Is The Hidden Cost Being Forgotten?

The Stargate Initiative has been making waves with its promise of a $500 billion investment to supercharge America’s AI capabilities. On the surface, it’s a forward-looking plan to cement global leadership in artificial intelligence.

But in the race to celebrate this monumental funding, we should not sideline the staggering energy cost of powering AI. By 2028, datacenters are expected to consume as much as 12% of the country’s total electricity. This is more than double their share today. Ambitious projects like Stargate could reshape the nation’s energy consumption and environmental footprint. As we are faced with increasingly frequent extreme weather and natural disasters, managing environmental impact of datacenters should take priority.

Will Stargate lead us to a sustainable future, or will it lock us into an era of escalating energy demand and ecological strain?

Datacenters as Strategic Assets in AI

The Stargate initiative reflects a global trend. International governments recognize that datacenters are essential for the advancement of AI, and will redefine world order. Datacenters provide the computational support for training and deploying AI models, AI agents and sophisticated robotics systems, which are key areas of technological development in the next decade. The European Commission is pushing for their datacenters to adopt renewable energy, reuse waste heat, and achieve carbon neutrality by 2030. China is developing the East-West Computing Resources Transmission Project to channel renewable energy including solar, wind, and water from the country’s western regions to process data tasks in the eastern provinces.

According to a recent report by the U.S. Department of Energy, total data center electricity usage climbed from 58 TWh in 2014 to 176 TWh in 2023 and estimates an increase between 325 to 580 TWh by 2028. It is clearly unsustainable if no measure is taken to make datacenters more energy-efficient.

The Environmental Costs of AI Hardware

Beyond electricity, the production of GPUs—the hardware powering AI systems—has its own environmental challenges. GPUs consist of silicon layered with “tantalum and palladium transistors and capacitors for better storage on a smaller chip,” made from chemicals including copper, boron, cobalt and tungsten. These rare earth metals are sourced primarily from Southeast Asia, often through unjust labor practices that exacerbated environmental degradation.

Cobalt mining in the Democratic Republic of Congo, which supplies over 70% of the global demand, is often tied to unsafe working conditions and child labor. Similarly, tungsten and tantalum extraction contributes to deforestation, soil erosion, and pollution of water sources. The processes involved in mining and refining these materials also release greenhouse gases thousands of times more potent than CO₂.

The very hardware driving the AI revolution risks becoming a significant source of environmental and ethical challenges, threatening to undermine the long-term benefits of the technologies it supports.

Moving Toward Sustainable AI Infrastructure

The rapid expansion of datacenters presents an opportunity to embed sustainability into the core of AI infrastructure development. Several existing strategies can serve as starting points for Stargate Initiative:

Renewable Energy Integration: Companies like Google and Microsoft are investing in renewable energy sources to power their datacenters. Meta’s facility in Luleå, Sweden, is an example of this approach, running on hydropower and using natural cooling from the region’s climate.

Heat Reuse: Datacenters in Denmark redirect excess heat from servers to nearby homes and businesses, creating a circular energy system.

In addition to operational improvements, addressing the lifecycle impact of GPUs is critical. Efforts to use recycled materials, as seen in Apple’s product designs, and innovations in energy-efficient GPU architecture, such as Nvidia’s Grace GPUs, are promising steps. Investments in alternative materials for semiconductors and advancements in recycling practices for rare earth metals could further reduce the environmental costs of AI hardware.

The growing energy demands of datacenters also highlight the need for scientific innovation. Photonic chips, which use light rather than electricity for data transmission, could significantly reduce power consumption.

Policymakers and industry leaders must also consider the impact on regional power grids. As electricity demand surges, investments in renewable energy storage and advanced grid management technologies will be essential to prevent over-reliance on fossil fuels.

Equitable Access to AI Resources

Stargate Initiative aims to advance national AI capabilities, but will its benefits be distributed equitably? Just as disparities in digital technology access marginalized some communities, unequal access to AI infrastructure could aggravate existing inequalities between regions and institutions.

For example, rural regions often lack access to cloud computing resources, which are essential for AI-driven solutions in fields like agriculture and healthcare. Globally, nations in the Global South may struggle to participate fully in AI research and applications due to limited infrastructure and funding.

To address these disparities, international framework should be established to provide public institutions, startups and nonprofits subsidized access to datacenters. On a global scale, partnerships should ensure that major tech companies take as much responsibility as possible to ensure underdeveloped regions benefit from AI advancements.

A Thoughtful Approach to AI Infrastructure

The Stargate Initiative represents an important step in developing the infrastructure necessary for AI advancement. However, its implementation will have implications beyond national competitiveness. Decisions made today will shape the environmental and societal wellbeing for decades to come.

Integrating sustainability into data center design, advancing energy-efficient technologies, and promoting equitable access can enable more responsible AI infrastructure deployment for the greater benefit of society.