The artificial intelligence infrastructure boom is creating an unprecedented strain on local water and power resources, forcing lawmakers and communities to demand that tech companies, not taxpayers, bear the costs of their expansion. As Nvidia reports record earnings driven by AI demand, Congress is preparing legislation to address the hidden environmental and financial toll of data center growth, while major tech firms race to secure nuclear power sources to fuel their computing ambitions.

Why Are Data Centers Consuming So Much Water?

Data centers require enormous amounts of water to cool their computer servers and to generate the electricity that powers them. The scale of this consumption has grown dramatically. U.S. data centers directly consumed 21 billion liters of water in 2014, but by 2023, that figure had jumped to 66 billion liters, according to a 2024 Berkeley Lab report. The trend is accelerating; hyperscale data centers, which are the massive facilities built by companies like Meta and Microsoft, are expected to consume between 60 billion and 124 billion liters of water by 2028.

The problem is that these costs often fall on local communities. When a data center connects to a public water system or requires infrastructure upgrades, residents can end up paying higher water bills to cover those expenses. Public concern about this issue has reached a tipping point. A Gallup poll conducted in April found that 50% of people opposed to data center construction cited the impact on local resources, including water, energy, farmland, and forests.

What Is Congress Doing to Protect Communities?

The House Committee on Energy and Commerce is moving to address the problem through legislation. The committee plans to hold a hearing on September 3 to discuss a draft of the Water Cost Accountability Act, which would ensure that communities do not bear the costs of data center water consumption or infrastructure upgrades.

The proposed bill would allow states receiving grants from State Revolving Loan Funds to prohibit public water systems from charging customers additional costs when data centers connect to the public water system or when infrastructure must be expanded. Instead, those costs would have to be recovered directly from the data center's owner or operator. The Environmental Protection Agency would also be required to provide Congress with a report on water sources used by large data centers, defined as those consuming more than 200,000 gallons of water per day on average.

"While data centers consume a fraction of the water of many other industries and have proven that new technologies are actually reducing new water consumption, nonetheless, Americans should have confidence that data center developers are acting responsibly with local water supplies," said Chairman Brett Guthrie.

Brett Guthrie, Chairman of the House Committee on Energy and Commerce

This legislative effort mirrors a similar bipartisan bill called the Ratepayer Protection Act, which was unanimously voted out of committee in July and aims to ensure that households do not bear the cost of energy consumed by data centers.

How Are Tech Giants Addressing the Power Crisis?

While Congress tackles water costs, major technology companies are pursuing a different strategy to solve their power problems: nuclear energy. Microsoft signed a 20-year agreement with Constellation Energy to reopen a retired reactor at Three Mile Island in Pennsylvania, a facility that has been offline since the 1979 accident that made it a symbol of nuclear safety concerns. Meta Platforms has taken a similar approach, announcing an agreement with nuclear technology firm Oklo to develop a nuclear facility in south central Ohio that will deliver 1.2 gigawatts of power to Meta data centers in the region.

The Oklo facility will center on small modular reactors, or SMRs, which can be built more quickly and cheaply because they are partly manufactured in factories and then delivered to the site. Pre-construction is scheduled to begin in 2027, with general construction expected in late 2028 or 2029, and the first phase of the nuclear plant is hoped to come online around 2030. However, Oklo does not yet have the necessary license from the Nuclear Regulatory Commission, and the company's small modular reactor technology remains largely unproven at scale.

The challenge is significant. Oklo's share price is down 45% this year as investors remain cautious about the prospects of its business model and the viability of unproven SMR technology. Meaningful scale-up by 2030 is unlikely, making a date several years later much more plausible for when these facilities will actually begin generating power.

How Is Nvidia's Growth Driving the Infrastructure Crisis?

The urgency of these infrastructure challenges becomes clear when examining the scale of AI chip demand. Nvidia reported second-quarter fiscal 2027 results on August 26, 2026, with revenue of $96.2 billion, up 106% year-on-year. Data Center revenue reached $89.0 billion, up 117%, as demand for AI computing continued to accelerate. The company expects $108 billion in third-quarter revenue, forecasting 70% revenue growth in fiscal 2028.

"AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue. And demand is accelerating," remarked Jensen Huang, founder and CEO of Nvidia.

Jensen Huang, Founder and CEO of Nvidia

Yet this explosive growth masks a growing climate challenge. Nvidia's Scope 3 emissions, which cover indirect emissions across a company's value chain, reached 10.7 million metric tons of CO2e in fiscal 2026, up from 6.9 million tons in fiscal 2025 and 3.6 million tons in fiscal 2024. That means Scope 3 emissions increased about 55% in one year and almost tripled in two years. The largest source of these emissions comes from purchased goods and services, which accounted for about 87% of Nvidia's reported Scope 3 emissions.

Steps Tech Companies Are Taking to Improve Energy Efficiency

Despite the emissions growth, Nvidia and other companies are investing in more efficient computing technologies. Nvidia's latest sustainability report acknowledges that AI demand will increase energy use and says energy is the foundation of the AI infrastructure stack. The company is trying to address this challenge through several approaches:

• Vera Rubin Platform: Nvidia's Vera Rubin NVL72 platform can deliver up to 10 times the energy efficiency of its previous Blackwell architecture, and its Vera CPU can run up to 50% faster with twice the energy efficiency of traditional CPU infrastructure.
• Inference Performance Gains: Nvidia says its Groq 3 LPX combined with Vera Rubin NVL72 can deliver up to 35 times more inference performance per watt than the Blackwell GB200 NVL72 for trillion-parameter models.
• Grid-Responsive Power Systems: Nvidia is collaborating with Emerald AI, the Electric Power Research Institute, and energy firms to create AI data centers that adjust power use according to grid conditions, potentially unlocking up to 100 gigawatts of U.S. power capacity by using existing infrastructure more efficiently.
• Onsite Power Generation: The company is exploring onsite power generation and energy storage to help data centers connect faster and reduce grid pressure.

However, efficiency improvements alone may not solve the problem. Higher performance per watt can reduce the electricity needed for a given amount of computing, but efficiency gains do not automatically reduce total emissions if companies deploy far more AI computing. Nvidia has set science-based emissions targets to cut absolute Scope 1 and Scope 2 market-based emissions by 50% by fiscal 2030, and to reduce Scope 3 emissions intensity from the use of sold GPU products by 75% per petaFLOP by fiscal 2030. Yet the company's Scope 3 goal is an intensity target, not an absolute emissions target, meaning Nvidia can reduce emissions per unit of computing performance while its total emissions continue to rise if it sells enough additional GPUs and systems.

The convergence of these trends reveals a fundamental tension in the AI boom. Tech companies are racing to build the infrastructure needed to power artificial intelligence applications, but the environmental and resource costs of that expansion are increasingly visible to regulators and the public. Congress is stepping in to ensure that communities do not bear these costs alone, while tech giants are betting on nuclear power and efficiency improvements to sustain their growth. Whether these solutions will prove sufficient remains an open question as AI demand continues to accelerate.