Kevin O’Leary says AI infrastructure is becoming less water-intensive, challenging one of the biggest environmental criticisms surrounding the rapid expansion of artificial intelligence. As governments and technology companies invest billions of dollars in AI data centres, concerns have grown over their electricity and water consumption. However, Kevin O’Leary, chairman of O’Leary Ventures and a well-known investor from Shark Tank, believes modern facilities have significantly reduced their water requirements through advanced cooling technologies.

Speaking on Glenn Beck’s radio programme, O’Leary argued that today’s AI data centres are fundamentally different from those built 10 to 15 years ago, which relied heavily on evaporative cooling. His comments come as his company develops major AI infrastructure projects, including a 1.4-gigawatt data centre in Utah, while the global AI industry faces increasing scrutiny over its environmental footprint.

Modern Cooling Systems Are Changing AI Data Centres

O’Leary said many public concerns about water consumption are based on outdated infrastructure rather than the latest generation of AI facilities.

• According to him, modern data centres use far more efficient cooling technologies, including closed-loop cooling systems, which significantly reduce reliance on evaporative cooling and minimize water use.
• Comparing his Utah project to a golf course, O’Leary said the facility’s water requirements would be similar despite delivering 1.4 GW of computing capacity, highlighting how cooling technology has evolved.
• He noted that older data centres, particularly those built in Virginia more than a decade ago, created the industry’s reputation for excessive water use, but today’s designs have changed that equation.

Industry Leaders Push Back Against Water Consumption Claims

The debate over AI’s environmental impact extends beyond O’Leary.

Earlier this year, OpenAI CEO Sam Altman also dismissed widely circulated claims that individual AI queries consume large amounts of water. He said those estimates are based on outdated evaporative cooling systems rather than modern facilities that use far less water.

Despite these assurances, environmental concerns remain. Previous investigations have found that several of the largest U.S. data centre campuses hold permits allowing daily water consumption exceeding what nearly 49,000 Americans would typically use in the same period, demonstrating why the issue continues to attract public attention.

Also Read: France’s Floating Igloo Embarks On Historic Polar Exploration Mission

AI Growth Must Include New Power Generation

• While Kevin O’Leary says AI infrastructure is becoming less water-intensive, he also believes future AI projects should take greater responsibility for their energy demands.
• Rather than relying solely on existing electricity grids, O’Leary argues that developers should build their own power generation alongside new data centres.
• He suggested that companies investing billions in AI infrastructure should contribute surplus electricity back to the grid, helping strengthen national energy security while supporting economic growth.

His proposal comes as utilities across the United States face mounting pressure from rapidly increasing electricity demand driven by AI computing.

Also Read: European Union Suspends Methane Penalties To Safeguard Energy Supply

Environmental Debate Is Far From Over

O’Leary’s remarks arrive as AI infrastructure investment accelerates worldwide. Although modern cooling systems may reduce water consumption, experts continue to examine the industry’s broader environmental footprint, including electricity use, carbon emissions, and impacts on local resources.

Meanwhile, O’Leary is also facing a defamation lawsuit linked to public comments about opposition to his Utah project. His legal team has stated that he intends to vigorously defend the case.

As AI continues transforming industries, discussions around sustainable infrastructure, efficient cooling technologies, renewable energy, and responsible resource management are likely to remain central to the future of data centre development.

Also Read: NASA Climate Records Moved To Swiss AI Supercomputer Following US Funding Cuts