Descriptions:
Peter Diamandis hosts a Moonshots podcast episode featuring Will Marshall, co-founder and CEO of Planet — the world’s largest Earth-observing satellite fleet, currently operating 200 satellites generating 25 terabytes of imagery daily and valued at $10 billion, with a 450% stock price increase over the prior year. Marshall introduces the concept of “Large Earth Models,” framing planetary intelligence as the next frontier in machine learning: indexing the Earth the way Google indexed the internet to enable real-time, searchable awareness of global conditions across agriculture, climate change, and infrastructure monitoring.
The conversation covers Planet’s move toward orbital AI data centers, the competitive dynamics of launch costs (the “SpaceX tax” every non-SpaceX operator currently pays) versus compute costs (the “Nvidia tax”), and the physics of satellite lifespans at 400–500km altitude — where atmospheric drag creates a self-cleaning orbit with typical 2–3 year satellite lifetimes that conveniently match GPU depreciation cycles. The panel also discusses Eric Schmidt’s Relativity Space and an upcoming Mars mission, the ongoing AI talent reshuffle across major labs, and Argentine President Javier Milei’s statements on AI personhood.
A major segment focuses on China’s GLM 5.2 open-weight model, which the panelists describe as matching or exceeding top models from OpenAI and Anthropic on several benchmarks — a development called “absolutely shocking” for an open-weights release. Additional topics include Bytedance’s new 4K video generation model and the collapsing price of AI inference alongside exploding capital expenditure, with nuclear and fusion power increasingly cited as the only plausible energy sources for next-generation data centers.
📺 Source: Peter H. Diamandis · Published June 26, 2026
🏷️ Format: Podcast
