Brad uses ChatGPT rather than Google to help his son Lincoln study for AP History, as it's a better knowledge-discovery tool, but this use case isn't monetized by Google anyway (knowledge graph searches), so it doesn't cannibalize Google's high-value searches.

Even if meta-cap companies are concerned about capex guidance and spending levels, the competitive environment creates a Pascal's Wager dynamic: if AI upside is potentially infinite and all competitors are investing heavily, no company can afford to pull back capital investment, even if returns are uncertain, because falling behind in AI capability would be catastrophic to their competitive position.

OpenAI put its Foundry (chip manufacturing) ambitions on hold and is instead talking to TSMC about building a chip, suggesting recognition that internal foundry operations would require excessive capital that conflicts with other priorities like model development and training, and indicating some restraint or reevaluation of spending despite reported high burn rates.

TSMC announced a broad 20% price increase across its foundry business, which Brad anticipated would occur and viewed as overdue evidence that semiconductor supply is tight and capacity constraints are real, indicating that mega-capex spending is running up against material supply constraints and cost escalation.

Nvidia Jenson Huang estimates that 9 trillion dollars of cumulative capex on 200 million GPUs is 'reasonable' and possibly even 'too small,' and the NPV math suggests that if AI achieves 'artificial superintelligence' and replaces 5% of the global workforce (worth ~$9 trillion annually in wages), then 9 trillion in cumulative capex over seven years would be NPV positive and justifiable as a single-entity investment.

Services markets are typically 10x larger than software markets; if AI replaces services (call center agents, insurance form fillers, back-office workers), the capex payoff is enormous because it targets the much larger services market, not just software spend.

The election is likely 70-80% priced into markets as a Trump victory, though not 100%, meaning there is 20-30% downside scenario if Harris wins, in which case the trades that have been unwinding based on Trump probability would sharply reverse (democratic spread trade is down 20-30% in 30 days), creating significant portfolio implications for investors betting on Trump policies.

Brett Taylor's company (Sherpalo Ventures or similar) is showing that customer service can be 100% replaced with AI while improving customer satisfaction, which would validate the massive capex spending thesis if that pattern holds across multiple verticals.

Mega-fund investors function as 'free rolls'—GPs get paid regardless of outcomes, treating the carry (20% of profits) as a call option, so they win if any one of their portfolio companies becomes a 10x+ power law outcome, but don't lose anything if the other investments fail.

Piff (Public Investment Fund) and other sovereign wealth funds are consolidating the number of GPS they work with and are increasingly prioritizing investments within their home countries over global diversification, which may reduce capital available for overseas venture funds.

Over the last decade, data center engineering and construction breakthroughs at Meta and Google were largely taken for granted in the background, but data center engineering is now becoming a primary source of competitive advantage in the AI era.

Only mega-cap companies at the scale of the Magnificent Seven can sustain $50 billion+ annual AI compute spend; for smaller companies, even at a $250B spend scale, the opportunity is inaccessible.

xAI announced expansion of the Colossus cluster in Memphis from 100,000 GPUs to 200,000 GPUs, and other mega-cap companies (Google, Meta, Microsoft, Amazon) are reporting capex at a ~$250 billion annual run rate, with all companies discussing how to secure facilities capable of powering 1-3 gigawatt clusters, indicating that the prior question of 'will we need 200,000 GPU clusters' is now settled and the focus is on facility capacity and power infrastructure.

A new browser company called Browserbase and conceptually similar approaches are exploring whether to build a new browser designed for agents (with composable APIs instead of buttons) rather than humans, representing an alternative architecture to the 'computer use' approach of scraping human-designed websites and clicking buttons, but this would require rebuilding all underlying transaction infrastructure (like Stripe) to work with agents.

There was speculation that Google would release 'Jarvis', their version of agentic AI, this week, but Google has already discounted this risk by trading at the lowest multiple in the Mag 7, suggesting markets are pricing in future search disruption.

Analysts in January 2023 forecasted Nvidia would do $26 billion in revenue for 2023, but the company actually did $60 billion—a $34 billion miss—representing the single greatest failure of forecasting the world has ever seen due to analysts' focus on crypto winners and inability to imagine the generative AI transformation.

Nvidia produces approximately 4 million dollars of revenue per employee and 2 million dollars of profits/free cash flow per employee, representing extraordinary operational efficiency and culture of execution.

Nvidia will never be worth more than a billion dollars was a prediction someone made about the company, and it is now worth over $1 trillion, demonstrating the misjudgment of fixed TAM assumptions.

U.S. productivity growth has slowed from 2.5-3% annually in the 1990s to 1.8% in the 2000s to the slowest rate on record in the past 10 years, suggesting the world may be on the verge of a dramatic productivity expansion if AI manufacturing of intelligence is as transformative as described.

The Cambrian moment in generative AI occurred with ChatGPT's launch in November 2022, and analysts' failure to forecast Nvidia's 2023 growth was attributable to their focus on cryptocurrency winners rather than imagination of the generative AI transformation.