KI-Blase wird PLATZEN! Aber ANDERS als die meisten denken - Dr. Andreas Beck & Gregor Broschinski

Main Argument

The discussion argues that the main risk in the AI boom may be debt and concentrated investment exposure, rather than an immediate collapse in the underlying technology. AI could generate major economy-wide efficiency gains, but it remains uncertain which companies will capture the profits.

The speakers compare today’s boom with the dot-com era: the technology may transform the economy even if many early leaders or heavily funded companies ultimately fail.

AI Valuations, IPOs, and Investment Risks

• Andreas Beck says AI adoption has not yet translated broadly into company profits. Much of the current activity involves investment in models and data-centre hardware. The key test will be whether businesses can use AI to raise productivity and earnings.
• Gregor Broschinski sees a concentration of risk rather than a classic bubble. Leading AI-related companies have substantial revenue growth, he says, but revenue growth does not necessarily translate into durable profitability.
• Ambitious market-size claims—such as a $30 trillion potential market based on cognitive work—do not mean that any particular AI provider can capture that value. Efficiency gains may be distributed among customers and across the wider economy.
• Beck is skeptical that the valuations proposed for OpenAI, Anthropic, and SpaceX will necessarily be sustainable a decade from now. Broschinski cautions investors against feeling obliged to subscribe to high-profile IPOs.
• The speakers compare the AI boom with the dot-com era. Both involve transformative technology and uncertainty about the winners. One difference, they say, is that many current public-market leaders have substantial existing businesses and profits. Still, rapid innovation could make today’s leaders obsolete and lead to large investment write-offs.
• Broschinski cites Nvidia as being worth roughly twice the combined value of the DAX at the time of the discussion. He describes its P/E ratio as roughly 14–18, depending on assumptions about future earnings growth. These are figures cited in the interview.
• The discussion cites an Anthropic revenue run rate of about $60 billion, said to be around seven times its 2025 revenue. The speakers question whether current sales can sustainably cover the sector’s enormous capital spending, depreciation, and financing costs.

Debt and Financing Risks

The speakers identify debt-financed infrastructure spending as a potential weak point in the AI boom:

• Hyperscalers were expected to issue about $250 billion of bonds in 2026, competing for capital with heavily indebted governments.
• The guests refer to substantial borrowing routed through special-purpose entities, potentially amounting to trillions across the sector. They describe a possible feedback loop in which suppliers help finance customers’ purchases.
• This arrangement is sustainable only if growth and cash flows are sufficient to cover interest and principal repayments. A slowdown or higher financing costs could expose weaknesses in the structure.
• Broschinski says a further one-percentage-point rise in interest rates could push some 10-year hyperscaler bond yields to around 6–7%.
• He also cites SpaceX as an example of volatile private-market pricing: reportedly trading slightly above its IPO price but around a third below its peak. The transcript’s wording around the relevant valuation is unclear.

Portfolio Construction and Investment Approach

Beck’s core recommendation is to avoid trying to predict the winners of a fast-moving technology cycle and instead diversify broadly:

• Invest across the global economy, rather than making concentrated bets on a handful of AI-related companies.
• Recognize that an MSCI World portfolio already has substantial exposure to the largest technology stocks. Broschinski estimates that roughly 25% is concentrated in seven leading companies.
• Consider broader diversification through:
  • Small-cap stocks, including an MSCI World Small Cap index.
  • Emerging markets.
  • Multifactor strategies that include factors such as value, size, and quality, rather than relying only on market-cap weighting.
  • A bond allocation, including government bonds, alongside equities.
• Broschinski suggests researching the broader AI value chain—not only the largest technology names. Areas mentioned include companies that adopt AI effectively, data businesses, storage providers, copper, energy infrastructure, and potentially quantum-computing companies. These are areas to investigate, not specific buy recommendations.
• Both guests stress that uncertainty around innovation cycles makes broad diversification especially important.

Inflation, Deflation, and AI Infrastructure

The guests distinguish between near-term infrastructure costs and longer-term productivity effects:

• Beck argues that productivity gains are inherently deflationary: cheaper and more efficient production can lower prices and increase what people can buy with their income. He says government spending, regulation, and bureaucracy can offset those gains.
• Broschinski expects inflationary pressures during the build-out phase, potentially followed by deflationary effects once the infrastructure is established.
• He points to demand for copper, memory chips, energy, and gas turbines as possible sources of near-term price pressure. He cites an estimate of about 50,000 tonnes of copper per gigawatt of data-centre capacity and says Siemens Energy is fully booked for several years.
• Beck identifies debtors—especially governments—as vulnerable in a deflationary environment because the real burden of debt rises. He says holders of financial assets, including cash and pension claims, could benefit from deflation relative to inflation.
• Broschinski argues that highly indebted governments have an incentive to tolerate inflation as a way of reducing the real burden of debt. He therefore doubts that deflation will be allowed to persist unchecked.
• The guests also discuss high government borrowing costs, including a French 10-year bond yield of roughly 4.5% as cited in the conversation, and the effect of a €1 trillion German special-debt package on long-term borrowing costs.

Real Estate

The guests see a more challenging outlook for offices than for housing:

• Broschinski says residential property is less directly exposed to AI because people will continue to need homes, though demographics and local supply still matter.
• He cites office vacancy rates of roughly 8.5–10% and says forecasts could take the rate toward 15% by 2030. Combining remote work, AI-related reductions in office employment, and demographic factors, he says vacancy rates of 20–30% are possible in some scenarios.
• He sees the strongest prospects in prime central locations, particularly in Munich, and advises caution toward peripheral and lower-tier office locations. Prime Munich rents of €60–70 per square metre were cited.
• Beck is more concerned that AI could make significant portions of today’s office stock obsolete, creating substantial write-down risks, including in prime locations.
• Broschinski adds that higher interest rates and tighter bank financing standards make refinancing office properties more difficult. Conversion of some offices to housing may be possible, but is not straightforward.

Employment and Economic Implications

• Beck cites estimates that 20% of jobs could be eliminated and questions whether the eventual impact might be larger. The discussion presents AI as potentially automating substantial amounts of knowledge and administrative work.
• Broschinski expects fewer workers to be needed in office-based fields, but also anticipates continued demand for trades and physical work, such as construction and maintenance.
• Both guests discuss uncertainty over how productivity gains and income will be distributed. Beck argues that a basic income cannot eliminate scarcity or determine who gets access to scarce goods. Broschinski says government reforms or new fiscal approaches may be needed if employment and tax receipts fall significantly.
• The conversation also highlights possible social and regulatory responses to concentrated AI power, liability risks, and job displacement.

Germany and Policy

The guests call for reforms to make Germany more competitive as a business location:

• Broschinski says Germany has fallen from the top five in international business-attractiveness rankings to around 24th–30th. He proposes a clear goal of returning to the top ten.
• Beck calls for a renewed national economic vision, spending restraint, and coordinated action among political parties, unions, and employer groups.
• They warn that a continued loss of competitiveness could mean fewer jobs, lower tax receipts, and less innovation.

Assets, Instruments, and Companies Mentioned

• Equities and companies: Nvidia, Meta, SpaceX, OpenAI, Anthropic, Siemens Energy, Google, Microsoft, Nestlé, Uber, WeWork, and Standard Oil.
• Indices and investment products: MSCI World, MSCI World Small Cap, DAX, global equity ETFs, and multifactor portfolios.
• Other assets and sectors: Government and hyperscaler bonds; copper; memory chips; gas turbines and energy infrastructure; data centres; data businesses; storage providers; quantum computing; residential and office real estate.

No explicit stock ticker symbols are given.

Disclosures and Cautions

• No explicit “not financial advice” disclaimer appears in the subtitles.
• The speakers caution against relying on high-profile IPOs, concentrated exposure to a few large technology companies, or headline revenue and market-size forecasts without evaluating their sustainability and financing requirements.
• Several figures are estimates or claims made during the discussion. Some company names and numbers in the auto-generated subtitles are unclear.

Presenters and Sources

Presenters: Leonard Schmedding (host), Dr. Andreas Beck, and Dr. Gregor Broschinski.

Sources or references cited: McKinsey (automation estimate); David Graeber (discussion of “bullshit jobs”); Alex Karp, likely the person referred to in the subtitles as “Alex CP.”