Apple's AI Agent Bet: Why the iPhone Base Changes Everything
As seen on the 24/7 Wall St. homepage on November 20, 2026.
Apple Plans To Build Its Own AI Agent to Rival OpenAI & Anthropic
- Mark Gurman reports Apple is building an autonomous AI agent to act on users' behalf
- Dave Lee: Apple's iPhone base could make its assistant the dominant consumer AI, like the Apple Watch tie-in
- Apple's capex stays tiny next to Google's, with on-device processing limiting data center reliance
- Anthropic's compute crunch: 75% of spend on R&D, $900B in committed compute, Neo Cloud dependence
Distribution decides this, not model quality: if Apple ships a capable agent, the iPhone installed base hands it consumer AI by default, a direct threat to the assistants Meta and Alphabet are spending far more to build. The counterweight is Anthropic, whose $900 billion in committed compute may not be deliverable.
Continue ReadingShow less
Mark Gurman reports that Apple is developing an autonomous AI agent designed to act on users' behalf, putting it in direct competition with offerings from OpenAI, Anthropic, Meta, and Alphabet. The episode, discussed on Bloomberg Businessweek with Bloomberg Opinion columnist Dave Lee, frames the move as a question of reach.
Dave Lee's core argument is that distribution is the decisive variable in consumer AI. The iPhone installed base gives Apple a structural advantage that neither Meta nor Alphabet can simply buy their way past, however much more capital they deploy on AI infrastructure.
Apple's approach leans on on-device processing, which keeps its capital expenditure small relative to rivals like Google. That architecture limits data center reliance and means Apple can roll an agent to hundreds of millions of devices without needing the cloud-compute budget its competitors are committing.
Sponsored
The counterweight to Apple's story is Anthropic, whose trading debut framed the second half of the conversation. Anthropic has $900 billion in committed compute and carries a dependence on Neo Cloud, raising real questions about whether that committed compute is reliably deliverable at scale.
The Apple Watch tie-in Lee cited illustrates the pattern: when Apple embeds a feature into the core iPhone experience, adoption tends to follow the installed base by default. An AI agent baked into the iPhone would operate under the same dynamic, making early positioning by rival assistant makers harder to defend over time.
Mentioned: AAPL, META, METAM, GOOGL, GOOG, GGLBP, GOOGN, GOOGM, GGLAP