The most underappreciated risk to generative AI is drought Image By Dan Romito Generative artificial intelligence represents one of the most transformative technological advancements of the modern era. New capabilities introduced by generative AI will reshape socioeconomic, geopolitical, and innovative dynamics over the next century. However, the critical infrastructure enabling these advancements – semiconductor chips and hyper-scale data centers – relies heavily on water. As global drought and water scarcity intensify, particularly in the United States, this dependency emerges as a significant and underappreciated risk to generative AI's continued development and sustainability. Next-generation semiconductor chips are at the heart of generative AI systems, providing the computational power required for machine learning and neural networks. Producing a single semiconductor chip requires approximately 2,200 gallons of “ultra-pure” water, which is essential for cleaning and etching the microscopic circuits on silicon wafers during chip fabrication. Chip demand drives water demand The water-intensive nature of semiconductor manufacturing raises concerns about resource sustainability, mainly as global demand for these chips grows exponentially and the state of water continues to deteriorate. The scale of this water dependency becomes even more apparent when considering electric vehicles (EVs), which incorporate generative AI systems for autonomous driving and other advanced features. Newer EV models can contain up to 3,000 semiconductor chips. The cumulative water footprint of producing these chips is staggering, highlighting the unspoken environmental costs of the technologies driving both AI and the broader digital economy. Beyond chip manufacturing, hyper-scale data centers – the backbone of AI operations – also contribute significantly to water usage. These facilities house the servers that train, deploy and run generative AI models, requiring immense computational power and energy. Approximately 40% of the electricity these data centers consume is dedicated to air conditioning and cooling systems, which often depend on water-intensive technologies to prevent overheating. As AI adoption grows, so does the demand for data center capacity, amplifying their environmental and water resource impact. The state of water resources in the United States, particularly in the Southwest and West, is increasingly dire. Prolonged drought conditions have led to significant water scarcity in states like California, Arizona, and Nevada. These regions are home to many semiconductor manufacturing facilities and data centers, creating a critical intersection of technological advancement and environmental vulnerability. The reliance on water in these drought-prone areas raises questions and risks about the long-term viability of AI infrastructure and the potential for water shortages to disrupt the industry. What policymakers can do Addressing this challenge requires proactive measures from both the technology sector and policymakers. Semiconductor manufacturers are exploring reducing water usage through recycling and more efficient production methods. Similarly, data centers are adopting innovative cooling technologies, such as liquid immersion cooling, to minimize water and energy consumption. However, these efforts must be scaled up significantly to mitigate the growing risk posed by water scarcity. Policymakers also have a crucial role to play in ensuring sustainable water management. Investments in water infrastructure, conservation initiatives, and policies promoting water efficiency are essential to balancing the needs of technological innovation with environmental stewardship. Encouraging the development of semiconductor manufacturing and data center operations in regions with more abundant water resources could also help reduce the strain on drought-affected areas. Generative AI has the potential to revolutionize industries and address some of humanity’s most significant challenges. However, its success depends on a resource that is becoming increasingly scarce. Recognizing and addressing the water intensity of AI-related technologies is essential to ensuring their sustainability and resilience in the face of mounting environmental pressures. Dan Romito is managing director overseeing the Consulting & Advocacy practice at Pickering Energy Partners. He previously worked at Nasdaq, and his writing has been published in Harvard Business Review, Bloomberg and CNBC. *The opinions expressed in this column are those of the author and do not necessarily reflect the views of EnergyPlatform.News.
