Alex Sacerdote - How to Invest Through Technology Cycles - [Invest Like the Best, EP.477]

Episode Summary

Executive Summary: Alex Sacerdote explains Whale Rock’s framework—S-curve adoption, competitive advantage, and underappreciated earnings power—and applies it to AI. He argues the AI stack is concentrating into an oligopoly, with Anthropic his highest-conviction position because of code leadership, enterprise brand, and scale. He also sees chips/infrastructure as the clearest near-term beneficiaries, while remaining cautious on AI application-layer software.

Main Topics: Whale Rock’s investing framework (Priority: 5/5): Sacerdote centers decisions on three filters: where a technology is on its S-curve, whether it has durable competitive advantages, and whether earnings power is underestimated by the market. Why Anthropic is the highest-conviction position (Priority: 5/5): He traces Anthropic from an early private investment to a core holding, arguing code, enterprise traction, recursive model improvement, and scale have made it one of the leading AI franchises. AI stack: chips and infrastructure vs. applications (Priority: 5/5): He believes the bottom of the AI stack—compute, data centers, networking, memory, cooling, and power—is still the most attractive way to invest, while app-layer software remains less proven and more vulnerable. S-curves, adoption timing, and market structure (Priority: 4/5): He uses historical examples like smartphones, cloud, EVs, and AWS to show how technology adoption accelerates when barriers fall, and why leaders often compound far longer than skeptics expect. Competitive advantages in digital businesses (Priority: 4/5): Sacerdote argues digital companies can build strong moats through network effects, scale, industry-standard status, brand, IP, and ecosystem lock-in, sometimes as powerful as offline moats. Research process and private-market access (Priority: 4/5): Whale Rock uses deep scuttlebutt-style research, thousands of meetings, and strong company relationships to gain access to private rounds and build conviction before capital is deployed. Risks to the AI thesis (Priority: 3/5): He highlights regulation, a slowdown in model progress, open-source catch-up, and the possibility that key players lose position as the main threats to the current AI trade.

Key Arguments: AI is creating a new stack, and investors should own the scarce layers first: chips, infrastructure, and the leading model providers. Anthropic became compelling because coding was the real unlock for AI monetization, and its models increasingly dominate high-value enterprise/code use cases. The AI model market is not pure commodity; different models have real performance distinctions by task, creating room for multiple winners. Enterprise AI adoption is still early—only a tiny fraction of knowledge workers are using it in a truly transformative way—so the growth runway is long. The best technology investments are often bought when earnings power is underestimated, allowing strong businesses to look optically cheap. Large-cap tech can still produce alpha because the market systematically underestimates the durability and scale of dominant digital franchises. AI infrastructure should benefit regardless of which model provider wins, because compute demand and hardware intensity are rising across the board. Application-layer software is riskier today because AI features are not yet generating meaningful revenue and incumbents may get squeezed before durable moats emerge. Effective research in technology still requires human judgment, company visits, and relationship-building; AI helps with blocking and tackling but does not replace insight.

Data Points: Whale Rock assets under management: more than $17 billion - Size of Whale Rock Capital Management Whale Rock performance over past three years: roughly 44% per year - Described as one of the best-performing funds AI expense review automation at Ramp (ad copy): 85% automated with 99% accuracy - Sponsor message, not part of interview content Anthropic valuation at investment: $180 valuation - Sacerdote says Whale Rock invested at this valuation Anthropic daily active users: 14–15 million DAUs - Used to illustrate early-stage adoption Anthropic capacity: about half of what they need right now - He says compute supply is still constrained OpenAI/Anthropic coding spend example: $100 a day on tokens - Internal/observed usage illustrating willingness to pay Annualized coding spend per user: $20,000–$30,000 per year - Derived from the $100/day token spend example Implied coding market size: about a half trillion-dollar market - Based on ~20 million coders paying the above level AI adoption among knowledge workers: 10 bips (0.10%) - He cites Sunder’s estimate of true AI usage Enterprise AI penetration: less than 1% - He describes enterprise application AI as very early Infrastructure layer penetration: 10% - His estimate of infrastructure/S-curve penetration AI model market size expectation: three to five trillion dollars - His estimate of the eventual opportunity Cloud TAM estimate: $800 billion - Benchmark used to compare AI’s potential NVIDIA valuation multiple in 2023: 4x earnings - Example of buying early in an S-curve at low apparent multiples Tesla valuation multiple in 2019: 5x earnings - Example of buying on EV inflection Apple valuation multiple: 4x earnings - Example of underappreciated earnings power Amazon/AWS purchase context: for free - His characterization of AWS optionality inside Amazon Celestica valuation when noticed: 8x earnings - Mentioned as a hardware supply-chain example Top research cadence: 2,000–3,000 face-to-face meetings per year - Whale Rock’s management-team engagement Private-company meeting share: 10%–15% - Portion of management meetings involving private companies General AI usage number: 800 million people - He says many are using AI in a limited search-like way Consumer AI adoption S-curve example: radio reached ~100% penetration in 7 years - Historical benchmark for fast consumer adoption EV adoption caution: 10–15% penetration - He notes EVs hit a wall around this level in his view Market share/AI “modified rule of 40”: AI sales share + category market share - His framework for judging AI exposure in software companies Rule-of-40 example: 30% sales in AI + 30% market share = 60 - Illustrative threshold he says is attractive

Pivotal Quotes: "We call it a backwards L curve." — Alex Sacerdote: Describing current AI adoption as very early but steepening fast "The big kicker was code. And this is the true unlock of AI." — Alex Sacerdote: Explaining why Anthropic and the coding market changed his conviction "The world doesn't think exponentially." — Alex Sacerdote: On why investors miss the long duration and nonlinear earnings power of S-curves

Implications: Listeners should take away that AI value may accrue first to compute, infrastructure, and a few model leaders, while software incumbents face near-term pressure. The main challenge is timing: adoption is early, but the leaders may compound much longer than the market expects.