Energy, Minerals, and the Physical Stack Behind AI
Turner Caldwell (Mariana Minerals) and Drew Baglino (Heron Power), both former Tesla executives, discuss how America's AI and reindustrialization ambitions are constrained by physical infrastructure gaps — specifically a 50-year lag in critical mineral supply chains and a power grid running on pre-WWII mechanical systems. They argue that the Tesla model of techno-optimism, risk appetite, and relentless execution can be applied to mining, refining, and grid-scale power electronics to close these gaps.
Summary
The conversation, hosted by Aaron Price-Wright for the A16Z podcast, frames America's AI dominance as fundamentally a physical infrastructure challenge rather than a software or algorithms challenge. The host argues that every breakthrough AI model and autonomous system has real-world requirements in materials, energy, and electricity transmission that are currently underbuilt or misaligned with demand.
Turner Caldwell, co-founder and CEO of Mariana Minerals, describes his company as a software-first minerals mining and refining operation. Roughly a quarter of the company consists of software and ML engineers building three core operating systems: Capital Project OS (agentic workflow automation for engineering and procurement), Plant OS (reinforcement learning for refinery control), and MineOS (autonomous short-interval control of mining operations). Crucially, Mariana does not sell software — it develops, builds, and operates mineral projects, including an operating copper mine in Utah and a lithium refinery under construction in Texas, with a goal of 10 projects in 10 years. Caldwell emphasizes that the U.S. is 50 years behind China in critical mineral supply, and that the core bottleneck is not just permitting but the slow pace of construction and ramp-up even after licenses are granted.
Drew Baglino, founder and CEO of Heron Power, focuses on the grid infrastructure layer. He argues that while Moore's Law-style improvements in power semiconductors have transformed consumer electronics and data centers over four decades, those advances have not been applied to the grid itself, which still relies on mechanical systems developed over 100 years ago. Heron Power builds solid-state transformers using silicon carbide semiconductors and software to replace steel, oil, and copper in power conversion at data centers and large-scale energy installations. Baglino notes that the world's leading silicon carbide producer is based in the U.S., making it strategically important to commercialize those applications domestically.
Both founders draw heavily on their Tesla experience — Baglino spent 18 years there, leading Megapack, the 4680 battery program, and the energy business; Caldwell led Tesla's minerals and metals team. They identify the key advantages the Tesla model provides: a deep belief that old, archaic systems can be innovated on; a high appetite for risk enabling fast decision-making; and a firm commitment to seeing projects through to their intended outcome rather than shelving failed attempts. Baglino also highlights Tesla's existential urgency, clear company mission as a talent magnet, and the high growth environment that retains ambitious people.
On workforce, both founders argue that analog industries are the key to building industrial talent in the U.S. during reindustrialization. Baglino drew manufacturing talent for the 4680 facility from high-speed bottling plants and syringe manufacturers. Caldwell notes that oil and gas talent translates well to mining, and that optimization algorithms for mining plants closely resemble those used in ride-sharing and ad tech. Both companies are planning facilities expected to create roughly 500+ jobs each in the near term.
On policy, Caldwell calls for applying the same toolkit used for oil and gas over the past 50 years to critical minerals — including incentive structures that mobilize private capital with long-term market confidence. Baglino advocates for durable industrial policy that suppliers and financiers can plan around, federally coordinated manufacturing and energy build-out zones to enable co-located supply chains, and a federal highway trust fund equivalent for the grid to create a master plan for linear infrastructure build-out.
About this episode
Erin Price-Wright speaks with Turner Caldwell and Drew Baglino about what it will take to close America's critical minerals gap and modernize the power infrastructure that underpins the AI economy. With the US more than 50 years behind China in critical mineral supply and grid infrastructure built on systems designed a century ago, they examine where the real bottlenecks are and how to move faster. The conversation covers how automation, reinforcement learning, and vertically integrated operations can compress the timelines for mining and refining, and why co-locating supply chains matters more than labor costs in the race to reshore manufacturing. Baglino explains how solid state transformers can replace aging mechanical grid equipment with silicon and software, while Caldwell outlines how Mariana Minerals is applying autonomous systems to remove the know-how bottleneck from critical mineral processing. They also discuss the lessons both founders carried from Tesla — techno-optimism, appetite for risk, and mission-driven talent — and what durable industrial policy, smarter permitting, and a federal grid investment framework would unlock for American competitiveness.
Key Insights
- Caldwell argues that the U.S. is 50 years behind China in critical mineral supply, and that the core problem is not just permitting but the slow pace of construction and operational ramp-up even after licenses are granted — a gap that software-driven autonomy is specifically designed to close.
- Baglino contends that despite four decades of Moore's Law-equivalent improvement in power semiconductors, the grid itself still runs on largely mechanical systems designed over 100 years ago, leaving it fragile, overbuilt, and dependent on overseas suppliers for critical infrastructure equipment.
- Both founders argue that the Tesla model's most transferable advantage is not any specific technology but rather a cultural commitment to fighting through project challenges as long as the outcome is worth it — contrasting sharply with incumbent industrial companies that tend to shelf failed innovation attempts rather than iterate through them.
- Caldwell claims that Mariana Minerals is using reinforcement learning to remove humans from the loop in refinery operations, arguing that the U.S. lacks a domestic labor pool with the embedded know-how to manage the constant variability of a complex refining circuit — making algorithmic control not just an efficiency play but a national supply chain necessity.
- Baglino argues that the labor cost differential between U.S. and Chinese manufacturing is less than 10% of cost of goods sold — and possibly under 5% — meaning the real competitiveness gap comes from co-located supply chains, where China's industrial zones concentrate all necessary inputs within a three-hour drive, an advantage the U.S. has not replicated.
Topics
Transcript
The U.S. is 50 years behind on critical mineral supply. We are too slow at designing, building, and ramping up new minerals capacity, even after we have licensed to operate. Even though there's so much innovation happening at the edge of the grid, on the other side of the wire, there's really been no change. You both came out of Tesla. What does the Tesla model give you that a traditional industrial company doesn't have? The belief that you can innovate on systems that are old and archaic. If the outcome is worth it, Tesla will fight through the challenges of getting to that outcome. We're making a big bet on autonomy in refineries, where we use reinforcement learning to…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The a16z Show
The Top 100 Consumer AI Apps: Who’s Actually Paying?
A16Z's seventh edition Top 100 Consumer AI Apps Report reveals that while half of Americans use AI, only 4.5% pay for subscriptions, with spending heavily concentrated among developers and power users. The report highlights emerging personal agents as a major trend, significant divergence between ChatGPT, Claude, and Gemini, and substantial untapped opportunities in categories like dating, social, shopping, and entertainment.
David George & Jack Altman on AI, Autonomy, and the Next $25 Trillion
David George from Andreessen Horowitz argues that AI, autonomy, and robotics will create over $25 trillion in market value over the next decade, with success across multiple layers of the stack—frontier models, open source, and applications will all flourish due to massive supply constraints and early-stage demand penetration among knowledge workers.
Beyond the God Model | Alex Atallah & Amjad Masad
Alex Atallah (OpenRouter) and Amjad Masad (Replit) discuss how the future of AI development lies in specialized, diverse models working together rather than relying on single AGI-like models. They argue that enterprises need independence from foundation model providers through model diversity, cost efficiency, and ownership of their own AI capabilities, while also addressing the risks of general-purpose agents and the importance of model specialization.
Why AI Agents Can Beat the Incumbents
Leo, an AI-native startup, is building multi-agent systems to automate end-to-end procurement processes for enterprises by coordinating work across multiple departments, stakeholders, and external systems that traditional procurement software cannot handle. The company differentiates itself from incumbents by owning the entire job workflow rather than being confined to a single system of record, and builds customer trust through human-in-the-loop approaches before scaling to fully autonomous negotiations.
The $1 Trillion AI Buildout | State of Markets
A16Z partners discuss the $1 trillion AI infrastructure buildout, arguing it's driven by real earnings growth rather than inflated valuations, with adoption still extremely early at the enterprise level. They highlight opportunities across consumer agents, robotics, autonomous vehicles, and enterprise diffusion, while noting that the market's 90% gain since ChatGPT reflects fundamental business performance rather than speculative excess.