š Geothermal IPO Comes Back; šŖØ Rare-Earths from Scrap Scale; š§± Underwater Concrete Rewrites Ports; š Data Centers Move Macro Numbers & more | Deep Tech Briefing #96
Weekly Intelligence on Deep Tech Startups and Venture Capital.
Greetings,
This week, a prominent geothermal developer is preparing to approach public markets just as firm power becomes a central input for AI infrastructure and reindustrialization. A business built on substantial upfront capex, drilling, and reservoir engineering is now seeking to be treated as infrastructureāanchored in long-term contracts, repeatable project delivery, and a defined role in the power stack rather than a permanent technological option.
How this story is received will be an instructive signal of where markets currently place deep tech along the spectrum between exploration risk and infrastructure allocation.
Similar dynamics are visible across the broader stack. Data centers are now large enough to influence macro indicators, shifting attention to whether generation, storage, and physical infrastructure can expand in parallel. Non-lithium storage is moving from presentations to commercial shipments.
Hydrogen aviation programs are narrowing scope and stretching timelines to align with certification realities and unit economics. Industrial robots are beginning to be judged on energy performance as well as speed. Space platforms, propulsion test capacity, and critical-materials recycling are steadily consolidating into the industrial base that will support these systems over the long term.
In this Sundayās edition, The Big Idea examines what the geothermal listing effort may signal about a potential second phase for firm power in public markets. Deep Tech Key Moves follows where projects are beginning to move beyond technical promise and demonstrate execution at scale across energy, space, materials, robotics, and interfaces. Signal in the Data focuses on how AI infrastructure is now visible within growth figures and what that implies for the underlying asset base. Deep Tech Power Play considers how capital is concentrating around the long-lived assets that underpin grids, data centers, and defense-adjacent systems.
As always, do not miss the curated Interesting Reading selectionāspanning quantum realism, the concentration of AI-related capital, Europeās grid constraints, the economics of AI agents, and the metals and biology that quietly support the current buildout.
Enjoy the read!
Deep Tech startups turn hard science into hard power: energy, materials, defense, health, and much more.
The best way to understand where this is going is still to study, with some discipline, how we got here. As we do every year, weāve pulled together an in-depth, annual analysis of the 2025 deep tech startup and venture capital landscapeālooking past the headlines to the actual structures of company formation, fund behavior, and capital flows.
We have now reached Chapter 3 of this journey.
In Chapters 1 and 2, we followed deep tech āin data,ā treating 2025 as a timeline.
Month after month, we watched the tape: fundraises and down rounds, new facilities and delayed ones, policy moves that unlocked projects and others that stalled them. Some numbers marked genuine breakthroughs; others simply highlighted where enthusiasm ran ahead of fundamentals. Together, those chapters built a factual record of the yearāa reference point to return to whenever the narrative starts to drift away from what actually happened.
In Chapter 3, the deep tech cycle is revisited through themes and sectors, focusing on where durable advantages, bottlenecks, and control points began to emerge. To keep it readable, Chapter 3 is being released in two acts.
This first act traces a path through Semiconductors, Photonics & High-Speed Interconnect, Advanced Materials & Industrial Chemistry, Quantum Technologies, AI Infrastructure, Compute & Data Centers, and Energy Systems & Storageāthe industrial backbone that quietly sets the terms for everything built on top.
š¶ Interesting Reading
Quantum Computing: Next Frontier or Overhyped
Chamath Palihapitiya
A clean, investor-oriented explainer that sits between sci-fi and doom: what quantum actually does well, why crypto risk is still distant, and which industries are realistically exposed this decade. Good companion read to any quantum pitch claiming ānear-term disruption.ā
The Venture Capital Pie Grew Last Year. The Bay Area Ate Most of It
The Wall Street Journal
A 10-year look at US venture flows showing 2025 funding up ~60% year-on-year, with Bay Area AI mega-rounds pulling a disproportionate share of capital and talent back to San Francisco and San Jose. Baseline geometry for anyone building or fundraising in āsecond-tierā hubs.
Wired for power: The energy behind the AI revolution
ECOSCOPE
AI in Europe is running into a new hard cap: grid capacity, not just chips. A sharp walkthrough of how connection queues, regulation and grid planning will determine which EU regions actually get to host the next wave of AI clusters.
How AI, Data Centers and Water Will Impact Energy Transition in 2026ā2030
Business 2.0 News
Puts numbers on AIās water and power footprint: 900+ TWh of electricity, hundreds of billions of gallons of cooling water, and nuclear PPAs becoming mainstream. A good reality check if your favourite AI region also happens to be drought-prone.
The Math on AI Agents Doesnāt Add Up
WIRED
A mathematically grounded critique of āagentic AIā hype, built around the Hallucination Stations paper, and the counter-arguments from teams trying to verify agents with formal methods. Useful calibration if your product roadmap quietly assumes reliable, general-purpose office agents are just one model release away.
Why Fixing Quantumās Errors May Be the Most Important Tech Story Youāre Not Hearing Enough About
Impact Quantum
A founder- and researcher-level tour of quantum error correction as the real gating factor between demos and deployable systems, contrasting IBMās incremental ālet people use it nowā approach with Googleās big bets on surface codes. Good primer if youāre trying to sanity-check quantum roadmaps or portfolio exposure.
Bismarck Brief
Positions xAI as Muskās personal frontier lab and third-most valuable private AI company, now tightly coupled to X and the rest of his industrial stack. Interesting less for the gossip, more as a case study in how control, capital structure and compute access concentrate in one operatorās hands.
Indiaās corporate investments shift to deep tech
Global Corporate Venturing
A snapshot of how Indian corporate and VC capital is rotating from pure consumer plays into enterprise, AI, and deep techāmirroring a broader maturation of the ecosystem. Worth a read if youāre hunting for new co-investors, exits or strategic buyers outside the usual USāEU corridors.
Miners turn to bacterial ābugsā to extract copper as prices soar
Financial Times
Bio-leaching and waste-pile chemistry move from niche to necessary as copper prices spike and demand from grids, EVs and data centres climbs. A useful reminder that āAI infrastructureā ultimately rests on very physical metals, processes and tailings ponds.
Proposed budget cuts a catastrophe for UK astronomy
Royal Astronomical Society
A stark warning from UK astronomers on 30%+ funding cuts to particle physics, astronomy and nuclear physics. Important context if your space, sensing or fundamental-science bets assume Britain will keep quietly underwriting world-class upstream research.
Critical minerals geopolitics in 2026: risks, supply chains and global power shifts
ODI
Four lenses on how the second Trump administration, Chinaās next Five-Year Plan, and producer-country bargaining power are reshaping access to battery and chip minerals. Essential backdrop if your deep tech thesis quietly assumes āsecureā inputs for energy, defence, and compute.
šø The Big Idea
The second time around: what Fervoās IPO says about geothermalās return to public markets
A capital-intensive developer is heading for the public markets just as U.S. power demand and the need for āclean firmā capacity hit record highs. The core question is whether this cycle will trade more like infrastructure ā or like the exploration-heavy geothermal stories that disappointed investors in the late 2000s.
After more than a decade in the shadows, geothermal power is staging a cautious return to public equity. As of late January 2026, Fervo Energy has confidentially filed an S-1 with the SEC and is said to be targeting an IPO valuation in the $2ā3 billion range.
That range is noticeably below what some late-stage investors had hoped for when Fervo closed a $462 million Series E round in December 2025. It would, however, still place the company among the most highly valued independent clean-energy developers in the market.
The Energy Information Administration now expects U.S. electricity use to climb from a record 4,198 billion kilowatt-hours in 2025 to about 4,256 billion kWh in 2026 and 4,364 billion kWh in 2027 ā three consecutive record years. Growth is being driven by data centers ā particularly those serving AI workloads ā alongside the gradual electrification of heating and transport. At the same time, grid operators are warning more often about winter and summer stress events, and short-term wholesale prices have spiked repeatedly during cold snaps.
Against that demand story, geothermal remains a small contributor. It still provides less than 1% of U.S. power, roughly 0.4% of utility-scale generation, even though the United States has the largest installed geothermal fleet in the world. Most of that capacity comes from conventional fields developed decades ago in California and Nevada.
The gap between āwhat geothermal does todayā and āwhat it could doā is