KI & Börse: Dein Depot ist nicht dein größtes Risiko

Core thesis

• The presenter argues that AI’s biggest risk to individuals is not stock selection inside your portfolio, but the concentration of overall wealth in “human labor” and pay-as-you-go retirement systems.
• If AI shifts value creation away from wages, these systems may come under pressure.
• Separately, AI may reduce achievable “alpha” by:
  • making markets more efficient, and/or
  • increasing fragility via synchronized models, which can erode diversification benefits.
• The implication: investors may be pushed toward broad diversification and market-average exposure.

Sources / framing used (interest-rate and macro logic)

The video frames its argument as an “early warning system” based on real interest rates, suggesting markets would reflect short timelines for disruptive AI (labeled “HI” = short-term high-income / breakthrough timelines).

Paper: “Transformative AI, Existential Risk and Real Interest Rates” (authors referenced in subtitles)

• Logic:
  • Saving acts like consumption smoothing.
  • If extreme future scenarios are plausible (e.g., abundance vs ruin), people may save less / borrow more.
  • That behavior would require higher real interest rates to restore equilibrium.
• Conclusion:
  • If markets truly believed in short-term extreme outcomes, real interest rates would need to be much higher today.

Model: “Mareska” (subtitle: model for 2025)

• Mechanism:
  • Households may “hoard assets” if the level of wealth at the AI breakthrough determines how much automated value they later control.
• Result in the model:
  • Interest rates rise to 10–16%, versus about ~3% in the base scenario.

Key market observations / numbers mentioned

• Real interest rates in the US and Germany are described as being in the low single digits.
• Therefore, the bond market suggests low probability of very short AI “high-income” timelines—not zero, but low.

Methodology / “framework” presented

1) Interest-rate test (macro valuation sanity check)

• If markets price short-term AI breakthrough timelines that significantly change consumption, then real interest rates would need to be much higher.
• The presenter treats current real rates as an update, not hard proof.

2) Market efficiency / alpha-eroding mechanism

• Uses the argument associated with Grossmann–Stiglitz:
  • Markets can’t be perfectly efficient; some inefficiency must remain.
• AI can behave like “real-time publication” of information, accelerating the removal of inefficiencies.
• Conceptual supporting evidence:
  • Cited idea that post-publication anomalies tend to weaken (subtitles cite McLean and Bonif), and AI may accelerate this.

3) Crisis synchronization / systemic fragility

• Cites a concept from the New York Fed (subtitles reference Lenzu):
  • If models become similar across participants, stress can increase correlated trading (described as “model monopolization” / synchronization).
• Whether the outcome is more efficient or more synchronized, the practical implication for individuals is similar:
  • avoid concentrated bets
  • prefer diversification

Tickers / assets / instruments mentioned

• No specific stock/ETF tickers were explicitly provided.
• Market/index references:
  • DAX
  • ATX
  • European industries, global equity, “true transformation winners,” broad market index
• Company examples mentioned (names only):
  • Apple, Microsoft
  • Nvidia
  • Cisco
  • Pets.com
• Instruments / systems:
  • Bonds, including short-term government bonds
  • Pay-as-you-go pension / retirement income (financing system, not an instrument)
• Infrastructure sectors framed as “where winners might be”:
  • Chips
  • Data centers
  • Energy

Performance/risk statistics & claims (numbers explicitly stated)

Active management underperformance

• In Europe, 97% of active funds underperformed benchmarks over the last 10 years (as stated in subtitles).

Skewness of equity returns (“winner-take-most”)

• Based on a study attributed to Bessenbeinder (as described in subtitles):
  • The majority of stocks returned less than short-term government bonds over their lifetimes.
  • Net wealth creation above the safe return comes from about 4% of companies.
  • Therefore, return distribution is highly skewed, and stock picking resembles a winner-take-most dynamic where losers drag.

Human-capital framing

• Example calculation (described as conservative):
  • Age 30
  • €3,000 net/month
  • 35 years of earned income ahead
  • Implies roughly ~€1.2 million human capital (not adjusted for inflation or raises).

Retirement system framing

• Pay-as-you-go pension described as a claim on future wages.

Explicit recommendations / portfolio implications

General conclusion

• Use diversified equity / market-average exposure for the long term, and avoid concentrated bets.

Scenario-based portfolio implication (as presented)

1. If you believe short-term AI breakthrough (“HI”)

   • Don’t concentrate into one stock.
   • Increase the savings rate and maximize the capital share of your balance sheet.
   • Diversify globally, since wealth at breakthrough matters more than the specific company.

2. If you think AI is normal tech spreading gradually

   • Standard advice applies: equity, time in the market, lower costs.

3. If you’re agnostic

   • Choose a portfolio that doesn’t require being right:
     • globally diversified equity
     • pay attention to human capital concentration risk

Cautions mentioned

• Don’t assume stock-picking with a chatbot will beat sophisticated markets.
• Don’t assume politics will fix retirement systems in time (framed as: relying on politicians to solve a century-long problem).
• Don’t equate being “right” about the technology with making money (examples mentioned: Cisco, Pets.com).

Disclosures / disclaimers

• Presenter states they don’t have a HI prognosis (no confident timeline prediction).
• Conflict disclosure:
  • “No commissions” and no affiliate-like setups:
    • no affiliate links
    • no Telegram group
    • only small YouTube revenue
• Financial-statement style disclosure:
  • Their view is labeled as “assessment and opinion”
  • They state they’re not aware of studies supporting some parts of the claim.

Presenter / sources mentioned (end of video)

• Presenter: Marcel (introduced as “Marcel, an honorary consultant at Capskale”).
• Named sources/concepts:
  • Grossmann and Stieglitz
  • Ciao Halperin and Masslich (paper cited: Transformative AI, Existential Risk and Real Interest Rates)
  • Mareska (AI wealth-hoarding model; 2025 scenario)
  • Paul Samuelson (“Markets are micro-efficient but macro-inefficient”)
  • McLean and Bonif (anomalies lose excess returns after publication; cited conceptually)
  • Lenzu (New York Fed; subtitles reference a downside “in 2026” and model monopolization/synchronization)
  • Trammel and Corineck (overview/empirical research on technology scenarios)
  • Bessenbeinder (return skewness; ~4% drive wealth creation claim)