Salt: The Thermodynamic Cost of Concentration

Historically, salt production was an energy crisis in disguise. While solar evaporation is viable in arid climates, large-scale production in temperate regions required massive thermal energy to boil brine. Mined rock salt (halite) or concentrated brine wells were strategically superior because geology had already performed the work of concentration, requiring far lower energy input per tonne.

Control over these low-entropy deposits defined military logistics. From Roman salaria to the 1864 Battle of Saltville, conflicts were fought not for the resource itself, but for the energy-efficient production sites that bypassed the massive fuel costs of seawater extraction.

The word “salary” comes from the Roman salarium — salt allowances for soldiers — highlighting salt’s strategic importance for food preservation, trade, and military logistics. The state focused on controlling the most efficient production sites rather than every salt pan.

Rare Earths: Vertical Integration and Energy Moats

REEs are not geologically rare; the U.S. Mountain Pass mine alone accounts for approximately 15% of global mined output. However, the separation and refining stages are extraordinarily energy-intensive, involving hundreds of solvent extraction stages. The separation stage alone demands 13 times the energy of initial mining. Primary energy intensity for REE production is estimated at 110,000kWh per tonne, with carbon emissions often exceeding 20 tonnes of CO2 per tonne of product.

China’s current hegemony accounting for 92% of refining capacity 94% of permanent magnet production is built on state-supported vertical integration and access to cheap, dispatchable baseload power. Western reliance stems from decades of offshoring these energy-intensive, ecologically demanding industrial processes.

The 2016 Pivot: Software as a Leapfrog Strategy

In response to deindustrialization, the 2016 U.S. strategy (evidenced by the Preparing for the Future of Artificial Intelligence report) bet on a computational “leapfrog.” The objective was to use AI to optimize industrial catalysts, accelerate geological exploration, and simulate refining steps—essentially using algorithmic intelligence to mitigate physical energy and material constraints.

The Feedback Loop: AI as an Energy Materializer

The strategy underestimated AI’s own thermodynamic requirements. Training frontier models consumes hundreds of gigawatt-hours, and U.S. data center consumption is projected to rise from 176TWh in 2023 to 580TWh by 2028 (up to 12% of total U.S. electricity).

This creates a self-reinforcing material-energy feedback loop:

• AI Deployment requires increased energy and cooling infrastructure.
• Energy Infrastructure (turbines, EVs, transformers) requires high-performance REE magnets.
• REE Production requires massive, cheap baseload electricity.