We spend a lot of time talking about AI on METAR because the AI-related stock market bubble has driven the growth in the S&P500 index.
Let’s remember that AI is a tool. The purpose of the tool is to improve the productivity of the real economy of goods and services – everything from pedicures to rocket ships, food, clothing, shelter, medical care, entertainment… everything.
We need to know GDP to understand whether the real economy is growing. Also, we need to know the number to understand the scale of expenditures relative to the entire economy. GDP is all goods and services produced in the U.S. (which includes exports) minus imports.
2Q26 GDP (nominal) was $32.5 Trillion. Current-dollar GDP increased at an annual rate of 7.9%.
Real GDP is adjusted for inflation. 2Q26 Real GDP = $24.3 Trillion Chained 2017 Dollars, Seasonally Adjusted Annual Rate. This is a big difference because inflation has eaten away the value quite a lot over the past 10 years.
- Real Growth Rate: 1.5% (seasonally adjusted annual rate).
- Quarterly Context: This reflects a slowdown from the 2.1% real growth recorded in 1Q26. The primary drivers of growth were consumer spending, non-residential investment, and exports, offset by lower government spending and higher imports.
Because AI spending has been in the hundreds of billions of dollars, I want to know how this affects GDP and also how well the entire economy has been growing without AI investments. Relatively few people work in the AI development. The vast majority of people work in non-AI jobs.
The BEA does not publish an official “ex-AI” GDP metric because AI spending is spread across multiple categories—such as information processing equipment, software IP, data center construction, and imported semiconductor hardware.
- A large portion of the hardware powering the AI buildout (such as advanced semiconductors and telecommunications equipment) is imported. The resulting spike in capital goods imports caused the trade deficit to subtract over 1.0 percentage point from overall 2Q26 real GDP growth.
Because the surge in imported AI hardware largely offset the gains in domestic AI-related capital expenditure, the net direct effect of AI factors on headline real GDP was roughly neutral to slightly negative for the quarter.
I asked Gemini to summarize the GDP story.
| Metric | 1Q26 | 2Q26 | 1H26 Annualized Average |
|---|---|---|---|
| Headline Real GDP | 2.1% | 1.5% | 1.8% |
| Ex-AI Real GDP | 0.3% | 0.0% | 0.15% |
Real GDP growth (ex-AI) came to a standstill in 2Q26. There’s some detail about exports from the government’s Strategic Petroleum Reserve (SPR) impacting GDP but the overall take-away is that GDP growth has stalled.
Outside of tech capex, consumer spending grew at a modest 1.8% rate, while residential construction and government spending acted as headwinds, leaving the non-AI portion of the economy essentially stagnant.
The Atlanta Fed’s Initial Third-Quarter GDPNow Estimate for 2026:Q3 Updated: July 30, 2026 is 5.0%. This is dramatically higher than the Blue Chip consensus which has been running under 2%. This includes the entire economy including AI.
With the economy slowing, let’s look at U-6 unemployment. This is the broadest measure of unemployment since it includes everyone who is actively looking for a job but can’t find a full-time job. ( Total Unemployed, Plus All Persons Marginally Attached to the Labor Force, Plus Total Employed Part Time for Economic Reasons, as a Percent of the Civilian Labor Force Plus All Persons Marginally Attached to the Labor Force.) This has climbed gradually since 2022 but is stable at 8%.
The unemployment rate only includes people who are actively searching for a job. Let’s look at the Labor Force Participation Rate which includes everyone. This plunged during Covid but a lot of people came back to work in 2022 - 2025. Many people left the labor force in 1H26, including prime-age workers. This may reflect the departure of immigrants from the work force who have not been replaced.
The big news on the financial front this week was the decision of the FOMC to hold the fed funds rate steady. Why should this be big news? Because the options market was expecting the Fed to raise the rate by 0.25%.
Inflation is still above the Fed’s target of 2% though it is declining. The decline may be temporary depending on whether oil begins to flow freely from the Middle East again, which depends on the on-again-off-again-on-again Iran war.
Fed Chair Kevin Warsh said that he wants the market to do price discovery without forward guidance from the Fed. Well, he got price discovery, all right. Bond traders didn’t like the FOMC inaction. They sold off the long-dated Treasury bonds. The Treasury bond yield curve had a “bear steepening” where the short end yield fell a little but the long end rose a lot. Long-dated bonds (especially the 10 year Treasury) have a much stronger impact on the economy than the overnight fed funds rate since mortgages and business loans are based on longer term yields.
Lenders are demanding a higher yield for the longer term (“term premium”) in addition to inflation because of very high forecast government deficits which competes with commercial and household borrowing. This term premium shows in the rising yield of the TIPS which is inflation adjusted.
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, showed that financial conditions are very loose and getting looser.
The stock market bubble is being inflated by an immense amount of borrowed money. Debit Balances in Customers’ Securities Margin Accounts equaled $1.5 Trillion in June 2026, about 5% of nominal GDP. This is pure speculation and resembles other speculative bubbles, like 1929 and 1999.
The stock indexes popped up a bit last week. Volatility declined. The Fear & Greed Index was in Fear but the trade turned toward risk-on.
USD bounced off the top of its channel. Gold and silver seem to have stabilized since their bubbles popped. Gemini analyzed demand and concluded that the floor for gold may be $3800 per ounce. Past gold bubbles have popped and stabilized at levels which were higher than pre-bubble. Of course, Gemini is AI and not a crystal ball so that could be wrong.
Oil, gasoline and diesel prices are rising. Natgas fell. Copper is rising even as gold is falling.
The big Macroeconomic risks for investors are a stock market crash and a financial crisis where companies become illiquid or (worse) insolvent. These are two different things. The 2000 dot-com stock crash was not a financial crisis. (But lenders lost huge on laying fiber-optic cables that were way ahead of profitable internet traffic.) The 1929 and 2008 stock market crashes were each tied in to a financial crisis.
The AI capex surge carries unprecedented concentration risk—driven almost entirely by four hyperscalers rather than broad-based business investment. Meanwhile, macro labor productivity numbers have yet to register any measurable uptick, leaving the economy with all of the debt and hardware expenses today, but none of the efficiency gains yet. It’s like the early internet or memory chip manufacturers. Excess capacity forces prices down and it could take years to catch up. Meanwhile, “frontier AI” hyperscalers are being undercut by much cheaper “open AI distilled models.”
The high level of debt financing the AI build-out is added to a tsunami of low-rate debt issued to non-AI companies during 2020-2022 which will be maturing in 2026-2028. Many of these companies are zombies that barely make their interest payments even on the low rate debt. Junk bond yields are rising (though not yet to crisis levels). Many companies are negotiating PIK (“payment in kind”) deals where they can’t pay maturing bonds but roll them into even higher debt at higher interest rates.
Zombie corporations aren’t the only ones facing a maturity wall; regional bank balance sheets remain constrained, tightening credit conditions for mid-sized non-AI businesses. Mid-sized businesses have to borrow locally because they can’t borrow on Wall Street. They hire a lot of people. Higher interest rates and tight credit conditions may force layoffs or even bankrupt local companies that need local lenders to meet payrolls.
The real economy already has zero growth.
A classic economic slowdown acts as a direct pin to the AI bubble. The core vulnerability of the current tech expansion is a massive monetization disconnect: hyperscalers and tech giants are committing $660B to $725B in annual capital expenditures for data centers and chips, while total global end-user spending on AI models and platforms sits at roughly $64 billion.
To justify current tech valuations and multi-hundred-billion-dollar infrastructure debt, end-user software monetization needs to explode exponentially. A classic recession starves that required revenue pool.
Zombie companies with rising debt loads will cut AI before they cut head count.
Real disposable income shrank and personal savings rate declined. When real disposable income contracts while the personal savings rate simultaneously falls (dropping toward historical lows around 2.7%), it signals a fundamental macroeconomic squeeze: households are drawing down cash buffers to maintain their baseline living standards against sticky inflation or rising debt-servicing costs.
Personal Consumption Expenditures (PCE) represent roughly 70% of U.S. GDP. When spending is financed by eroding savings rather than real income expansion, spending eventually hits a hard wall once liquidity reserves deplete.
This is a classic recessionary pathway. It will take a while to develop.
The METAR for next week is cloudy. The charts aren’t sunny but the short-term forecast is stable.
Wendy