AI Compute Oligarchy

Energy is becoming the new frontier of AI.

X-energy, a Maryland-based small modular reactor (SMR) company, has just raised $700 million in a Series D round led by Jane Street, with investors including ARK, Ares, Emerson Collective, and XTX Ventures. Last year, Amazon led a $500 million round and is also a customer.

The reason is simple. Hyperscalers anticipate needing more power than the grid can supply.

• Right now, only 2,259 MW of AI compute is operational.
• There are 35,736 MW in the pipeline.

This represents a 15.8 times expansion, with 94% of AI power infrastructure still unbuilt.

According to McKinsey's research, we are heading toward a supply deficit. Even if every announced project is completed, data center power will not meet demand by 2030. In the United States alone, the shortfall is expected to exceed 15 gigawatts.

The infrastructure oligarchy that nobody is talking about is forming. A small group of companies, including Meta, Oracle, OpenAI and Microsoft, xAI, and Crusoe, are becoming the gatekeepers of future AI capability.

This gap is now driving a surge of investment into next-generation energy companies such as SMRs, on-site generation, advanced geothermal, hydrogen turbines, and AI-optimized microgrids. The competitive viability of small nuclear reactors is still debated.

However, one thing is clear. Capital availability will not be the bottleneck.

We are in an AI power race, and it will reshape who controls the future of training and inference at scale.

If this chart helped you understand where AI is heading, share your thoughts in the comments.