We are at a critical inflection point where the demand for metals underpinning everything from AI data centers to electric vehicles is skyrocketing, yet it can still take over 15 years to permit and build a new mine in the United States. This stark contradiction between urgent technological need and glacial industrial pace forms the core of a conversation with Turner Caldwell, founder of Mariana Minerals. A Tesla veteran who worked his way upstream from factory design to battery mining, Caldwell explains that critical minerals are the hidden bedrock of modern life, embedded in every phone, laptop, and renewable energy system. The discussion dives deep into why this industry is so resistant to change and how his new company aims to rebuild America’s capacity to mine and refine these essential materials through vertical integration and a software-first approach.

The path from raw ore to a usable component in a battery or magnet is a long, complex, and shockingly bespoke journey. Each mining site has a unique chemical signature, requiring a custom-built processing “flow sheet.” This lack of standardization, combined with a shrinking talent pool and a cultural aversion to risk, has calcified the industry. Major mining companies operate with a decentralized, low-risk appetite, making it nearly impossible for innovative tech startups to sell point solutions. Caldwell observed this firsthand at Tesla, where the incentive structure between manufacturers (who want lower costs at higher volume) and miners (who charge more for increased demand) was fundamentally misaligned. This led him to believe that the only way to drive efficiency is to own and optimize the entire chain, from discovery to refined metal.

Mariana’s thesis is that the only way to disrupt this massive, technologically stagnant market is through vertical integration and a radical focus on software. The company is building two core systems: a “Capital Project OS” to streamline the notoriously slow and data-lagged construction of mining infrastructure, and a “PlantOS” designed to autonomously control refining operations. By using reinforcement learning and LLMs, they aim to optimize complex, variable-rich processes—like adjusting to changing ore grades—much faster than human operators ever could. The goal is to enable a small, elite team to accomplish what traditionally requires thousands, thereby making smaller, “subscale” mineral deposits economically viable to mine and reducing America’s dependency on foreign supply chains, particularly China’s dominant and highly efficient industry.

Surprising Insights

• Inverse Economics: In mining, increased demand from a customer typically leads to higher prices, the opposite of the manufacturing world where scaling up drives costs down through economies of scale. This creates a fundamental incentive misalignment between miners and their industrial customers.
• The Bespoke Nature of Dirt: Every mining deposit is chemically unique, requiring a completely custom-designed processing circuit. There is no one-size-fits-all solution, which massively complicates scaling technology and expertise across the industry.
• China’s Overwhelming Advantage: The edge isn’t just policy support; it’s an immense, skilled talent pool. Caldwell cited a Chinese-built nickel refinery in Indonesia that mobilized 13,000 workers on-site during construction—a scale of labor coordination and availability currently unimaginable for a similar project in the West.
• The “Orphan” Assets: Major mining companies only want giant, multi-billion dollar deposits, leaving a long tail of smaller, high-quality resources unexploited because they are deemed “subscale.” This represents a significant opportunity for more efficient operators.
• Human as Temporary Robot: Refining plants are often operated via “teleoperated humanoids,” where workers manually adjust set points in response to changing conditions. The industry has historically automated discrete control loops (like temperature) but not the holistic, multivariable optimization of the entire plant.

Practical Takeaways

• Advocate for Permit Reform: Support policies that streamline the permitting process for mineral exploration and mine development, especially on federal lands, to unlock domestic resources and attract talent who want to see their projects built.
• Push for Demand-Side Support: Encourage government programs that provide price floors or fixed-price offtake agreements for critical minerals. This de-risks the commodity cycle and mobilizes the vast pools of private capital currently avoiding mining investments.
• Embrace Vertical Integration for Critical Paths: For industries dependent on constrained commodities, consider following the Tesla playbook of integrating upstream where there are incentive misalignments or a lack of innovation from traditional suppliers.
• Apply AI to High-Latency, Multivariable Systems: Look for industrial operations with complex, interdependent processes and long feedback loops (like chemical refining) as prime candidates for reinforcement learning optimization, which can find counterintuitive, global optima beyond human intuition.
• Build a Culture that Trusts the Counterintuitive: To leverage AI and ML effectively, companies must foster a culture where operators trust and act on algorithmic recommendations, even when they defy conventional wisdom, moving beyond local optima.