Behind the video
A.I. Is Getting Expensive, Even For Google
Overall Thesis
Alphabet is dramatically ramping AI capital expenditures and, for the first time in decades, issuing new stock rather than relying solely on internal cash flow and debt. The host argues this spending is being driven largely by investor sentiment and competitive pressure rather than proven AI profitability, and highlights signs (like Uber's AI spending blowouts and 'token maxing') that enterprise AI ROI may be overstated.
Narratives
GOOGLAlphabetThe host details Alphabet's $85 billion stock offering and surging AI capex, noting the company chose equity over additional debt to fund an increasingly expensive and uncertain AI buildout. He frames this as a cautious, low-risk strategy for Alphabet given its inflated market cap, while flagging real risk that heavy AI spending (and reliance on customers like Anthropic) may not pay off.
Key Arguments
- Alphabet's capex jumped from $52B (2024) to a planned $185B (2026), far outpacing its ~$165B operating cash flow.
- The company raised $85B in equity (diluting shares by ~2%) rather than taking on more debt, despite having room to borrow more at low rates.
- Much of Google's AI-related revenue growth comes from selling cloud compute to external AI companies like Anthropic rather than from Google's own Gemini products.
- At current AI operating profit run-rates (~$16B/year), it could take roughly 7 years for Alphabet to recoup its AI investment.
- Issuing stock while Alphabet's market cap is inflated by AI enthusiasm lets the company raise large sums while diluting shareholders only modestly.
Risks acknowledged
- If the AI boom continues, Alphabet stands to profit substantially from its infrastructure investment.
- By avoiding additional debt, Alphabet would not face bankruptcy risk even if the AI bubble bursts and Anthropic defaults on its spending commitments.
- Alphabet's CEO has himself suggested that trillions in AI data center spending, including by his own company, may not be fully rational.
UBERUberThe host uses Uber as a case study in AI spending excess, noting the company blew through its entire 2026 AI budget in four months and has since capped per-employee AI spending, illustrating that AI productivity gains have not clearly translated into measurable business value.
Key Arguments
- Uber gave engineers unlimited access to Claude Code and burned its full 2026 AI budget in the first four months of the year.
- Uber's COO acknowledged it is hard to draw a direct line between AI token usage metrics and actual useful feature output.
- Uber has instituted an AI spending cap of $1,500 per employee per month, suggesting prior spending was far higher.
Risks acknowledged
- Uber's leadership still views AI as potentially transformational even if near-term productivity links are unclear.
Hedges & Caveats
- The host frames the AI bubble bursting as a hypothetical scenario ('let's say hypothetically'), not a stated expectation.
- He notes Alphabet's own CEO has acknowledged AI spending may be irrational, without the host endorsing a specific stock price outcome.
- Analysis is focused on capital structure and business economics rather than an explicit price or return forecast.