Here’s the problem with most direct air capture: you spend enormous amounts of energy releasing the CO₂ you just captured. Heat the sorbent to 900°C (solid sorbents) or boil a solvent (liquid systems) — either way, the energy cost dominates the economics. That’s why DAC still costs $400–$1,000 per tonne. But what if you could release captured CO₂ by just… making the air humid? Researchers at Arizona State University, led by doctoral researcher Gayathri Yogaganeshan, have published new work explaining exactly why certain commercial polymers can capture CO₂ from dry air and release it when exposed to moisture. No heat. No pressure. Just wet/dry cycling. And for the first time, they’ve mapped the structural differences that make some materials dramatically better at it than others. ...

Microsoft just signed what might be the largest carbon removal contract in history — and nobody in Washington had anything to do with it. E&E News reports that Microsoft has agreed to store 4.9 million tonnes of CO₂ with Vaulted Deep, a Houston-based startup that uses geological storage. Neither party confirms the price, but industry experts estimate the deal could be worth approximately $1 billion. Vaulted Deep CEO Julia Reichelstein called the deal’s impact “Revolutionary.” ...

While the US spent 2025 freezing $3.5 billion in DAC hub funding, the Netherlands quietly started building the kind of CDR infrastructure that actually scales. Platform Zero just organized a CDR strategy session in South Holland with an attendance list that tells you everything about how Europe does industrial policy differently. At the table: Skytree (DAC), SeaO2 (ocean alkalinity enhancement), TNO (the Netherlands’ heavyweight national research organization), Provincie Zuid-Holland, Gemeente Rotterdam, the Port of Rotterdam, InnovationQuarter, the Netherlands Enterprise Agency, Paebbl, Carbyon, everox, Zero Emission Fuels, EBN B.V., and Division Q. ...

The CO₂ in your beer probably came from an ammonia plant. Or an ethanol refinery. Or — if you’re drinking Almanac Beer Company’s new “Flow – Clean Air Edition” — straight from the sky above a parking lot in Alameda, California. Almanac just launched what appears to be the world’s first commercial beer carbonated entirely with CO₂ captured from ambient air via direct air capture. Their partner, Berkeley-based Aircapture, installed a modular DAC unit in the brewery’s parking lot — it looks like an oversized HVAC unit with a chimney — that pulls CO₂ from the air and liquefies it to beverage-grade purity (99.999%). The beer is already available at 800+ retailers across California, including Safeway, Whole Foods, Total Wine, and BevMo. ...

📊 v2 Update (March 23, 2026) This census now uses LLM classification (Gemini Flash) instead of keyword-based search. Every paper was re-classified by title + abstract across our 7 CDR pathways. Starting from 37,133 candidate papers, we filtered 12,384 as NOT_CDR, leaving 24,749 high-confidence CDR papers. Researcher count updated to 122,674. Pathway assignments recalculated. Full methodology below. This is more rigorous but catches different papers — earlier keyword searches found 21,804. ...

📊 v2 Update — LLM-Based Reclassification This post updates v1 with fresh census data (80,382 researchers, 24,749 papers). The methodology has shifted from keyword-based to LLM-based classification of research papers — meaning we’re now using AI to understand semantic context instead of just matching terms. This catches CDR research that traditional keyword methods miss. Country and institution assignments use ORCID self-reported affiliations where available, falling back to OpenAlex’s “last known institution” otherwise. Still not perfect, but significantly more accurate than before. Your corrections still make v3 better. Bluesky · X ...

⚠️ Updated for v2 (2026-03-23) This post now uses LLM classification (title + abstract) instead of keyword search. That’s a major upgrade — but it exposed a real problem with DAC data that I need to flag transparently. Rankings are based on number of CDR papers found in our classification. Institution and country data uses ORCID self-reported affiliations where available (66% coverage), with OpenAlex as fallback. ORCID links let you verify every name. If you see errors, tell me — I’ll fix them. Bluesky · X ...

Update: V2 Data (March 2026) We’ve reclassified 122,674 researchers using an LLM-based methodology instead of keyword-only matching. 42,292 researchers lost their CDR designation because their papers weren’t actually about CDR — they were tangentially related but didn’t cross the threshold. This leaves 80,382 researchers with a genuine CDR pathway. The dabbler percentage dropped from 69% to 52.8%, which at first looks like bad news. But the real story is subtler — and slightly more hopeful. Previous version: first shot · Feedback? ...

⚠️ v2 Update — Improved Methodology This analysis now uses LLM classification instead of keyword filtering, which reclassified ~2,945 papers and shifted author trajectories. The core finding holds: CDR science is growing, but the talent retention story remains fragile. Bluesky · X This is Part 5 of the CDR Researcher Census series — the synthesis. This entire census project started with a question: Is CDR science growing fast enough to deliver gigaton-scale removal by 2050? ...

If you’ve published anything related to carbon dioxide removal, you’re probably in here. The CDR Researcher Lookup lets you search 122,674 researchers by name. Each profile shows: Your CDR pathway — which removal method your work maps toCommitment level — dabbler, part-time, focused, or dedicatedTrajectory — is your CDR output growing, stable, or declining?Career stage — early career through eminent, based on h-index and career spanInstitution and country — verified via ORCID where availableYour CDR papers — the specific publications that put you in this datasetWhy This Exists#When Lück et al. (2025) mapped 53,000 CDR papers in Nature Communications, they answered “how much CDR research exists?” I wanted to answer the next question: who are the people doing it? ...