A rising stock market can make an investment boom look stronger than the system beneath it. Nvidia and SpaceX put a sharper question in front of us: can the customers using this infrastructure generate enough cash to support the money committed to building it?
Assessment: the evidence supports concern about concentration and overbuilding. It does not establish that a dot-com-style crash is imminent. The useful work is to examine where growth depends on expectations, financing and payments arriving on time.
Stock market concentration: the index can outrun the typical company
Reuters reported on October 12 that the S&P 500 had gained about 14% in 2026, while its median constituent had gained roughly 4% and more than 40% of current constituents had fallen. Its analysis excluded companies already removed from the index that year. [1]
That is a warning about market breadth: how widely gains are shared. A few large companies can lift a capitalization-weighted index while many others struggle. Their success matters, but an index record cannot establish that every industry, household or business is advancing with them.
The dot-com lesson: real innovation can still attract too much capital
San Francisco Fed research describes how the late-1990s technology boom combined genuine innovation with investment in capacity that exceeded sustainable demand. The internet's usefulness did not ensure that every network, company or share price would earn its expected return. [2]
There is also evidence against declaring a broad valuation bubble. FactSet's October 2 report put the S&P 500 at 19 times expected next-year earnings, close to its ten-year average of 19.1. That dated snapshot is not today's quote. More importantly, expected earnings are forecasts: an apparently reasonable multiple becomes less reassuring if the profits fail to arrive. [3]
Nvidia: strong earnings, longer payment terms and financing exposure
Nvidia reported $96.2 billion in revenue and $63.7 billion in operating income for the quarter ended July 26. Those results demonstrate substantial operating performance; they do not settle what an investor should pay for future growth. [4]
Its CFO commentary also reported 60 days sales outstanding, up from 45 days in the previous quarter. Nvidia attributed the increase to extended payment terms on agreements with certain investment-grade customers. A longer collection period needs context; it does not by itself establish bad debts. [5]
The same document disclosed guarantees capped at $105 billion supporting SB Energy's Ohio data-center project under leases to OpenAI. Obligations become effective in phases as conditions are met, with the first expected in Nvidia's fiscal 2029. This is conditional exposure, not $105 billion already lost or immediately payable. [5]
Interpretation: a supplier helping customers secure infrastructure becomes more exposed to their execution and payment capacity. Follow both the sale and the obligations supporting it.
SpaceX: the capital expenditure test
SpaceX's second-quarter filing reported $7.8 billion in revenue, a $541 million net loss and $18.4 billion in capital expenditure, including $15.8 billion for AI. It also reported approximately $100 billion in cash, cash equivalents and marketable securities. [6]
That liquidity is a substantial cushion. Spending above current revenue can finance productive expansion, but future returns still have to justify the commitment. Compare the timing of investment, customer receipts and financing obligations. Capital expenditure is not the same measure as operating loss, and cash on hand is not proof of an adequate investment return.
How an AI investment slowdown could spread
Risk scenario, not a forecast. If customers earn less from AI services than expected, they may delay expansion. Equipment orders could weaken, infrastructure developers could face unused capacity, and lenders could demand tougher terms. Financing pressure could then force further spending cuts.
The effects could reach construction contractors, power suppliers and local service businesses. That is the economic connection worth investigating. It is not established merely by a falling technology share price.
The broader evidence also matters. BEA's September 30 estimate showed second-quarter real GDP growing at a 2.2% annualized rate, with contributions from consumption, investment and exports. This does not measure growth without AI, but it cautions against assuming the rest of the economy has stopped functioning. [7]
Four AI bubble warning signs worth tracking
- Cash conversion: customer revenue grows while collections fall behind or payment terms keep lengthening.
- Capacity utilization: cancellations, idle equipment or falling service prices undermine expected returns.
- Financing resilience: refinancing becomes harder as guarantees, leases and debt payments approach.
- Market breadth and earnings: fewer companies support the rally while profit forecasts weaken.
These are proposed monitoring questions, not findings that all four conditions are present. Improving customer cash generation, productive use of new capacity and broader earnings growth would weaken the bubble argument.
The Infrastructure Literacy test: choose one growth promise. Identify the paying customer, the infrastructure required, the party funding it and the evidence that the customer can sustain the payments. An investment boom becomes durable when the cash generated by useful work can carry its commitments.
Sources and evidence cutoff
- [1] Reuters / Noel Randewich: market breadth analysis, republished by AOL · 2026-10-12.
- [2] Federal Reserve Bank of San Francisco / Kevin J. Lansing: Speculative Bubbles and Overreaction to Technological Innovation · 2008-06-20.
- [3] FactSet / John Butters: S&P 500 Earnings Season Preview, Q3 2026 · 2026-10-02.
- [4] Nvidia: second-quarter fiscal 2027 earnings release · 2026-08-26.
- [5] Nvidia: CFO commentary, second-quarter fiscal 2027; commitments and guarantees on pages 3-4 · 2026-08-26.
- [6] SpaceX: second-quarter 2026 earnings release, filed with the SEC · 2026-08-04.
- [7] Bureau of Economic Analysis: second-quarter 2026 GDP, third estimate · 2026-09-30.
Evidence cutoff: October 13, 2026. Market figures retain their source dates; company results cover different fiscal periods. This is a selected-source comparison, not a complete valuation, credit review or prediction of a market peak. The historical Fed research is not a current Fed assessment or endorsement.
Explore the operating side in SpaceX and Starlink Mobile: Who Controls the Network?, then try Sales Are Up. Why Is Cash Tight? to trace the gap between growth and collected cash.