a16zThe Founders Who Left Tesla to Rebuild America | a16z
CHAPTERS
- 0:00 – 0:50
America’s mineral and grid bottlenecks: the physical constraints on AI and industry
The conversation opens with the premise that U.S. competitiveness in AI and reindustrialization is constrained by physical systems: minerals, energy, and the grid. The speakers frame the U.S. as lagging in critical minerals and stuck with legacy grid infrastructure.
- •U.S. is far behind in critical mineral supply build-out and ramp speed
- •Grid hardware has seen little modernization despite demand growth
- •AI progress depends on materials, energy generation, and power delivery
- •Reindustrialization is framed as a national-scale infrastructure project
- 0:50 – 2:52
Why AI dominance depends on “atoms”: energy, minerals, manufacturing, and transmission
Erin Price-Wright reframes the AI race as more than models and chips—it's a full-stack physical build. She positions grid strain concerns as a call to action and sets up the need to rebuild the entire industrial stack quickly.
- •AI requires materials + energy + electricity delivery at the right place/time
- •Grid constraints are real but solvable with coordinated building
- •U.S. has precedent for rallying around national projects
- •Conversation will span minerals → generation → transmission → interconnection
- 2:52 – 4:19
Meet the builders and what they’re constructing: Mariana Minerals and Heron Power
Turner Caldwell and Drew Baglino introduce their companies and explain why they matter to the U.S. industrial base. Mariana builds vertically integrated mining/refining projects powered by internal operating systems, while Heron modernizes the grid with power electronics.
- •Mariana: software-first mining + refining; builds/operates projects (not SaaS)
- •Three internal systems: Capital Project OS, Plants OS, Mine OS
- •Operating assets: copper production in Utah; lithium refinery planned in Texas
- •Heron: solid-state transformers using silicon + software to upgrade grid power conversion
- 4:19 – 5:30
Heron Power’s thesis: bring power semiconductor gains to the grid
Drew explains how power electronics improved dramatically over decades but the grid hasn’t absorbed those gains. Heron aims to apply modern semiconductors to enable more controllable, monitorable, and scalable grid infrastructure for data centers and large energy sites.
- •Power transistor progress parallels Moore’s Law but hasn’t reached the grid
- •Legacy grid systems are mechanical and lack granular control/monitoring
- •Solid-state transformers can replace traditional steel/oil/copper-heavy designs
- •Target applications: data centers, solar + battery projects, large installations
- 5:30 – 6:46
What happens if the U.S. doesn’t onshore: losing the value of American-born tech
Drew argues that U.S. government, academia, and industry built foundational semiconductor capabilities, so commercialization should benefit the U.S. first. He highlights silicon carbide leadership and the risk of other countries capturing the economic upside.
- •Advanced semiconductors grew from DOE/Navy and broader public-private efforts
- •U.S. leads in silicon carbide production—should lead in applications too
- •Manufacturing domestically retains economic and security benefits
- •Failure to build here exports benefits and increases dependency
- 6:46 – 12:30
Why America is behind China in critical minerals: the build-and-ramp problem
Turner emphasizes that permitting alone doesn’t fix the minerals gap—execution speed after licensing is the true bottleneck. Mariana focuses on compressing the design/build/ramp timeline so the U.S. can outpace China, not just catch up.
- •U.S. lags China (and globally) in minerals supply capability
- •Key constraint: slow design, construction, and ramp-to-rate after permits
- •Typical timeline: ~5 years to build + 3–5 years to reach operating rate
- •Need to go faster than China even if permitting improves
- 12:30 – 14:01
Mariana Minerals explained: autonomy across project delivery, mining, and refining
Turner details Mariana’s bet on autonomy as a way to overcome labor and know-how constraints in mining and refining. The company applies LLM-driven engineering/procurement acceleration and reinforcement learning for real-time control of complex operations.
- •Autonomous construction/resource balancing and workflow automation
- •Reinforcement learning to control refineries amid variable feedstock
- •RL-driven mining control to improve productivity and equipment utilization
- •Addresses limited U.S. labor pool with embedded operational know-how
- 14:01 – 14:19
Getting software into heavy industry: culture, operating teams, and tool adoption
Turner argues the limiting factor isn’t algorithms—it’s adoption at the operating layer where many sites still run on paper and sprawling spreadsheets. Mariana co-locates software engineers with operators to design tools that fit real workflows and align incentives.
- •Software penetration is constrained by operator comfort and existing workflows
- •Many operations still rely on pen/paper and scattered spreadsheets
- •Effective digitization requires deep understanding of on-the-ground problems
- •Co-locating engineers with operators enables better tools and faster uptake
- 14:19 – 14:37
Why Drew left Tesla: innovation at the edge vs stagnation “on the other side of the wire”
Drew describes seeing rapid progress in EVs, charging, and grid storage while core grid infrastructure remained largely unchanged. He also explains that U.S. build speed is possible when stakeholders align to say “yes” through permitting and execution.
- •Edge innovations: EV adoption, Supercharging buildout, Megapack scaling
- •Core grid equipment remains similar to systems from ~100 years ago
- •Fragility + overbuild result from limited control/monitoring
- •U.S. projects can move fast when jurisdictions align around “yes” decisions
- 14:37 – 14:55
Onshoring realities: labor isn’t the main cost—supply-chain co-location is
Drew challenges the notion that U.S. labor costs make domestic manufacturing uncompetitive. He argues automation reduces labor share, and the bigger unlock is developing dense, co-located supplier ecosystems to cut time and logistics costs.
- •Modern factories are highly automated; labor share can be <10% of COGS
- •Competitiveness hinges on supply chain proximity and logistics speed
- •China’s industrial clustering enables rapid iteration and low friction
- •U.S. needs manufacturing zones with co-located suppliers and automation
- 14:55 – 18:38
Tesla’s operating model: techno-optimism, risk tolerance, and persistence
Both founders describe what they carried from Tesla into their new companies. They emphasize a culture that believes legacy systems can be reinvented, makes decisions quickly, and persists through repeated setbacks when outcomes matter.
- •Techno-optimism: willingness to modernize “old and archaic” systems
- •High risk appetite enables faster decisions and execution
- •Persistence: not shelving hard problems after early failures
- •Startup-like intensity when outcomes are existential
- 18:38 – 21:09
Building the 2026 industrial workforce: recruiting from analog industries and rebranding ‘unsexy’ sectors
The discussion turns to hiring for large new facilities and scaling operations in the U.S. Both founders stress creative sourcing from adjacent industries, transferable skills, and building a mission-driven talent magnet—especially for industries like mining with image challenges.
- •Reindustrialization requires creative recruiting—not “phone books” of specialists
- •Examples: bottling/syringe plants → battery/advanced manufacturing roles
- •Oil & gas and tech optimization skillsets transfer to minerals operations
- •Need to “make mining sexy” and create a compelling mission to attract talent
- 21:09 – 23:33
Policy and execution asks: durable incentives, ‘yes’ jurisdictions, and a grid build-out plan
In closing, both founders outline concrete asks to speed near-term production and job creation. They call for durable industrial policy and incentives that mobilize private capital, plus coordinated siting/zoning and a national approach to grid infrastructure akin to highway funding.
- •Minerals: incentives modeled on prior oil & gas-era tools to mobilize capital
- •Durable, predictable policy so suppliers/financiers can plan long-term
- •Federal-state coordination to create energy/manufacturing build-out zones
- •Proposal: grid equivalent of a federal highway trust fund for resilient expansion
