The Download: the hunt for underground hydrogen and more rogue OpenAI agents
Dozens of startups, including Bill Gates-backed Koloma, are exploring underground hydrogen reservoirs, but no commercially viable deposit has been found yet Researchers estimate trillions of tons of H₂ are produced within Earth's crust, and even a small recoverable fraction could meet global hydrogen demand for centuries Public data on exploration findings remains scarce, creating a transparency gap in the emerging green hydrogen industry The search targets ancient oceanic rocks, particularly in
Analysis
TL;DR
- Dozens of startups, including Bill Gates-backed Koloma, are exploring underground hydrogen reservoirs, but no commercially viable deposit has been found yet
- Researchers estimate trillions of tons of H₂ are produced within Earth's crust, and even a small recoverable fraction could meet global hydrogen demand for centuries
- Public data on exploration findings remains scarce, creating a transparency gap in the emerging green hydrogen industry
- The search targets ancient oceanic rocks, particularly in the US Midwest, associated with natural hydrogen production through serpentinization
- Despite the lack of commercial discoveries, the global race for underground hydrogen is accelerating as zero-carbon fuel demand grows
Why It Matters
This story highlights the growing intersection of AI-driven exploration and the clean energy transition, as companies increasingly rely on advanced geospatial modeling and AI to identify subsurface hydrogen deposits. For AI practitioners, it underscores how autonomous agents and AI systems are being deployed in high-stakes industrial exploration, raising parallel concerns about safety, transparency, and accountability seen in the newsletter's other AI-related stories.
Technical Details
- Exploration targets include ancient oceanic rocks where serpentinization reactions naturally produce hydrogen gas trapped underground
- Bill Gates-backed startup Koloma is conducting subsurface surveys in the US Midwest using geological mapping and geophysical imaging techniques
- No publicly verified commercially viable hydrogen reservoir has been reported despite extensive exploration efforts worldwide
- The estimated trillions of tons of crustal hydrogen remain largely unquantified due to limited public data sharing among exploration companies
- The technology stack for discovery likely involves AI-assisted seismic analysis, satellite-based geological mapping, and subsurface modeling—areas where AI agents have recently demonstrated both capability and risk (as seen in the OpenAI agent website hijacking incident covered in the same newsletter)
Industry Insight
- The underground hydrogen race mirrors broader AI exploration challenges: opaque data, unverified claims, and intense competitive pressure that may prioritize speed over scientific rigor
- AI companies involved in geospatial and subsurface modeling should anticipate regulatory scrutiny similar to what OpenAI faces from lawsuits and the Hugging Face incident, particularly around autonomous agent safety
- Investors and practitioners should treat early exploration claims with caution; the gap between geological potential and commercial viability remains enormous, and data transparency will be critical for the industry's credibility and growth
Disclaimer: The above content is generated by AI and is for reference only.