Find cheap unused land with high sun exposure (e.g. desert areas)
Build there solar farms and datacenters (for AI/LLM inference; using either ASICs for high energy efficiency or GPUs)
Use solar to power these datacenters (essentially free energy)
Charge per token, or potentially future auction bidding (pay more for higher intelligence access)
Latency of distant locations may not be an issue… intelligence processing usually takes time anyway + Starlink can handle it
My opinion: demand for intelligence (AI/LLM inference, compute) will grow in coming years, as will energy demand. Possibly more demand than supply.
Some actors will pay a premium for intelligence as their use case/business justifies it.
Cutting-edge LLMs (e.g. GPT-10..) might stay closed, available only via developers. These models will be sought after as they’ll outperform others – for some use cases, it’s worth the price (privacy loss, dependency, etc.)
I have always been driven by the need to create — games, AI agents, ideas. That’s why I started Keen Software House: to create games that only existed in my head. After Space Engineers took off, I founded GoodAI to develop AGI, to help humanity and understand the universe.
These days I’m focused on Space Engineers 2, the VRAGE3 engine, AI People, and autonomous agents in general — powering NPCs in our games, or swarms of autonomous and intelligent drones.
It’s all part of my long-term plan: to make civilization stronger, greater, and more resilient.
Our home base is a 17th-century Oranžérie in Prague — but we’re a remote-first, global team of 100+ programmers, artists, designers, and engineers.
I am proudly European , and in the last few years, I’ve come to love South Africa and its people.
I love this idea – do you want to do it? Also, we could use open source LLMs like Llama or DeepSeek
Thanks! I would use DeepSeek at this moment, but in general I would be agnostic, and always use what’s best (things will be changing rapidly…)