Datacenters in the Desert: Feasibility and Sustainability of LLM Inference in the Middle East

TL;DR

This paper investigates the feasibility of sustainable LLM inference in desert-based datacenters, focusing on the Middle East. It employs a controlled empirical setup using the DeepSeek Coder 1.3B model evaluated on HumanEval, with CodeCarbon estimating energy use and $CO_2$ emissions across four regions (UAE, Iceland, Germany, Texas) while holding workload and hardware constant. Findings show that carbon emissions track grid carbon intensity more strongly than local efficiency improvements, with Iceland and Germany exhibiting substantially lower emissions than UAE and Texas, though electricity costs vary significantly. The work suggests that desert deployments can be climate-conscious with co-located clean energy and advanced cooling, contributing to a diversified, resilient global AI infrastructure.

Abstract

As the Middle East emerges as a strategic hub for artificial intelligence (AI) infrastructure, the feasibility of deploying sustainable datacenters in desert environments has become a topic of growing relevance. This paper presents an empirical study analyzing the energy consumption and carbon footprint of large language model (LLM) inference across four countries: the United Arab Emirates, Iceland, Germany, and the United States of America using DeepSeek Coder 1.3B and the HumanEval dataset on the task of code generation. We use the CodeCarbon library to track energy and carbon emissions andcompare geographical trade-offs for climate-aware AI deployment. Our findings highlight both the challenges and potential of datacenters in desert regions and provide a balanced outlook on their role in global AI expansion.

Figures (1)

• Figure 1: Carbon and energy usage for equivalent inference workloads across regions.