AI capex vs ROIC: cross-sector value-investing risk
Key Questions
What is the scale of AI capital expenditures?
Annual AI capex reaches $660 billion, with $320 billion in 2025 and $7 trillion total projected. Much risks zero ROI.
Why do 95% of AI projects fail to deliver value?
Projects like Oracle's $523B backlog yield zero FCF; Palantir at $330B valuation questioned by Burry, DB $100M SaaS shorts. Hype exceeds cash flow.
How does AI capex impact ROIC compared to Berkshire?
Mag7 depreciation, Intel negative ROIC, Arm P/E 198x contrast BRK's 20.4% and BHE stability. Munger's ROIC filter highlights risks.
What stresses do AI data centers pose to insurers?
Data centers strain insurers' capacity and credit opacity, echoing 2008 risks; private credit pulls back. Dimon notes trillion-dollar resiliency needs.
How do private vs public markets differ in AI exposure?
Private markets have lighter AI concentration than public. Practical insurer uses lag hype biases.
What is CoreWeave's funding situation?
CoreWeave secured $8.5 billion in loans amid AI infrastructure boom. Survival hinges on adoption phase unwind.
Why is Oracle's AI backlog concerning?
$523B backlog exists but generates zero free cash flow yet. Investors question near-term ROI realization.
What investment lessons from Munger apply to AI?
Munger's ROIC metric filters hype; AI biases mirror past bubbles. Focus on durable value over capex frenzy.
$660B annual/$320B 2025/$7T total vs 95% zero-ROI (Oracle $523B zero FCF/Palantir $330B/Burry/DB $100M SaaS shorts); CoreWeave $8.5B loans/Intel neg/Arm P/E198x/Mag7 depr/BRK 20.4%/BHE; data centers stress insurers capacity/credit opacity 2008 redux/pull priv credit/Dimon trillion$ resiliency; priv markets lighter AI conc vs public; DLR/Munger ROIC/AI biases/insurers practical use; materials sector AI demand noted.