📨 More Investment, Questionable Returns

Big Tech's AI spending hits $600B in 2026, increasingly debt-funded. Private credit faces turmoil as software loan defaults rise. Harvard study challenges AI productivity assumptions. Meanwhile, VC concentration in AI reaches $1T, raising systemic risk concerns.

Anthropic's Claude Opus 4.5 achieved significant performance gains on METR's widely-cited AI capability benchmarks, though actual results carry substantial uncertainty (2-20 hour range vs. reported 5 hours). METR, which evaluates frontier AI systems, tests primarily coding tasks. Researchers caution against overstating capabilities, as metrics measure task difficulty, not workplace automation readiness or revenue potential. [MITTechReview]

U.S. venture capital funds now hold over $1 trillion in AI-related investments: the highest level ever recorded. AI companies represent 40% of all startup market value. However, experts warn of systemic risk: too many startups are competing in crowded sectors like software, life sciences, and finance. If valuations prove inflated, the concentrated bets could trigger significant losses across the venture capital market. [WSJ]

Harvard Business Review study of 200 employees over 8 months found AI adoption led to expanded job scope, longer hours, and increased multitasking rather than reduced workload. Workers took on unfamiliar tasks and worked after hours. Engineers spent more time reviewing AI-assisted code. Findings challenge productivity gain assumptions underlying corporate AI investments.

Big Tech will invest over $600B in AI infrastructure in 2026, led by Amazon's $200B commitment. Companies are shifting from cash reserves to debt financing, with potential borrowing of $140B annually through 2028. Spending primarily targets servers and chips, benefiting suppliers like Nvidia. Wall Street projects plateau by 2027 amid profitability concerns. [Reuters]

Blue Owl Capital's stock dropped 50% over one year as investors withdrew funds. Software companies (comprising 20% of private credit loans) face AI disruption risks. Recent bankruptcies and write-downs at BlackRock and Apollo exposed opacity in the $3 trillion sector. Analysts note limited visibility into private borrowers' financial health raises contagion concerns. [NYT]

Recent M&A activity reflects strategic positioning across AI infrastructure and security, with deals ranging from semiconductor connectivity to identity protection and retail AI platforms:

Waymo secured $16B funding at $110B valuation, doubling from previous round, with new investors including Dragoneer, Sequoia, and DST Global. Expansion planned for U.S., London, and Tokyo amid Tesla robotaxi competition [FT]. Alphabet reported $400B annual revenue: services division $350B, Cloud $70B run rate, YouTube $60B (exceeding Netflix's total business). Day AI raised $20M for AI-powered CRM platform that analyzes emails, documents, and messages to automate data entry and provide business insights; co-founded by former HubSpot CPO Christopher O’Donnell. Crypto.com acquired AI.com domain for $70M in cryptocurrency (record domain purchase) signaling brand positioning ahead of Super Bowl advertising and potential AI sector expansion [TechCrunch].

Andreessen Horowitz doubled federal lobbying spend to $3.5M in 2025, becoming primary advisor to Trump White House on AI policy. Firm co-founder Marc Andreessen donated $2.5M to Trump, maintains regular contact with president. A16z blocked federal AI safety standards and helped draft executive order deterring state regulation. Critics note firm holds stakes in 10 of top 15 VC-backed companies. [Bloomberg]

China is implementing artificial intelligence to detect and prevent corruption in public project bidding. The AI systems will analyze tender documents, monitor review committees, and identify suspicious activities like bid-rigging. This initiative follows a successful case in Zhejiang province, where AI helped uncover bribery, leading to an official's imprisonment. [Reuters]

NASA's Perseverance rover has successfully completed its first drive on Mars using routes planned entirely by artificial intelligence. The AI analyzed terrain data to identify hazards and plot a safe course, which the rover then followed autonomously. This marks a significant step in robotic space exploration technology. [NASA]

👀 AI benchmark interpretation: Distinguish between technical capability metrics and actual workplace automation potential when evaluating AI company claims and investment theses.

👀 VC concentration risk: Monitor startup valuation corrections in crowded AI sectors (software, life sciences, finance) as $1 trillion exposure creates systemic vulnerability.

👀 Productivity paradox evidence: Track whether AI workplace studies continue showing workload expansion rather than reduction, potentially impacting enterprise adoption rates and ROI projections.

MIT Technology Review's AI conference for executives and researchers. Theme: "The Implementation Era" - transitioning from AI pilots to production (95% pilot failure rate noted). Features curated sessions on AI strategy, regulation, agentic systems, and breakthrough research. Intimate format with accessible speakers for meaningful networking beyond sessions. Join in-person or online.

Valence Thoughts

Lewis Carroll on the Red Queen effect: 

"Here we must run as fast as we can, just to stay in place. And if you wish to go anywhere you must run twice as fast as that."

Don’t run harder -  run smarter.

That’s it for this week.

Until next time,

The CLNM Capital

[Edited with the assistance of LLMs]

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