Linas Beliūnas Analyzes Potential SpaceX-xAI Merger as Planetary-Scale Vertical Integration

In a recent LinkedIn post, Linas Beliūnas discusses the implications of a hypothetical $250 billion acquisition of xAI by SpaceX, framing it not merely as financial engineering but as a potential planetary-scale vertical integration play for artificial intelligence. Beliūnas posits that this ambitious move could circumvent terrestrial limitations on power and infrastructure required for advanced AI development.

AI’s Thirst for Power: The Space-Based Solution

Beliūnas highlights the immense and escalating power demands of modern AI, particularly large-scale training runs. He points out that current AI infrastructure, exemplified by xAI’s operation of over 100,000 Nvidia H100s, is moving beyond the capacity of terrestrial power grids. As Linas Beliūnas notes:

“Next training runs need gigawatts, not megawatts. The grid isn’t built for that.”

This sets the stage for his argument that space offers a viable, and perhaps necessary, alternative. Beliūnas explains the advantages of space-based operations for AI infrastructure. According to Linas Beliūnas, solar power in orbit can generate significantly more energy annually than ground-based arrays due to the absence of weather disruptions and the continuous sunlight cycle. Furthermore, he emphasizes the cooling benefits:

“Space offers free cooling. Heat radiates into vacuum instead of burning water and electricity on chillers.”

Starship and Starlink: Enabling the Space AI Infrastructure

The feasibility of such a large-scale space-based AI operation, Beliūnas argues, is increasingly supported by advancements in space technology. He specifically points to the potential impact of reduced launch costs associated with SpaceX’s Starship program. As Linas Beliūnas outlines, the economics of transporting hardware to orbit are becoming more justifiable compared to the extensive, decade-long grid upgrades required on Earth.

Beyond infrastructure, Beliūnas also touches upon the synergistic potential with existing and future SpaceX services. He suggests that the combination of Starlink’s global connectivity and X’s payment rails could create a powerful ecosystem. This integration, in his view, would enable not just AI models that answer questions but also autonomous AI agents capable of executing actions and transactions.

Overcoming Terrestrial Bottlenecks

Beliūnas contrasts this potential space-centric approach with the challenges faced by companies operating solely on Earth. While competitors are contending with land acquisition, permits, and substation development, this proposed SpaceX-xAI integration aims to shift the primary bottleneck to orbital operations. In Linas Beliūnas’s view:

“Everyone else is fighting for land, permits, and substations. This stack is trying to move the bottleneck to orbit.”

He concludes by acknowledging the dual nature of such a venture, suggesting it could be either the most groundbreaking infrastructure play in technology or an exceptionally costly method for revaluing an AI company. However, he leans towards the likelihood that it will be both.