Episode Summary
Executive Summary: The episode examines how AI is reshaping U.S. tech giants, investor sentiment, and the next phase of computing. Guests argue that hyperscalers are still heavily investing in AI infrastructure, are supply constrained, and are only beginning to unlock an application layer. While tariffs and macro volatility are affecting the narrative and some costs, core AI capex plans remain intact, and the real debate is which consumer and enterprise applications will ultimately capture the most value.
Main Topics: AI as the Third Computing Shift (Priority: 5/5): Eric Sheridan frames AI as the next major computing wave after desktop/web 1.0 and mobile/web 2.0, with AI and spatial computing forming the basis of a new Web 3.0 era. Infrastructure, Platform, and Application Layers (Priority: 5/5): The discussion distinguishes between AI infrastructure (model building and training), the emerging platform layer, and the still-early application layer where consumer and enterprise use cases will eventually dominate value creation. CapEx Commitments and Supply Constraints (Priority: 5/5): Large tech companies remain committed to high AI-related capital spending through 2025 and likely into 2026, though growth rates may normalize; they are still constrained by supply and capacity. Tariffs and Macro Volatility (Priority: 4/5): Tariffs are beginning to affect AI capex through higher input costs for data center buildouts, while macro uncertainty is more likely to pressure operating expenses than long-duration AI investments. Search, Advertising, and Consumer Behavior (Priority: 5/5): The transcript explores whether chatbots will cannibalize search, but the argument is that search is evolving rather than disappearing, and Google still dominates monetization of commercial queries. Investor Sentiment and Valuation (Priority: 5/5): Investors have moved from enthusiasm for infrastructure spend to scrutiny over scaling and returns; sentiment is now muddied by tariffs, while major AI beneficiaries still trade at relatively reasonable valuations. Winners in the Application Layer (Priority: 4/5): The biggest open question is which AI applications will become indispensable consumer or enterprise products, analogous to Uber or Airbnb reshaping behavior in Web 2.0.
Key Arguments: AI adoption is progressing unusually fast, but it is still in early innings; the market is moving from model training to inference and then to applications. Large tech firms are continuing heavy AI capex because they view it as a multi-year strategic imperative, not a short-term macro trade. Investor focus is shifting from who sells the picks-and-shovels infrastructure to who can convert AI into scalable products and revenue. Tariffs matter first through higher construction and component costs, not necessarily by reducing the amount of AI buildout planned. Search is not being “killed” by chatbots; instead, the user interface is changing, and the number of queries to computers is increasing. The commercial monetization of AI-driven search-like behavior still largely accrues to Google today. The most important future battleground is the consumer AI assistant and enterprise application layer, where new winners may emerge. Unlike prior tech cycles, incumbents have enough capital to defend and shape the transition rather than being quickly displaced.
Data Points: ChatGPT monthly active users: Over 800 million - Used to illustrate how quickly the AI cycle scaled after launch. Time since ChatGPT emergence: About 2.5 years - Shows how rapidly the AI shift has advanced compared with prior computing cycles. Meta capital intensity: Approaching 40% - CapEx as a share of revenue, described as peak capital intensity. 2025 CapEx growth expectation: Still elevated through 2025 - Companies reiterated commitment to spending on AI workloads this year. 2026 CapEx growth expectation: Mid-teens growth - Eric expects growth to slow from this year’s 40%-60% range. Google Cloud growth: 28% - Latest quarter growth cited as evidence of AI and cloud demand. AWS growth: 17% - Latest quarter growth cited as investor-relevant AI infrastructure demand. Azure growth: Well into the 30s - Used to show strong hyperscaler demand for AI-related workloads. Alphabet CapEx: $75 billion - Estimated 2025 spending on AI and infrastructure. Meta CapEx: ~$70 billion - Estimated 2025 spending on AI and infrastructure. Amazon CapEx: $100-$110 billion - Estimated 2025 spending on AI and infrastructure. Combined capex of Alphabet, Meta, and Amazon: Over $250 billion - Illustrates the scale of incumbent investment capacity in AI.
Pivotal Quotes: "we are first batter, second strike of the first inning" — Andy Jassy, as cited by George Lee: Used to describe how early the AI cycle still is. "The application will play out in your consumer computing and your enterprise computing." — Eric Sheridan: Summarizes the expected evolution from infrastructure to real-world AI products. "The pie of us asking computers questions has exploded." — Eric Sheridan: Explains why search and query behavior is growing rather than disappearing.
Implications: AI investment remains durable, but the next value shift depends on applications, not just infrastructure. Investors should watch for real user behavior changes, enterprise monetization, and which AI assistants become daily defaults.
About Goldman Sachs Exchanges
In each episode of "Exchanges," people from the firm share their insights on developments shaping industries, markets and the global economy.