An AI bubble? What the giants’ third-quarter 2026 results reveal

Tech giants spend hundreds of billions of dollars a year on artificial intelligence infrastructure. The third quarter of 2026 provided a reason to pause: several of them have negative free cash flow, and their debt is growing. This doesn’t mean the bubble is “bursting,” nor does it mean everything is fine. This article summarizes verifiable figures and arguments from both sides. It is not an investment recommendation.

What’s Verified and What Isn’t

• “Leaks from financial reports”: these were not leaks. The data comes from publicly released quarterly results from late October 2026.
• “The bubble is bursting”: that is an interpretation, not a fact. The data show increasing pressure on cash flow, but revenue from cloud services is also growing.
• Sources: I cite figures based on company announcements and their earnings reports. Analyst estimates vary, and I provide a range for the amounts.

What the Numbers Say

• Alphabet (Google): In the third quarter, it had capital expenditures of $51.4 billion and free cash flow of -$9.1 billion, marking the second consecutive quarter of negative free cash flow. Over the past twelve months, it was still positive, at approximately $19.7 billion.
• Amazon: Free cash flow over the past twelve months was approximately $-42.4 billion, whereas a year earlier it was positive (around $14.8 billion). The company still expects approximately $220 billion in cash capital expenditures in 2026. Meanwhile, cloud revenue (AWS) grew by 39%. Despite this, the stock fell 7% following the earnings release.
• Meta: Free cash flow was approximately -$2.1 billion in the third quarter.
• Microsoft: is the exception. Free cash flow was positive, at approximately $18.6 billion, with capital expenditures of $45.2 billion. However, according to some sources, it relies more heavily on leasing, i.e., renting equipment from others.
• Debt: According to S&P Global Market Intelligence, the five largest companies (Microsoft, Alphabet, Amazon, Meta, and Oracle) borrowed a combined total of approximately $326 billion in 2026, roughly three times last year’s figure. On October 27, the Bank for International Settlements reported that these companies’ bonds with maturities of ten years or more accounted for roughly 24% of corporate issuances with that maturity.
• Total issuance: figures vary depending on which companies are included. The sum of four of them exceeds $740 billion; estimates for five companies range from approximately $800 billion to $900 billion. A year ago, the figure was around $400 billion.
• Future estimates: S&P Global Ratings expects these companies to have negative operating free cash flow in 2026 and 2027, and the agency does not foresee a turnaround until 2028–2029. Bank of America estimates a total negative free cash flow of approximately -64 billion dollars in 2026. These are projections, not actual data.

Argument for the “bubble”

• Expenses are growing faster than operating cash flow. For the first time, this is also affecting companies that have long been synonymous with cash cows.
• Financing is shifting toward debt, equity, and leasing. This increases sensitivity to changes in interest rates and investor sentiment.
• Model-developing labs do not have traditional profits from other businesses and depend on external capital until their AI ventures generate enough cash.
• The market reaction was sharp: according to one summary, the Nasdaq fell by 5.3% in October. This figure comes from a financial portal, so take it as a rough estimate.

An argument against the “bubble”

• Cloud revenue is growing at a double-digit rate. AWS’s 39% growth is proof that demand exists.
• A negative free cash flow is not, in and of itself, proof of a bad investment. Companies incur costs upfront before deploying infrastructure, and revenues come later. Alphabet has maintained a positive cash flow over the past twelve months, and Microsoft has also posted a positive cash flow for the quarter.
• According to some sources, the companies also have a significant volume of contracted orders to cover their expenses. However, the quality of these orders cannot be assessed based on the available data.

How does this compare to the dot-com bubble of 2000?

Some analyses compare current valuations to those in 1999. According to one source, the CAPE ratio stood at 41.6 on October 8, 2026, while in December 1999 it reached 44.2. Such comparisons are crude. The difference is that today’s market leaders have actual revenue and profits, but the similarity lies in the rapid growth of spending before a proven return on investment.

What does this mean for regular users?

• Pressure to generate a return on investment could lead to higher prices for AI services or restrictions on free versions. However, this has not been proven.
• If funding becomes tighter, the construction of data centers and the release of new models may slow down.
• Companies may place greater emphasis on efficiency: smaller and cheaper models.

What to Take Away From This

The data supports a more cautious tone than “unbridled enthusiasm,” but it does not support a definitive “the bubble is bursting.” The question is whether AI revenues will catch up with expenses before the patience of creditors and investors runs out. No one can answer this question with certainty today. I am not a financial advisor, and this text is not a recommendation to buy or sell any securities.