Strategy Update: November 2025
We did the math: USD 10 Billion in Losses, USD 500 Billion Valuation, 20 Nuclear Reactors – AI Breakthrough or Madness?
REVIEW
THE FINANCIAL MARKETS IN OCTOBER
Earnings season is in full swing and American companies are surprising with impressive figures. 87% of large corporations are reporting profits above expectations – a quota that hasn‘t been achieved since 2001. The picture for revenues is similar, with 83% positive surprises. Profits are growing by 9.2% in the third quarter compared to the previous year, which already means the ninth consecutive quarter with profit growth. Also unusual is the fact that profit expectations have been adjusted downward only minimally this year. Particularly remarkable is that profit margins remain stable at 12.8% despite concerns about tariffs and rising costs – that‘s better than the five-year average. The IT sector leads growth at 22%, driven by the semiconductor industry with 48% growth. The financial sector follows with 20% growth across all sub-sectors. The energy sector is suffering from falling crude oil prices and reports a profit decline of -4%. For stock valuation, it should be noted that the price-earnings ratio at 22.7 is significantly above the historical average. For the coming quarters, analysts expect sustained growth of 7-10%. Economic momentum thus seems to be accelerating rather than weakening.
This raises the question of why the American central bank is additionally stimulating the economy with further interest rate cuts. The Federal Reserve has lowered its key interest rate by 0.25% to 3.75-4.0%. This is the second rate reduction this year. However, Fed Chief Jerome Powell quickly dampened hopes. Another interest rate cut in December is „not certain,“ he said unambiguously. Powell‘s caution is due to the US government shutdown, which is blocking important economic data. The official employment reports are particularly missing. Without these, the central bank is driving in the fog, as Powell himself said. The Fed is observing a slight deterioration in the labor market, but experts expect significantly worse figures. Large corporations like UPS, Amazon, and Intel have recently announced tens of thousands of layoffs. Inflation is also a concern. Consumer prices have risen to 3.0%, moving away from the 2% target. This double uncertainty means that a December rate cut is not guaranteed. Even bond investor Jeffrey Gundlach estimates the odds at only around 50%. From December, the Fed will no longer reduce its balance sheet but will instead reinvest in bonds. This additional liquidity could support bond markets.
Between August and October 20, gold reached its record high of 4,381 USD per ounce. But then came a sharp correction. On Monday, the price was back below the 4,000 USD mark. Gold demand has never been higher in a quarter than between July and September. It stood at 1,313 tons and was 44% higher in dollar terms than in the previous year. The main driver was FOMO – „Fear of missing out.“ Investors bought gold out of fear of missing something. ETF inflows rose by 134% compared to the same period in the previous year. Up to the record high, the gold price increase was the fourth-strongest increase in gold price history. Experts consider a drastic correction like in the 1980s to be unlikely and expect the gold price to exceed the 5,000 USD mark. Chinese investors continue to form the most important buyer group.
OUTLOOK
THE RATE CUT PARADOX: WHY AMERICAN MONETARY POLICY COULD BACKFIRE
Autumn brings more than just falling leaves. Now, as the cold approaches and winter casts its shadows, it‘s time to sit down consciously and critically examine your own wealth allocation. This annual stocktaking is not a ritual of fear, but rather an opportunity for clarity. It‘s about readjusting your strategic asset allocation while confronting a fundamental question that sits at the core of every investment decision: What return do I actually need from a particular asset class to justify the associated risk?
This consideration leads us directly to the concept of risk premium—that difference between the expected return of a risky investment and a safe investment. Anyone investing in stocks should know what additional return they receive for the higher risk. Anyone engaged in bonds should understand whether the interest income compensates for default risk.
Autumn invites us to perform precisely this recalculation and thus approach the coming months with strategy instead of hope. Risk premiums help challenge your gut feeling with naked numbers. The risk premium is thus not merely a mathematical quantity. It is a protective mechanism against our own cognitive distortions — a tool that holds our gut feeling accountable.
The year 2026 is anything but a uniform narrative about risk premiums. While one asset class breathes a sigh of relief, another is suffocating. While opportunities emerge here, they disappear there. And it is precisely this divergence that forces the attentive investor to challenge his gut feeling through the numbers. Equity risk premiums are on the upswing. In the United States, Asia, and emerging markets, companies are experiencing a quiet profit revolution. Margins are rising, profits are flourishing, and this development is driving risk premiums upward. This is not speculative hope; this is the mathematical reality of the earnings growth model: better profits today justify higher risk premiums tomorrow. Yet the mechanics differ significantly by region. In Europe, Switzerland, and the United States, lower real interest rates act as an additional driver, pushing equity risk premiums above GDP per capita growth. China points to something exciting: profits are returning. This could be a harbinger of higher risk premiums. Japan demonstrates resilience—even rising interest rates couldn‘t brake the rally because robust profits played the counterprogram.
Government bonds, meanwhile, are experiencing a quiet resurrection. Coupon yields are finally attractive again and interest rate structures are normalizing. After years in negative territory, risk premiums are climbing back. The central question for every investor: Do rising corporate profits justify the risen stock prices? The numbers say yes. But what if the numbers tell only half the truth – read more about this in the focus section.
FOCUS
WE DID THE MATH: USD 10 BILLION IN LOSSES, USD 500 BILLION VALUATION, 20 NUCLEAR REACTORS – AI BREAKTHROUGH OR MADNESS?
Tech giants are investing more money than they earn, AI startups are burning billions, and a single island determines the future of the global economy. Valuations have reached historic proportions—77% of the annual U.S. economic output. Is this justified or are we witnessing the biggest bubble of all time? The AI Revolution: Are we already in a Bubble? The AI revolution is transforming our world as profoundly as the internet once did. On financial markets, technology companies are reaching historic valuations. This raises critical questions: Do these valuations justify future performance? Is market concentration sustainable? Or are we experiencing an AI bubble?
MARKET VALUE: WHEN NUMBERS MAKE YOU DIZZY
The scale is staggering: The 10 largest U.S. companies— mostly from the tech sector (Apple, Nvidia, Microsoft, Amazon, Alphabet, Meta, Broadcom, Tesla, Berkshire, and Oracle)—together account for about USD 26.9 trillion in market capitalization.
That equals 77% of the United States’ annual GDP. For comparison: During the tech bubble in 2000, the top 10 U.S. firms represented only 28% of U.S. GDP. Even more extreme is Nvidia: The company alone has a market cap of USD 4.4 trillion—about 15% of U.S. GDP. At the turn of the millennium, Cisco was the largest company, valued at USD 0.5 trillion, just 5% of GDP. Nvidia now carries a 7–8% index weight and generates 6–7% of all S&P 500 revenues. It is undeniably profitable— USD 160 billion in annual revenue and USD 100 billion in EBITDA—but this enormous market power rests on a fragile foundation.
EXTREME CONCENTRATION: THE TAIWAN DILEMMA
Nvidia disclosed (on an anonymous basis) that its two largest customers—presumably Foxconn and Quanta— account for 40% of its revenue. The top six customers make up 80%. Geographically, the concentration is extreme: Nvidia’s top three customers are all Taiwanese manufacturers (Foxconn, Quanta, and Wistron) that use Nvidia chips in their products. This highlights why the U.S. is determined to protect Taiwan from Chinese annexation at all costs.
The entire Nvidia story rests on a Taiwan dilemma:
• One island (Taiwan)
• Three main customers (all in Taiwan)
• One main supplier (TSMC)
• One prayer that geopolitical tensions stay lower than geological risks
Why do we focus so much on Nvidia? While the AI industry includes many players—most notably in “AI Compute” like OpenAI, Midjourney, and Anthropic—all of them combined generate less than USD 40 billion in annual revenue and are unprofitable.
Reality: The entire AI ecosystem depends on Nvidia’s profitability.
“BURNING MONEY”: WHEN INVESTMENTS EXPLODE
David Einhorn of Greenlight Capital has already warned: The current AI investment boom could become the greatest cash burn since the dot-com era. The figures are truly breathtaking.
Tech giants’ CapEx spending has reached historic levels:
• Microsoft and Google: about 50% of profits on CapEx
• Meta, Oracle, Amazon: about 70% (projected)
• By 2025: Big Tech could spend up to 130% of profits on CapEx
For comparison: AT&T invested 72% of profits before the telecom bubble, and Exxon spent 65% during the shale boom.
OpenAI is the most extreme example. In 2024, it signed USD 1 trillion worth of compute contracts, giving it access to over 20 gigawatts of computing capacity over the next decade—the equivalent output of 20 nuclear reactors. Each gigawatt of AI compute costs roughly USD 50 billion. The catch: OpenAI currently generates about USD 12 billion in revenue, loses USD 10 billion annually, yet is valued at USD 500 billion. Its contractual obligations far exceed its income.
FINANCING: “CREATIVE” MONEY FLOWS
OpenAI already has USD 4 billion in bank debt, plus USD 47 billion in venture capital funding in the past 12 months.
The creative financing doesn’t stop there:
• Nvidia plans to invest USD 100 billion in OpenAI over 10 years—so OpenAI can buy Nvidia chips.
• AMD grants OpenAI warrants to purchase up to 10% of AMD’s shares for just one cent each, while OpenAI buys computing power from AMD.
• OpenAI, together with SoftBank and Oracle, launched the “Stargate Initiative”—a USD 500 billion project for U.S. AI infrastructure.
• Industry insiders warn: “The company is in a far more capital-intensive business than Google or Microsoft ever were—and it was born without cost discipline.”
COMPUTE OVERSUPPLY: WHEN REALITY HITS
The situation among the hyperscalers (Oracle, Google, Meta, etc.) is no less concerning.
Problem 1: Limited Ad Revenue
Most revenue still comes from advertising. Ad spending in the U.S. has been stable at 1.5% of GDP (about USD 400 billion). Google alone plans to invest USD 85 billion in 2025. Exponentially growing CapEx cannot be financed by stagnant ad income.
Problem 2: Margin Pressure in Cloud Markets
New competitors like Oracle and specialized “neo-cloud” providers such as CoreWeave undercut prices by up to 40%. A historical parallel: AT&T once had “better margins than the drug trade”—until new entrants crashed the market.
The uncomfortable truth: Amazon, Google, and Microsoft are building data centers in advance of actual demand. Each hyperscaler will spend USD 80–120 billion on CapEx in 2025—roughly equal to Amazon Cloud’s total annual revenue.
VC investor Marc Andreessen notes that nearly 50% of U.S. economic growth now comes from tech CapEx.
REAL-WORLD PROBLEMS: WHEN THE POWER RUNS OUT
Energy demand is exploding: Data centers already consume 4.5% of total U.S. electricity, projected to reach 9% by 2030. The consequences are severe—according to Bloomberg, electricity prices within 5 km of data centers have risen by 267% on average. A single data center uses 300–400 MW up to 1 GW — the equivalent to a nuclear reactor. At this scale, today’s power grids would collapse. Building new generation capacity takes years and is highly political. Even basic components like power transformers are becoming scarce. Extreme electricity price volatility in affected regions will be inevitable.
CONCLUSION: THE BUBBLE IS PROBABLY REAL
After analyzing all the numbers, the fact-based conclusion is clear: We are probably in an AI bubble.
Evidence includes:
• Extreme market concentration beyond any historical precedent
• Only Nvidia is profitable—everyone else is losing money
• The entire AI ecosystem depends on one company and one island (Taiwan)
• Unprecedented “cash burn” from CapEx spending
• Massive overcapacity in data centers
• Circular financing without real demand—Nvidia invests USD 100 billion in OpenAI so OpenAI can buy Nvidia chips
• Energy supply and construction resources are finite
Investor takeaway: Past bubbles show they often last longer than expected. Exiting too early can cost more performance than the eventual crash destroys in value. The best approach: regularly take profits without fully exiting and implement systematic hedging strategies.
THE OTHER SIDE: WHY AI MIGHT NOT BE A BUBBLE
There are valid arguments against the bubble thesis. A crucial difference from the dot-com era: back then, working business models didn’t exist—today, they do. Nvidia already earns USD 100 billion EBITDA, and Microsoft and Google are seeing significant revenue growth from paid AI services. McKinsey estimates USD 13 trillion in additional global GDP by 2030 from measurable productivity gains.
What would need to happen for AI not to be a bubble?
• Diversification away from Taiwan dependence
• ROI increases in traditional industries through AI
• More energy-efficient AI technology
• Profitable business models for AI startups
The uncomfortable truth: The internet was also a “bubble”—but it still changed the world. Many dot-com companies failed, yet Amazon, Google, and eBay rose from the wreckage.
OUTLOOK: If quantum computers become commercially viable, they could fundamentally redefine technological reality—comparable to the leap from Newtonian mechanics to quantum physics. While AI determines how we work, quantum computing may decide what we can explore at all. For today’s AI leaders and their massive CapEx commitments, that could pose immense long-term pressure.