The Wall Street Journal recently reported that Sam Altman, CEO of OpenAI, aims to secure up to $7 trillion for an ambitious technology initiative designed to significantly enhance global semiconductor capacity, with funding from investors including the United Arab Emirates. This project aims to supercharge AI model capabilities.

However, the environmental ramifications of such a colossal undertaking are undeniable, as noted by Sasha Luccioni, the climate lead and researcher at Hugging Face. Luccioni highlights the staggering demand for natural resources this project would entail. She emphasizes that even with renewable energy, the required volume of water and rare earth minerals would be overwhelming.

For context, Fortune magazine in September 2023 disclosed that AI technologies contributed to a 34% rise in Microsoft’s water usage. Additionally, it was reported that Meta’s Llama 2 model consumed twice the water of its predecessor, and a study found that the training of OpenAI’s GPT-3 used 700,000 liters of water. The scarcity of rare earth minerals like gallium and germanium is exacerbating the global semiconductor dispute with China.

Luccioni critiques Altman’s approach for not prioritizing more efficient AI development methods, suggesting instead that his strategy is perceived by some as visionary despite its brute-force nature.

The shortage of GPUs, crucial for AI development, is a well-discussed issue in Silicon Valley, particularly the scarcity of Nvidia’s H100 GPU, essential for training large language models (LLMs). Meta’s CEO, Mark Zuckerberg, recently outlined the company’s AI ambitions, emphasizing the need for top-tier computing infrastructure, including the acquisition of approximately 350k H100 GPUs by year-end, contributing to a total of around 600k H100 equivalent units.

Furthermore, Luccioni raises concerns about the lack of transparency regarding the environmental impact of AI, particularly the carbon footprint associated with Nvidia’s product lifecycle. Despite Nvidia’s 2023 Corporate Responsibility Report detailing efforts to monitor and report on the environmental impact of their supply chain, Luccioni argues that overall, companies are becoming less transparent about the environmental costs of AI.

In conclusion, while Altman’s project garners attention and possibly hype akin to Elon Musk’s ventures, Luccioni remains skeptical about its feasibility, questioning the long-term sustainability and transparency of such ambitious technological endeavors in the face of significant environmental concerns.