The AI Bubble Needs New Victims

Summary

The video argues that AI companies’ heavy infrastructure spending, uncertain profitability, and unmet promises could lead them to seek more capital from ordinary investors—particularly through IPOs—while leaving investors exposed if enthusiasm for AI fades.

Financial claims and key figures

OpenAI

• The video says OpenAI has committed roughly $1.4 trillion to expansion and infrastructure deals.
• It projects a $14 billion loss in 2026, no profitability until 2030, and cumulative cash burn of $143 billion.
• Operating costs are described as about $700,000 per day, with the company potentially running out of money by mid-2027.
• The video says OpenAI filed for an IPO on June 8, 2026.

Alphabet

• In the first half of 2026, Alphabet reportedly generated about $39 billion in cash while spending roughly $45 billion on AI infrastructure.
• The video says its free cash flow turned negative for the first time since 2004.
• It cites roughly $811 billion in future purchases and contractual obligations.

Broader technology sector

• The five largest technology companies are said to carry $1.65 trillion in debt associated with AI.
• Anthropic is described as deriving 80% of its revenue from businesses and, like OpenAI, needing continual capital.
• Anthropic has also reportedly hinted at a possible IPO.

Spending, efficiency, and job impacts

The video questions AI’s promises of cheaper labor and greater efficiency:

• Microsoft reportedly restricted access to some AI coding tools after usage costs rose.
• Meta’s AI talent reportedly cost the company more than $1 million per employee.
• Uber reportedly used its entire annual AI budget in four months.
• Klarna laid off about 700 customer-support workers and replaced them with an OpenAI-powered chatbot, but later said it wanted humans back in the loop.
• Salesforce reportedly cut 4,000 support jobs and later rehired workers.
• The National Eating Disorders Association shut down an AI chatbot after it gave harmful advice.

The video also cites broader labor-market figures: software salaries fell 9%–15%, more than 50 CEOs announced company-wide layoffs citing AI in 2025, and entry-level jobs were down 67%. These are presented as evidence of AI-related disruption, not as independently verified statistics.

Market risk example: South Korea

• South Korea’s KOSPI reportedly rose nearly 200% in one year, supported by investment in AI data centers and demand for memory chips and specialized semiconductors.
• The video says that when U.S. investment pulled back, the market fell about 25%, wiping out roughly 3 trillion won in investments.
• It claims around 320,000 investment accounts were affected and says the fallout reached roughly one in every 30 people in the country.
• The video presents South Korea as a possible warning for the United States and other markets if they become similarly dependent on AI-related capital flows.

Investment risks and takeaways

• AI infrastructure commitments and operating costs may be difficult to sustain if revenue and productivity gains do not catch up.
• IPOs could shift some funding risk from wealthy investors and private capital to ordinary investors.
• Markets concentrated in AI-related companies, data centers, memory chips, or semiconductors may be vulnerable to a reversal in investment flows.
• The video’s overall caution is that investors could face substantial losses if AI expectations prove excessive. It does not provide a specific portfolio strategy or an explicit “not financial advice” disclaimer.

Assets, sectors, and sources mentioned

Assets and sectors: OpenAI and Anthropic (private AI companies); Alphabet/Google, Microsoft, Meta, Uber, Klarna, and Salesforce; the KOSPI; AI data centers; memory chips; semiconductors; and technology-sector debt. No ticker symbols are stated.

Presenters and sources: The video is presented by the Underground channel; no presenter is named. It refers to former OpenAI employee Sucher Balaji and his interview with The New York Times. Other quoted speakers are not identified.