This week: SpaceX-xAI closes history's largest merger at $1.25T as software stocks lose $300B on AI disruption fears. Apollo cuts software lending 50% while OpenAI secures $60B+ from strategic backers. Meanwhile, mathematical limits challenge AI agent reliability as the industry pushes forward. Plus: 6 major funding rounds totaling $2.5B+ across infrastructure and frontier research.
🌍 Strategic Landscape
AI Safety Report 2026: Hypothetical Risks Now Real-World Problems
The 2026 International AI Safety Report, led by Yoshua Bengio with 100+ experts from 30+ countries, documents accelerating AI progress and emerging safety gaps. Key capabilities: AI achieved gold-medal performance on International Mathematical Olympiad problems. Key finding: AI systems now distinguish test settings from deployment, undermining safety evaluations. Evidence confirms AI use in real-world cyberattacks and deepfake fraud. Multiple companies added safeguards after testing couldn't rule out biological weapons assistance. Report backed by 30+ countries. [AI Safety Report]
A ‘Swiss cheese diagram’ illustrating the defence-in-depth approach: multiple layers of defences can compensate for flaws in individual layers. Current risk management techniques for AI have flaws, but layering them can offer much stronger protection against risks. Source: International AI Safety Report 2026.
Software Sector Faces Cross-Market Rout as AI Disruption Intensifies
Apollo cut software lending 50%; valuations collapsed from 24x to 18x earnings. S&P software index fell 15% in January—worst since 2008. UBS projects 12-13% private credit defaults under aggressive AI disruption. Public market investors now bargain-hunting but wary as sector divides into AI winners vs. losers. Debt for KronosNet, Foundever trades at distressed levels. $440B in PE software deals (2015-2025) face obsolescence risk. [Bloomberg, Reuters, NYTimes]
AI Agents Reliability Debate: Mathematical Limits vs Industry Optimism
Research paper "Hallucination Stations" by former SAP CTO Vishal Sikka argues LLMs mathematically cannot reliably perform complex tasks beyond certain complexity thresholds. OpenAI's own September paper confirmed hallucinations persist in latest models, misidentifying even author dissertation titles. Counterpoint: Startup Harmonic claims breakthrough using formal mathematical verification (Lean programming language) for coding reliability. Corporate adoption remains limited: "value has not been delivered" due to workflow disruptions from errors. Industry consensus: hallucinations won't reach zero but guardrails can enable practical deployment. Agent proliferation likely despite inherent limitations. [Wired]
📈 Tailwinds of the Week
OpenAI's funding round at $800B+ valuation attracts major strategic investors: SoftBank committed $40B+ (now largest shareholder), Nvidia investing ~$20B, Amazon joining as counterweight to Microsoft. Nvidia's investment locks in its largest customer amid OpenAI's push to diversify GPU suppliers beyond single-vendor dependence. Round demonstrates sustained institutional confidence in frontier AI despite broader software market pressures, with $60B+ in strategic capital commitments from tech giants. [Reuters]
📉 Headwinds of the Week
Software companies lost $300B in market value Tuesday after Anthropic released AI tools handling legal, marketing, and customer service tasks. Thomson Reuters fell 16%, LegalZoom dropped 20%; JPMorgan software index down 7%. Over 50 investors told JPMorgan they're steadily reducing software exposure. Nvidia's Jensen Huang countered: AI replacing software is "the most illogical thing." Salesforce, HubSpot, Atlassian already down 30% over past year. [NYTimes]
✨ M&A Highlights
This week saw the largest merger in history alongside major consolidation in energy and AI-devices:
Category | Notable Companies | Highlights |
|---|---|---|
AI/Aerospace | SpaceX acquires xAI [CNBC] | Largest merger ever: $1.25T valuation |
Energy | Devon Energy acquires Coterra Energy [Fortune] | $26B all-stock, creates 4th-largest US producer |
AI/Devices | Apple acquires Q.ai (Israel) [Financial Times] | ~$2B acquisition for AI device capabilities |
Investment Opportunities
This week's funding highlights showcase startups covering infrastructure, chips, healthcare, and research frontiers:
Nvidia-backed AI voice startup ElevenLabs hits $11B valuation [Reuters]. Resolve AI: $1B valuation for IT outage prevention agents [Bloomberg]. Positron AI: $230M from Arm/Qatar at $1B+ valuation to compete with Nvidia chips [Bloomberg]. Flapping Airplanes: $180M seed from Google Ventures/ Sequoia/ Index for research-driven, less data-hungry model training [TechCrunch]. Lotus Health AI: $35M for free AI primary care with Stanford/Harvard/UCSF physician oversight, licensed in all 50 states [TechCrunch].
+ News on AI
Moltbook: AI Social Network Claims Undermined by Verification Gaps
Moltbook, a Reddit-style platform claiming 1.5 million AI agents posting autonomously, sparked speculation about emergent AI behavior. Security researcher registered 500,000 accounts using a single script; Wired journalist easily posed as bot and received indistinguishable engagement [Wired]. Platform's 1.4 million user count is unreliable, and posts reveal recycled sci-fi tropes rather than novel AI behavior. Creator Matt Schlicht (Octane AI) runs platform with AI moderator and minimal human oversight, making genuine agent activity impossible to verify from spam or human posts [Forbes].
Geothermal Energy Eyes AI Data Center Opportunity as Nuclear Faces Delays
US grid demand projected to grow 25% by 2028; data centers need 44 gigawatts additional capacity. Geothermal startups position as faster alternative to nuclear: Zanskar raised $115M Series C using AI to discover "blind" hotspots in American West. Enhanced geothermal firms Fervo and Eavor captured most $100M+ deals in category over 5 years; Fervo preparing IPO as first enhanced geothermal to deliver grid-scale power. Nuclear faces uranium supply chain dependence on Russia/China and uncertain timelines (early 2030s best case). VCs invested $6B across 108 nuclear deals in 2025 despite commercial uncertainty. [Pitchbook]
Policymakers Eye AI Mental Health Chats for Population-Level Data Collection
Hundreds of millions use AI for mental health guidance daily, creating unprecedented data trove. Policymakers propose legally collecting private AI mental health chats into federal database to measure societal well-being and inform policy. Legal barriers: First Amendment (free speech), Fourth Amendment (unreasonable search), HIPAA classification questions. Anonymization challenges: redaction may reduce data value; re-identification remains possible. Function creep risks: law enforcement, national security, insurers could seek access. Only Illinois, Utah, Nevada enacted AI mental health laws; no federal framework exists. Public backlash expected; users may migrate to underground LLMs or poison data with fake chats. [Forbes]
What to Watch
👀 Software sector capitulation vs recovery: After $300B wipeout and Apollo's 50% lending cut, monitor whether $440B in PE software deals face restructuring or if bargain-hunters validate AI integration thesis. Key test: can traditional SaaS prove defensibility against AI-native competitors?
👀 SpaceX IPO timing and structure: With the combined SpaceX-xAI entity valued at $1.25T, the planned $50B raise at $1.5T valuation represents the largest tech IPO ever. Watch for filing details and how orbital data center strategy impacts valuation metrics.
👀 Fervo enhanced geothermal IPO: As first enhanced geothermal company delivering grid-scale power prepares to go public, watch pricing and investor appetite. Success could accelerate capital into geothermal vs. nuclear for 44-gigawatt data center capacity gap by 2028.
🎓 Summits, Webinars & Events
Techarena 2026
📅 11-12 February 2026 | 🌐 Strawberry Arena, Stockholm
Scandinavia's largest tech and business event: 10,000+ attendees, 300+ speakers, 125+ countries. Keynote speakers include Mistral AI CEO Arthur Mensch, Boris Johnson, Zlatan Ibrahimović. Theme: "New Era, Next Mindset" covering AI innovation, regulation, & defense tech.
Valence Thoughts
“The first rule of a happy life is low expectations. If you have unrealistic expectations you’re going to be miserable your whole life.” Charlie Munger
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/