> ## Content Index
> Fetch the complete content index at: https://tokyobrainie.ghost.io/llms.txt
> Use this file to discover other available public pages before exploring further.

# AI Bubbles Break on Cash Flow, Not Technology
- URL: https://tokyobrainie.ghost.io/ai-bubbles-cash-flow-en/
- Published: 2026-10-09T23:30:00.000Z
- Updated: 2026-11-11T00:05:09.000Z
- Description: A real technology does not guarantee that every company financing it will survive. What Barron’s 2000 cash-runway investigation can teach us about the AI capital cycle.
- Author: TokyoBrainie
- Tags: Signal Room, English

Languages: [日本語](https://tokyobrainie.ghost.io/ai-bubbles-cash-flow-ja/) · English · [简体中文](https://tokyobrainie.ghost.io/ai-bubbles-cash-flow-zh/)

First published: 10 October 2026\. Historical research note.

## What happened

Published on March 20, 2000, this Barron’s article examined the cash balances and burn rates of 207 listed internet companies. At least 51 appeared capable of exhausting their cash within twelve months. The core measure was simple: **cash runway equals available cash divided by average monthly cash burn.**

The work was conducted by Barron’s and Pegasus Research International. Its estimates put Amazon’s runway at about ten months, eToys’ at eleven months and CDNow’s at one month.

## Why it mattered

The March 20, 2000 publication came about ten days after the Nasdaq peak. On the same day, MicroStrategy revised previously reported revenue, adding to doubts about internet stocks.

The article shifted attention from dreams to cash, page views to operating cash flow, total addressable markets to survival until profitability, and rising share prices to the ability to raise the next round.

The internet was not a false technology. Many companies simply lacked enough money to reach the future they promised.

## What it got right—and its limit

Loss-making companies depended on rising prices and repeated issuance. When financing closed, many were sold, merged or liquidated. But runway is not a predicted date of death: new equity, credit, strategic investors, cost cuts and better revenue can change the result. Amazon survived. Runway is therefore a stress indicator for dependence on capital markets.

## Dot-com versus AI

The dot-com companies often burned cash on advertising and payroll. Today’s AI cycle requires GPUs, data centers and power, financed through bonds, leases, SPVs and project finance. The vulnerable structures are debt-funded AI clouds, concentrated customers, projects ordered before utilization is secured and companies whose free cash flow remains negative after capital expenditure.

### Comparing the two investment cycles

- **Dot-com spending:** losses driven by advertising and payroll. **AI spending:** enormous capital expenditure on GPUs, data centres and power.
- **Dot-com funding:** dependence on new equity. **AI funding:** bonds, leases, special-purpose vehicles (SPVs) and project finance.
- **Dot-com revenue:** often very limited. **AI revenue:** sales can be substantial while returns on investment remain uncertain.
- **Dot-com metrics:** page views and memberships. **AI metrics:** GPU utilization, inference demand and token prices.
- **Dot-com stress:** cash exhaustion. **AI stress:** falling utilization, depreciation, refinancing and customer concentration.

Large technology companies can finance AI from their existing businesses, so they should not be equated with weak dot-com start-ups.

### Structures requiring closer scrutiny

- AI clouds buying large GPU fleets with borrowed money
- Data centres dependent on a handful of customers
- Companies ordering equipment before securing long-term usage contracts
- Collateral lending that overestimates GPUs’ resale value
- Companies growing revenue while free cash flow after capital expenditure stays negative
- Start-ups relying on their next equity issue or refinancing

## Japan market

Japanese equipment, testing and optical-connectivity companies are profitable suppliers, not dot-com replicas. Yet tighter financing can delay data centers, weaken GPU, HBM, optical and power orders, lower capex forecasts and compress valuation multiples before earnings decline.

The transmission sequence is: worse funding conditions for AI clouds → delayed data-centre projects → slower orders for GPUs, HBM, optical connectivity and power equipment → lower semiconductor capital-expenditure forecasts → valuation-multiple compression before reported earnings decline.

## What to watch｜A modern Burning Up checklist

- Cash and short-term investments
- Operating cash flow
- Free cash flow after capital expenditure
- Annual cash burn
- Debt maturities over the next three years
- Lease liabilities and purchase commitments
- Customer concentration
- GPU and data-centre utilization
- Gross margins
- Time the business can continue without another equity issue
- Whether stopping capital expenditure would also stop growth
- Economic profitability after allowing for stock-based compensation

## Source

[Barron’s｜Burning Up](https://www.barrons.com/articles/SB953335580704470544?ref=tokyobrainie.ghost.io)

## Verdict

**A bubble breaks not when the technological dream is disproved, but when the money sustaining it stops. Before asking whether AI is real, ask whether the company can survive until profitability.**