A transaction-based personal AI agent business could generate roughly $45 billion annually for Apple, according to a new analysis from Evercore ISI that maps out the company’s path through an emerging competitive landscape.

Evercore ISI analyst Amit Daryanani laid out the financial case in a November 9 research note, arguing that Apple’s existing infrastructure gives it structural advantages that web-based AI competitors would struggle to replicate.

The transaction model assumes 125 million monthly active users each spending $600 on eligible purchases, with Apple collecting a hypothetical 5% fee on those transactions, a scenario Evercore describes as illustrative rather than company guidance.

A separate subscription-based scenario points to roughly $21 billion in annual revenue, built on 125 million paying customers each contributing $14 per month to an AI-enhanced service tier.

Apple’s most powerful asset in this race may be its payment layer, with Apple Pay combined with deep application integration and access to personal user data forming a competitive moat that purely web-based agents cannot easily match.

The scale of Apple’s compatible user base further separates it from current rivals, with between 550 million and 560 million iPhone owners worldwide running hardware capable of supporting Siri AI features, out of a total active base of 1.4 billion iPhone users.

That footprint dwarfs the estimated audiences of competing personal agent platforms, with Meta’s Muse sitting at around 10 million users and Instinct registering approximately 100,000.

Siri, however, remains largely request-driven in its current form, lagging behind rival agents on persistent task execution, third-party application connectivity, and memory retention across sessions.

Evercore identified those three areas as the most critical gaps requiring development before Apple can realistically compete at the higher end of the personal agent market.

More advanced iterations of Siri could eventually complete tasks in the background even when users are not actively engaging with their devices, a capability that would significantly expand the platform’s commercial utility.

Delivering that level of sophistication will demand greater computing investment, spanning cloud infrastructure, model processing, and additional memory and storage capacity, costs Apple would distribute across on-device processing, its Private Cloud Compute system, and third-party cloud providers.

Evercore maintained its Outperform rating on Apple shares alongside a $380 price target, with Daryanani arguing that strong execution on AI agents could both unlock new revenue streams and deepen customer loyalty across the broader Apple hardware ecosystem.