Global Household Wealth Jumped in 2025, Raising Stakes for Profitability of AI
Household wealth grew to $570 trillion in 2025, an increase of 7.3% from 2024
By Paul Hannon, The Wall Street Journal, July 23, 2026
- Global household wealth grew 7.3% to $570 trillion in 2025, driven largely by rising equity prices, according to McKinsey Global Institute.
- Expectations of bumper profits from artificial intelligence pushed equity prices to new highs, raising the risk of a sharp correction.
- Paper gains from rising asset prices drove 60% of the wealth increase, compared with an average of one-third between 2000 and 2024.
…
It is, of course, good news when people get richer. But there are signs that the particular way in which they are getting richer now may not be sustainable. MGI estimates that only 20% of the rise in household wealth was driven by new investment in real assets such as machinery and equipment, homes and buildings. From 2000 to 2024, the average was 30%.
By contrast, it reckons that 60% of the rise in wealth was driven by “paper gains”—or rising asset prices—compared with the average of one-third between 2000 and 2024…
“When real estate and equity values rise faster than GDP, capital may disproportionately go to asset repurchases, sometimes with a lot of leverage,” MGI said. “For households, wealth rises, but merely on paper, with heightened risks of eventual corrections.”…
Foreign investors are heavily exposed to America’s AI boom. … More than a third of U.S. equities are owned by foreign investors.
Should asset prices fall sharply, households around the world would feel less wealthy, and likely cut back on spending, weakening economic growth. [end quote]
Leverage - buying stocks on margin - magnifies gains on the way up and losses on the way down. My grandparents warned me against leverage. On a trip to Paris in late 1929, they met American expatriates who had lost everything in the stock market crash and couldn’t even afford a ticket to return home.
The rapid growth of the S&P500 index has been driven by AI and AI-related stocks.
In 2023–2024, the “Magnificent Seven” accounted for up to 60%+ of the S&P 500’s total return. AI-related stocks and infrastructure beneficiaries have driven roughly 40% to 50% of the S&P 500’s growth from early 2025 through mid-2026.
At the same time, Debit Balances in Customers’ Securities Margin Accounts rose to a breathtaking record of over $1.5 Trillion or about 5% of GDP. This is the leverage that is inflating the AI bubble.
Hyperscale AI capex spending surged past $600–$700 billion annualized during this period, fueling revenue directly for AI hardware, networking, and data center supply chains. Although several of the hyperscalers are paying most of the cost out of free cash flow, they have started to borrow heavily. High-grade tech bond issuances surged past $200 billion to $400+ billion in 2025–2026, including mega-deals by Meta, Alphabet, Amazon, and Oracle.
Alphabet’s latest quarterly report showed negative free cash flow due to high capital spending.
That spending provided income which circulated within the “AI ecosystem” but did not come from end-user customers. Direct end-user spending on generative AI models and software is generating ~$40B–$64B annually in revenue for model makers and platforms. That is expected to grow 50%–60%+ year-over-year if the business model continues as predicted.
Even if the business model growth continues as predicted (monopoly/ oligopoly hyperscale AI with high token prices) it will take many years for the end-users to pay for the data centers.
The problem is the the business model is NOT going according to plan. Token prices are already falling exponentially.
Most threatening of all, China completely changed the paradigm only a week ago by providing open-source AI that enables cheap “distilled” AI to replace expensive “frontier AI” for many business, medical and defense applications. I wrote about this in detail a few days ago.
If there’s one thing we have all seen with technology, it’s that a new, cheaper, better tech paradigm can wipe an old tech paradigm off the map faster than you can say “Wang Computers” or “Betamax.”
The “distilled AI” paradigm could make the huge, stunningly expensive frontier AI data centers obsolete before the permits are issued and the ground is broken. They just won’t be needed.
The biggest existential threat to hyperscaler projections isn’t just open-source model distillation—it’s the physics of energy interconnects. Substation transformers, high-voltage power lines, nuclear regulatory approvals, and natural gas turbine lead-times take 3 to 7 years to deploy. If hyperscalers cannot secure hundreds of megawatts of continuous baseload power, hundreds of billions in hardware capex will sit underutilized. When physical energy constraints collide with high debt service, infrastructure returns degrade rapidly.
Chip manufacturers have gone through this many times. Spend billions on a new fab. Too many chips. Prices plunge. Competition leapfrogs technology. See INTC.
Think of what would happen to Nvidia if data centers were canceled.
Beyond traditional corporate bonds, an increasing share of data center real estate, power infrastructure, and energy grid interconnects is being funded through specialized private credit, JV debt stacks, and securitized credit (Data Center ABS) rather than direct balance-sheet cash. If this goes south it will drag a lot of financiers along with it, just as the Global Crossing bankruptcy did in 2000 with the same too big- too soon dynamic.
Not to mention the stock market.
As the news of the threat to high-cost frontier AI spreads, investors will start to back away. The time frame is uncertain.
~15% to 18% of the entire U.S. stock market value is held directly inside passive S&P 500 index funds and ETFs (e.g., SPY, VOO, IVV, and mutual funds). If leveraged speculators are forced to sell during a downturn due to margin calls, ALL the stocks in the SPX will be sold together. The babies will be thrown out with the bathwater.
DJIA and SPX have plateaued but the NAZ is feeling the pain. NASDAQ New Highs-New Lows are plunging as is Bullish Percent.
VIX is low but the volatility is rising in individual stocks as shown in VIXEQ. The money isn’t leaving the stock market but is rotating to different stocks.
Stock, Treasury prices and junk bond prices are all dropping together. The Fear & Greed Index is in Fear.
As bond prices fall their yields rise. The Treasury yield curve is rising along its entire length, especially at the long duration end. Even if the Federal Reserve keeps the fed funds rate constant the longer durations are controlled by the market, not by the Fed.
There has been a dramatic trend change in the Treasury yields. The consistent drop in yields between 1981 and 2020 has ended. The trend has turned toward higher yields. This is true of both the nominal and real (inflation-adjusted) yields. The 10 year TIPS yield is also rising, showing that the rising yields are due to factors other than inflation. The most obvious reasons are the exploding government deficits, withdrawal of price-insensitive buyers (such as the Fed and foreign trade partners) and gigantic borrowing by AI hyperscalers on a governmental scale. The 30 year mortgage rate is rising along with the 10YT and is over 6.5%.
The Chicago Fed’s National Financial Conditions Index (NFCI), which provides a comprehensive weekly update on U.S. financial conditions in money markets, debt and equity markets, and the traditional and “shadow” banking systems, shows that financial conditions are very loose and getting even looser. However, the Leverage Subindex remains positive, confirming that debt-to-equity and balance-sheet leverage metrics are still running tighter than historical averages. The Leverage Subindex is a leading indicator of financial stress that shows up before stock prices are affected. The overall Index shows loose conditions because the high level of the stock index overwhelms the Leverage Subindex in the Chicago Fed’s model.
The High Yield Index Option-Adjusted Spread - that is, the interest rate on junk bonds relative to Treasuries - is still low. That’s a relief because many zombie companies whose cash flow barely pays interest on their debts have ultra-low-interest 2020 bonds rolling over this year. But the rise in the underlying Treasury yield has pushed the yields of junk bonds over 14%. If the spreads rise many companies will go bankrupt although they will try to extend by several methods. About 2.2 million people work for zombie companies, most of which are small local companies.
The Cleveland Fed predicts much lower inflation in 3Q26. But that doesn’t include today’s war and tariff news.
The Atlanta Fed’s Latest GDPNow Estimate for 2026:Q2 is 1.7%.
With the Iranians and Houthis blocking the Strait of Hormuz and threatening the Red Sea southern exit the price of oil is rising again. Ships are still sailing but the situation is uncertain. The price of gasoline is rising along with oil but so far no extra penalty for refined products.
https://www.nytimes.com/2026/07/25/business/energy-environment/houthi-blockade-red-sea-maps.html
The Administration late Thursday unveiled new tariffs under Section 301 to replace the President’s Section 122 tariffs that were lapsing. These latest duties will affect some 60 economies, hitting nearly all U.S. imports, with rates from 10% to 12.5%, and they come on top of a slew of other border taxes.
https://www.wsj.com/opinion/donald-trump-tariffs-301-canada-f3e90f13?mod=hp_opin_pos_2
The METAR for next week is cloudy. I think the situation is a metastable state - it’s inherently unbalanced and will topple suddenly if an outside force overcomes the small activation energy. But it’s possible that the huge cash buffers from the hyperscalers’ existing businesses will carry through until AI mass adoption gradually pays for the investments.
Wendy