Report published August 13, 2026
AI Bubble? Assessing Capex Fragilities, Valuation Peaks, and Market Contagion
Source and citation context
- Issuer
- 360 ONE Asset
- Report date
- August 13, 2026
- Analysis as of
- August 13, 2026
Finvaulta summarizes 360 ONE Asset's analysis. Attribute opinions, forecasts, time-sensitive values, and chronology to the issuer and report date; do not treat this page as an independent verification or a current market-data source.
360 ONE evaluates whether the AI buildout has entered bubble territory, citing circular financing, rising debt issuance by hyperscalers, and competition from low-cost Chinese open-weight models. It cautions that an unwind in semiconductor and hardware stocks could trigger contagion in highly concentrated Asian markets, though India remains relatively insulated.
Key Takeaways
- 1.AI capex by hyperscalers is increasingly debt-funded and relies on circular financing between AI labs and cloud providers, exposing credit markets to potential buildout vulnerabilities.
- 2.Cost pressures are intensifying on Western frontier labs as Chinese open-weight models lag performance by only ~6 months while operating at a fraction of the cost, driving a ~45% decline in blended token prices since May 2026.
- 3.Semiconductor equity rallies mirror the late-1990s dot-com boom, where price-to-book ratios sit at or above historical peaks despite stable P/E multiples, leaving little room for earnings disappointment.
Table of Contents
- Key Insights
- AI Bubble?
- AI adoption is increasing, but it has yet to become broad-based
- AI is affecting workers through slower hiring, not outright layoffs
- Hyperscaler capex is increasingly being financed through debt
- Corporates cite AI-driven productivity gains and labour substitution
- Chinese models trail US models in performance by just ~6 months
- Sharp rally in semiconductor stocks bears a striking similarity to the 1990s
- P/E multiples haven't re-rated much, thanks to strong earnings
- New tech follows a boom-bust cycle driven by inflated expectations
- Taiwan and South Korea are also driven by the AI-led investment cycle
- Indian markets are significantly less concentrated and more insulated
- What could derail the AI cycle?
- Disclaimer
Report data
Capex by US Hyperscalers and Other AI Firms & Change in Long-term Debt — as of August 13, 2026.
| Metric | Estimate | Context |
|---|---|---|
| Decline in Blended Price per Million Tokens | -45.0% | Blended price paid per million tokens fallen as enterprise usage rotates to cheaper open-weight models |
| Chinese Models Performance Lag vs US Frontier Models | 6.0 months | Lag in capability benchmarks according to Epoch Capability Index |
| Market Concentration of Top 1-2 Companies in Taiwan | 39.0% | Share of top 1-2 companies in total equity market capitalization |
| Market Concentration of Top 1-2 Companies in South Korea | 47.0% | Share of top 1-2 companies in total equity market capitalization |
| Market Concentration of Top 10 Companies in the US | 33.0% | Share of top 10 companies in total US market capitalization |
Document Preview
Access the Full Report
Get unlimited access to institutional research reports. Create an account to get started.
Reported Data Context
- Decline in Blended Price per Million Tokens: -45.0 % (May 2026 to August 2026) · Source: Bloomberg, Silicon Data LLM Token Expenditure Index, 360 ONE Asset Research
- Chinese Models Performance Lag vs US Frontier Models: 6.0 months (August 2026) · Source: epoch.ai, Artificial Analysis, 360 ONE Asset Research
- Market Concentration of Top 1-2 Companies in Taiwan: 39.0 % (As of August 13, 2026) · Source: Bloomberg, 360 ONE Asset Research
- Market Concentration of Top 1-2 Companies in South Korea: 47.0 % (As of August 13, 2026) · Source: Bloomberg, 360 ONE Asset Research
- Market Concentration of Top 10 Companies in the US: 33.0 % (As of August 13, 2026) · Source: Bloomberg, 360 ONE Asset Research
