📨 AI's Climate Dilemma & The End of Solo AI

AI investment is surging beyond automation. OpenAI’s revenue hit $20B, while Humans& secured a $4.5B valuation for human-centric collaboration. The market is shifting toward high-utility, specialized applications that empower rather than replace human expertise.

AI's climate impact defies simple categorization. Energy costs are measurable: data centers consume 4% of U.S. electricity, potentially doubling by 2030. Training GPT-3 used 700,000 liters of freshwater. Meta's $10B Louisiana facility runs on new natural gas plants. However, Google's DeepMind increased wind power grid value 20%, and AI flood alerts saved $200M in Bangladesh crops. ECMWF operates AI models outperforming traditional forecasts. Key uncertainty: emerging technologies like Emerald AI's GPU optimization could drastically alter demand projections — Boston Consulting estimates range 200-1,000 terawatt hours by 2030. [Bloomberg, Forbes, Wired, MIT Tech Review]

Microsoft CEO Satya Nadella at the WEC in Davos warned that AI becomes a bubble if only tech firms invest without workflow transformation. He said AI must deliver improvements people can feel in areas like health and education, otherwise it risks losing public support for the energy needed to run it. Only 10-12% of companies report AI revenue benefits; 95% of pilots fail. Nadella argues large firms must flatten hierarchies and redesign workflows, or smaller AI-native competitors will achieve scale faster. [Fortune]

Big Tech will spend over $300B on AI infrastructure this year, but investors are demanding actual profits. So far, revenue comes from cloud services and ads, not AI itself. Apollo Global warns of a $1-1.5T funding gap between AI spending plans and available capital. OpenAI's reported $100B fundraising attempt will test investor confidence. If AI disappoints, falling stock prices could hurt consumer spending and economic growth, creating political pressure ahead of midterm elections. [Politico]

Europe needs €3T over 5 years for AI, data centers, and energy infrastructure but lacks long-term investment capital. The US has securitized $63.6B in data center debt since 2018; Europe managed just $0.8B. Strict insurance regulations prevent European insurers from investing in these projects (they hold only 0.4% in securitizations versus 17% for US insurers). Private credit firms are pivoting to infrastructure financing, and regulatory reforms are under review, but current proposals fall short of meeting Europe's massive investment needs. [The Financial Times]

TSMC, the world's largest chipmaker, reported record earnings and expects nearly 30% revenue growth in 2026, spending up to $56B on expansion. CEO C.C. Wei confirmed strong AI chip orders after verifying cloud providers' AI investments generate real business value: "It looks like it's going to be endless" [ArsTechnica]. Separately, Taiwan agreed to invest $250B in U.S. chip production under a new trade deal, with the U.S. limiting reciprocal tariffs to 15% [CNBC].

SaaS stocks are posting their worst start to a year since 2022, with companies like Adobe, Salesforce, and ServiceNow down 18-45% despite strong earnings. The threat is real: Anthropic built its Cowork productivity tool in just 10 days using Claude Code (work that previously took months) [Forbes]. Companies are questioning expensive subscriptions when they can build custom solutions themselves. Morgan Stanley's software basket now trades at record-low 18x forward earnings [Bloomberg]. Exception: Companies controlling critical data ("systems of record") are thriving as AI agents need their data to function [Morningstar].

This week’s activity highlights a focus on energy efficiency, governance, and service automation to drive operational scale.

Human-centric AI is attracting massive early-stage capital, evidenced by Humans&'s $480M seed round at a $4.5B valuation [Reuters]. Meanwhile, the AI-powered climate tech sector is experiencing a record surge in VC funding and megadeals, signaling a robust and growing market [Pitchbook]. Established players like OpenAI also demonstrate explosive revenue growth [OpenAI], underscoring the sector's overall financial power and potential for practical, enterprise-level adoption.

The UK's ARIA agency received 245 proposals for "AI scientists”: systems that design experiments, run tests, and analyze results independently. ARIA funded 12 projects with £500K each after being overwhelmed by high-quality submissions. These AI systems handle repetitive lab work while human researchers oversee strategy. Projects include robot chemists and battery research tools. However, current AI agents still fail 75% of the time when running full scientific workflows, often taking shortcuts or declaring false success. [MIT Tech Review]

Anthropic's Economic Index report introduces "economic primitives” — 5 key measures tracking how AI affects work: task difficulty, education level needed, purpose (work/school/personal), AI independence, and success rates. Key findings: AI speeds up complex, college-level tasks 12x faster than simpler work. Currently, AI augments rather than replaces workers, with 52% using it as a thinking partner— mass job loss fears don't match the data yet. However, the real risk is for junior employees: if AI handles entry-level tasks used for training, an entire generation may lack crucial work experience, creating a new economic challenge. [Anthropic Economic Index Report]

Stanford researchers developed SleepFM, an AI model that analyzes one night's sleep to predict over 130 health conditions. Trained on 585,000 hours of sleep data from 65,000 people, it achieves 84% accuracy predicting mortality risk, 85% for dementia, and 81% for heart attacks by detecting subtle patterns in brain waves, heart activity, and breathing. As wearable technology improves, sleep tracking could become a powerful tool for early disease detection, transforming how we monitor health through everyday devices. [Nature Medicine]

👀 The Infrastructure-Revenue Correlation: Monitor the tightening link between compute capacity and Annual Recurring Revenue. As OpenAI’s growth tracks directly with GW expansion, infrastructure is no longer just a cost, it is the primary revenue driver.

👀 The Autonomy vs. Collaboration Schism: Watch for a strategic divide in capital allocation. While major labs chase full autonomy, a new "human-centric" wave is emerging, prioritizing AI as a coordination layer for complex human teams.

👀 Vertical Consolidation in High-Stakes Sectors: Expect intensified M&A and investment in specialized AI for energy, governance, and health. The market is moving from general-purpose models toward deeply integrated, sector-specific "agentic" solutions.

The world's largest tech conference returns to the Middle East. With AI dominating the agenda: expect insights into how Middle Eastern sovereign wealth funds are deploying capital in AI. CLNM Capital CEO Mian Manoff will be attending and available for meetings with portfolio companies and partners.

If you missed the International AI Summit, catch up with the Report & Video Highlights: 250+ global leaders gathered in Brussels on Dec. 11 2025 to tackle AI governance, geopolitics, and market dynamics.

"If your views on other things can be predicted from your views on one thing, you need to be very careful that you’re not in the grip of an ideologue.” Kevin Kelly

That’s it for this week.

Until next time,

The CLNM Capital

[Written with the assistance of LLMs]

Want to share this newsletter with a friend? Use this link: https://clnm.beehiiv.com/