pressure on local water sources. 

However, the company has not been transparent with the local community regarding its plans for the facility, which is expected to reduce the strain on the Memphis Sand aquifer by only 9%—even as xAI announced plans to expand its supercomputing operation to one million GPU’s more than 10 times its current capacity.

The environmental impact extends beyond water usage. The supercomputing industry is highly carbon-intensive. According to a report by Morgan Stanley, data centers worldwide are projected to emit 2.5 billion tons of CO2 by 2030. Many of these data centers are being built in the U.S. South, a region already vulnerable to rising temperatures. The carbon emissions from these projects will only worsen the heat risks in areas that are already hot. The effects of extreme heat are not felt equally. 

In Atlanta, for instance, where some neighborhoods were redlined in the 20th century, residents in predominantly Black communities—who often lack access to air-conditioning and shaded green spaces—are already disproportionately at risk for heat-related illness and death. The continued expansion of data centers and the heat generated by their carbon emissions will only make these inequities worse. While Musk’s xAI may tout its role in advancing AI innovation, these environmental tolls cannot be ignored.

Moreover, it’s not even clear that we need this kind of computing power. Earlier this year, the launch of DeepSeek, a more resource-efficient AI model developed in China, shocked the tech industry. DeepSeek’s breakthrough suggested that the future of AI could be far less resource-hungry than what Musk and others are pushing. Since DeepSeek’s success, tech giants like Microsoft have canceled data center leases in the U.S., signaling that perhaps the need for supercomputing power is not as pressing as Musk’s ambitions would have us believe. 

This raises the question: Are these massive data centers even necessary, or are they just part of a generative AI hype bubble? And will the communities hosting these data centers be left to foot the bill once the bubble bursts?

Of course, many will argue that the global AI race is too important to let these concerns get in the way. But we cannot allow Silicon Valley’s “move fast and break things” mentality to dictate how our cities evolve and how resources are allocated. Urban leaders must step up to demand oversight and accountability in the development of these data centers.

AI is undeniably vital to the future of human progress, but it’s too important to be left in the hands of unaccountable tech oligarchs. Atlanta, Memphis, and other cities across the country need to assert control over the expansion of data centers and insist on transparent, responsible development. 

https://www.fastcompany.com/91307367/ai-megaprojects-come-with-hidden-costs?partner=rss&utm_source=rss&utm_medium=feed&utm_campaign=rss+fastcompany&utm_content=rss