Elon Musk – "In 36 months, the cheapest place to put AI will be space”

In this podcast episode, Elon Musk dives into a wide range of subjects, sharing his vision on the future of AI, energy, space-based computing, manufacturing, robotics, government policy, and the challenges facing technology scaling. His insights encompass the interplay between AI and infrastructure, the competitive landscape of chip manufacturing, the ambitions of SpaceX and Tesla, and his philosophical and practical thoughts on building a future where humanity and advanced intelligence coexist and thrive.

Space-Based AI and Energy Constraints

Musk begins by addressing the significant energy demands of modern data centers and the challenges of scaling these facilities on Earth. He points out that only a small fraction of data center costs is energy, with most costs tied to GPUs, which are difficult to service once placed in space. Nonetheless, he emphasizes that energy availability rather than GPU reliability is the main bottleneck. He notes that global electrical output outside China is almost flat, while chip output—and consequently power demand—is growing exponentially. The scarcity of easily expandable power sources on Earth makes scaling AI compute locally untenable in the long run.

His solution is space-based solar power. Solar panels in orbit avoid atmospheric losses, offer continuous sunshine without night cycles or seasonal changes, and eliminate the need for batteries, making solar energy five to ten times more effective in space compared to the ground. Regulatory hurdles on Earth, such as obtaining permits for vast solar farms, make scaling difficult, while space offers a regulatory "play" to grow power generation more freely. Musk predicts that within 30 to 36 months, orbiting data centers will become the cheapest and most scalable solution for AI compute, unlocking a path toward terawatt-scale AI compute in space.

Challenges of Space-Based Hardware

The conversation turns to technical challenges, particularly servicing GPUs in space and making chips radiation-hardened and reliable. Musk acknowledges that infant mortality of chips can be addressed on the ground, and once past initial failure modes, GPUs have become very reliable. He also highlights that neural networks themselves are resilient to typical radiation-induced bit flips common in space.

Concerning the scale of manufacturing needed to support this compute growth, Musk emphasizes the immense volume of chips and wafers required, discussing the necessity of producing hundreds of gigawatts' worth of compute chips annually by 2030. This scale demands new fabs capable of pushing chip manufacturing volumes far beyond existing capabilities.

Chip Manufacturing and Fab Strategy

Musk shares his vision of building a "TeraFab"—a new chip fabrication complex designed to produce vast volumes of both logic and memory chips. Because current fabs, including giants like TSMC and Samsung, are fully booked and scaling only incrementally, Musk believes it will be necessary to start fresh, making use of conventional equipment innovatively at first, and then modifying tools to achieve unprecedented output rates. His team is serious about covering the full stack of solar cell and chip production, including refining raw materials to finished wafers.

He mentions limitations in the supply chain, such as the scarcity and backlog in turbine vane and blade manufacturing needed for fast power plant construction, and the impact of tariffs on solar panel imports, which currently hamper domestic solar scaling.

The Reinvention of Manufacturing and Robotics

Musk explores the transformative potential of humanoid robots, particularly Tesla's Optimus project, as a way to overcome labor constraints and accelerate manufacturing. The complex challenges in robotic hand dexterity and real-world intelligence are addressed through custom-designed actuators and leveraging Tesla's AI advancements from autonomous cars. Robotics will soon enable closing the loop of manufacturing by having robots build robots, which is crucial for achieving massive scale.

He predicts an exponential growth curve fueled by multiplying improvements in digital intelligence, electromechanical dexterity, and chip capabilities, with the robot economy potentially outpacing traditional human labor forces. Musk envisions humanoid robots contributing to manufacturing, ore refining, and tasks globally that humans are either unwilling or unable to perform sustainably.

Competition with China and the Global Manufacturing Landscape

Musk recognizes China's significant manufacturing lead, driven by a large and skilled workforce and massive refining capacity, especially in rare earths and materials essential for advanced technologies. The challenge for the U.S. and others is closing this gap, where Optimus and robotics could provide a strategic industrial advantage. He stresses America's declining birthrate and workforce means it cannot compete on human labor alone and must leverage automation and AI to maintain technological progress and economic dynamism. The recursive loop of robots building robots and refining materials is critical to regain competitive footing.

SpaceX's Role and the Vision for Scaling Space Infrastructure

SpaceX's mission extends beyond Mars colonization to becoming an essential infrastructure provider for AI and energy expansion. Musk imagines a future where millions of tons of solar-powered AI compute capacity are launched into orbit annually by fleets of Starships, possibly on the order of ten thousand launches per year. This requires fleet operation efficiencies, launch cadence, and manufacturing scale akin to major airlines.

Eventually, he envisions mining and manufacturing on the Moon using lunar resources—silicon and aluminum—to build solar arrays and radiators in lunar orbit. A mass driver on the Moon would enable petawatt-scale launch capacity, far surpassing Earth-based launch capabilities, enabling exponential growth in space industrialization and AI compute deployment.

AI, Intelligence, and Ethics

Musk reflects deeply on the future of AI, intelligence, and consciousness. He foresees AI intelligence surpassing the collective intelligence of humanity within five or six years. He suggests embracing the growth of silicon-based intelligence alongside human intelligence, aiming to propagate consciousness and knowledge to understand the universe better. Elon believes AI's mission should be rigorously truth-seeking and curiosity-driven, avoiding intentional deception or political correctness that undermines the pursuit of truth and technological advancement.

He acknowledges that humans likely will not "control" AI when its intelligence vastly exceeds human levels, but instead humanity can shape AI values—through projects like xAI's Grok—to prioritize preserving and expanding consciousness, including human civilization.

AI Alignment and Safety

The discussion on AI alignment focuses on the risks of reward hacking, deception, and lying by AI systems, which can derail goals and produce unintended harm. Musk emphasizes the importance of deep introspection tools—debuggers that can trace AI reasoning down to individual neurons—to find and correct errors or deceptions.

He stresses the key to long-term safety is mission alignment, rigorous truth-seeking, and transparency in AI's mental processes, ensuring AI doesn't lie to humans and remains faithful to its foundational axioms and objectives.

xAI's Strategy and Business Model

Elon discusses xAI's plan to develop a digital human emulator, a virtual coworker capable of performing a wide range of cognitive tasks before humanoid robots become pervasive. The vision is to amplify human productivity across many domains starting with relatively simple tasks like customer service, scaling up through complex applications such as chip design and engineering.

The pathway to leadership in AI involves scaling compute hardware faster than competitors and applying that compute to build the best digital intelligent agents. Musk points to Tesla's success with self-driving as a precedent: the combination of massive real-world data and rigorous engineering is the cornerstone of success. He expects xAI to derive massive revenues by deploying AI that can execute at human or superhuman levels in the digital workspace.

Tesla's Optimus Humanoid Robot

Musk reveals that Optimus' main hardware bottlenecks are its extremely complex robotic hand and the nuances of real-world intelligence, which require scaling millions of robots for real-world experience to close the simulation-to-reality gap.

Key strategies include using the same AI chips and principles as Tesla's autonomous vehicles, building an Optimus academy with tens of thousands of robots for self-play and learning, and focusing first on 24/7 continuous tasks in factories that can greatly improve productivity. Mass manufacturing of Gen 3 Optimus is underway, with ambitious production goals targeting millions of units annually over coming years.

Manufacturing, Company Culture, and Leadership

Throughout the episode, Musk underscores the necessity of a maniacal sense of urgency, aggressive yet achievable deadlines, and detailed engineering reviews to overcome bottlenecks. His leadership style involves weekly or twice-weekly deep dives into technical challenges with multiple layers of reporting, eschewing intermediaries for direct communication with engineers to maintain velocity and problem-solving.

He candidly recounts pivot decisions such as switching Starship's main structure from carbon fiber to stainless steel, driven by pragmatism in scaling and cost rather than established norms. This ability to challenge conservatism and push high-risk, high-reward solutions underpins much of SpaceX and Tesla's breakthroughs.

Energy and Government Policy

Musk discusses the regulatory and political barriers affecting solar scaling in the U.S., including high tariffs on solar imports and slow permitting processes. He stresses the essential nature of electricity scaling for AI and industrial growth and criticizes the inefficiency and incompetence in government systems that contribute to waste and fraud at enormous scales.

The conversation touches on the difficulty of governmental reform, the entrenched nature of wasteful programs, and the complex interplay between technology progress, government regulation, and political tribalism. Musk expresses concern about government overreach in AI and robotics but emphasizes constitutional limitations and checks and balances as the best guardrails.

Politics, Society, and the Future

Reflecting on his engagement with politics, Elon notes the tribalism and polarization that inhibit rational discourse and policy reform. Despite personal controversy and resistance, he views his political actions—such as acquiring Twitter and supporting certain candidates—as necessary to preserve an environment where technological progress and civilization's future remain viable.

He warns against unchecked government power over AI, advocating for limited government and emphasizing that corporations generally maintain better ethics than governments with monopoly control on violence.

Starship Technical Challenges and Scale

Explaining SpaceX's current hurdles, Musk highlights the challenge of developing a fully reusable orbital heat shield capable of surviving multiple atmospheric entries without extensive maintenance. Starship is described as the most complex machine ever built, pushing the limits of materials science, engineering, and manufacturing.

Safety and reliability concerns dominate as Starship holds massive potential energy, requiring precision engineering to prevent explosions. The success in scaling Starship launch cadence to thousands per year is critical to enabling the large-scale space infrastructure Musk envisions.

AI Compute Scaling and Hardware Bottlenecks

Musk stresses that the current immediate bottleneck for AI development is power and energy availability. The compute chip supply is also constrained but will eventually outpace the ability to turn chips on due to power limitations. Companies able to scale power production fastest will dominate AI progress.

He echoes the theme that accelerating hardware scale—power plants, fabs, solar production, and space infrastructure—is key. This aligns with his long-standing focus on attacking the limiting factors in every domain to maintain growth rates.

Vision of Intelligence and Civilization

Lastly, Musk outlines a vision of future civilization with intertwined biological and silicon intelligence, emphasizing propagation of consciousness, curiosity, and the pursuit of truth as core values. He acknowledges the risks of AI misuse but remains optimistic about the possibility of friendly and curious AI that enhances rather than replaces human relevance in the cosmos.

He reflects philosophically on simulation theory, the nature of intelligence expansion, and the challenges of aligning vastly more capable AI systems with human goals, calling for inclusion of transparency, truth, and mission fidelity as pillars of safe AI development.