Historical Behavior

Apple's consistent strategy in AI investment has been "late entry + acquisitions." In the 2010s, it acquired Siri, in 2016 it acquired Turi (a machine learning platform), and in 2020 it acquired NextVR (spatial computing), never developing large models from scratch. During Cook's tenure (2011 to present), capital expenditures have consistently accounted for less than 10% of revenue, while Amazon, Microsoft, and Google have all exceeded 15% in the same period. In 2024, Apple briefly invested in Didi Chuxing (rumored to be $3 billion), but completely exited by 2026, demonstrating its financial discipline in new ventures.

Capital Path

Clear funding flows: buybacks + dividends + service ecosystem. In fiscal year 2025, Apple repurchased about $92 billion in stock and paid out about $15 billion in dividends, totaling over 75% of free cash flow. This "contractionary capital allocation" sharply contrasts with Nvidia's "expansionary" approach (with capital expenditures expected to exceed $40 billion in 2026). Apple has chosen to outsource AI R&D to Google, essentially trading a fixed cost of "$1 billion/year" for flexible returns without bearing data center depreciation and chip inventory risks.

Analogy + Industry Position

A similar case is Netflix, which chose to rent AWS rather than build its own data centers in the 2010s, successfully avoiding massive capital expenditures and focusing on content production. Apple is currently in the "AI application layer harvest phase"—with underlying models provided by Google/Microsoft and hardware ecosystems supported by its own devices, AI features (like Siri upgrades and photo recognition) are becoming catalysts for iPhone upgrades. Compared to Nvidia (selling shovels) and Microsoft (selling cloud services), Apple is at the bottom of the industry chain but has the highest profit margins.

Structural Judgment

This essentially represents a transfer of pricing power due to technological substitution. The marginal cost of large AI models approaches zero (Google's Gemini has served 1 billion users), but Apple converts AI features into "upgrade necessities" through hardware bundling (2.5 billion devices). This explains why Apple's market value surpassed $5 trillion while Nvidia's stock price corrected—markets realize that the ultimate "AI monetization" is not at the infrastructure level, but at the endpoint level. As technology becomes homogenized, ecosystem lock-in and hardware penetration rates become sources of pricing power.

ABAB News · Cognitive Laws

Asset-light does not mean no investment; it means letting others bear depreciation.
The ultimate goal of AI is not in computing power, but in device penetration rates.
Cash flow is more resilient than growth rates across cycles.