Mission Gobi: AI Data Centers in Deserts Powered by Renewable Energy

What happened?

At the VivaTech conference in Paris in June 2025, Envision founder and CEO Lei Zhang presented 'Mission Gobi', a plan to build 5 GW of AI computing capacity in deserts and arid regions by 2030. The initiative is based on the premise that AI should follow energy, not the other way around, leveraging the abundant solar and wind resources of deserts to power data centers without competing with the general power grid. According to The Register, Zhang argued that AI is both an energy revolution and a computing revolution, comparing the GPU to the steam engine: just as the steam engine turned coal into motion, the GPU transforms electricity into intelligence. The key, according to Zhang, is not to produce more energy, but to use it efficiently, following the example of James Watt, who revolutionized industry by improving the efficiency of the steam engine rather than generating more steam.

Why is it important?

The energy consumption of AI data centers is growing at an unsustainable rate. According to Goldman Sachs, data center energy demand in the US will reach 31 GW in 2025 and 66 GW in 2027, although only about 72% of planned centers will come online on time because electricity, not construction, is often the limiting factor. The International Energy Agency (IEA) estimates that data centers consumed about 1.5% of global electricity in 2024, and that figure will rise to 3% by 2030, with AI-specific demand tripling. The power density of AI servers has multiplied elevenfold between 2020 and 2025, and is expected to quadruple by 2027, straining supply chains for power electronics and transformers needed to maintain cluster stability. While AI models are updated every six months and chips every year, power grids have barely changed in decades, creating a structural mismatch. This generates tensions with communities, who see AI competing for electricity with homes, hospitals, and factories, raising questions about who will bear the costs and whether AI infrastructure should depend on a grid that already serves essential uses. Mission Gobi proposes a solution: build new renewable capacity dedicated exclusively to AI, expanding total supply rather than displacing other uses.

What consequences will it have?

If Mission Gobi materializes, it could mark a paradigm shift in data center location. Instead of concentrating in urban areas with good grid connectivity, AI data centers could migrate to desert regions with high renewable potential. This would reduce pressure on existing power grids and interconnection timelines, which are currently the main bottleneck for new projects. Additionally, by generating their own energy, operators could significantly reduce operating costs, as electricity accounts for up to 60% of a data center's expenses. However, it also poses challenges: long-distance data transmission, water availability for cooling in arid areas, and the need for network and transportation infrastructure. Historically, the industry has followed a logic of proximity to users to minimize latency, but with AI, latency is not as critical for training as for inference. This opens the door to remote locations. In comparison, projects like Google's in Finland or Microsoft's in Arizona have already explored the use of renewables, but not at the scale or with the desert focus of Mission Gobi. If successful, it could inspire similar initiatives in other deserts around the world, such as the Sahara, Atacama, or Australia, transforming the global map of AI computing.

What should readers know?

Envision's initiative is not the only one. Other tech companies like Google, Microsoft, and Amazon are also investing in renewable energy for their data centers, but the proposal to build AI capacity in deserts is novel. It is important to note that the plan is still in the announcement phase and there are no concrete details on timelines, investments, or partners. Additionally, the technical and economic feasibility of operating data centers in remote locations with high network latency must be evaluated. Nevertheless, the idea of 'following the energy' could be a key solution for the future of sustainable AI. Readers should watch for Envision's next steps, such as potential partnerships with local governments or energy companies, and how cooling technologies evolve in arid climates. In a context where AI energy demand could outstrip supply, Mission Gobi represents a bold bet that, if successful, could redefine the relationship between technology and the environment.