Skip to content
IFAI

Industry

Meta poaches Apple's AI models chief

Published Reach 48

Meta just recruited Apple's head of AI models, Ruoming Pang, to join Mark Zuckerberg's new AI superintelligence team. Pang previously oversaw the training of foundation models behind Apple Intelligence and on-device features. Why it matters: After raiding DeepMind and OpenAI, Meta lands specialized expertise to boost its small on-device AI capabilities.

Meta poaches Apple's AI models chief
Image: tomshardware.com

Mark Zuckerberg is on an absolute hiring tear in Silicon Valley, and this time, he went shopping straight from Apple's backyard.

Apple's head of AI models, Ruoming Pang, is leaving the iPhone maker to join Meta, according to a report from Bloomberg on Monday. Pang is set to take on a leadership role inside Zuckerberg's newly formed AI superintelligence unit.

Inside Apple, Pang ran the in-house team responsible for training the foundation models powering Apple Intelligence and various on-device AI features.

Apple's AI struggles trigger an exodus

It is no secret that Apple's in-house models have failed to impress. By most accounts, they lag far behind frontier systems from OpenAI, Anthropic, and Meta itself.

The gap is reportedly wide enough that Apple has considered turning to external third-party models to power its upcoming Siri overhaul.

Now, the lack of momentum seems to be catching up with team morale. Sources told Bloomberg that Pang's departure could be just the first of many exits from Apple's troubled AI division.

Meta stockpiles edge computing muscle

Pang is the latest high-profile addition to Zuckerberg's aggressive talent spree. Over the past few months, Meta has poached top researchers and leaders from Google DeepMind, OpenAI, and Safe Superintelligence.

While Apple's large models may have underwhelmed, Pang brings rare and deep expertise in designing compact, efficient models built specifically for on-device execution.

For Meta, which is aggressively pushing AI onto smart glasses, headsets, and consumer hardware, Pang's background in small models gives the superintelligence group crucial engineering know-how.

Why it matters

This departure shines a harsh spotlight on Apple's internal generative AI bottleneck.

We think losing the lead architect behind Apple Intelligence's foundation models right as the platform is rolling out is a major setback. It suggests that internal friction over model performance is real, and it reinforces the narrative that Apple may have to lean heavily on third-party partners like OpenAI rather than mastering the technology in-house.

For Meta, Zuckerberg's playbook looks remarkably sharp. By pairing massive compute with open checkbooks, he is scooping up specialized veterans across every layer of the stack, from frontier reasoning to local device execution.

The talent war is entering a brutal consolidation phase, and companies struggling to show frontier-level progress risk seeing their best minds walk straight to the competition.

Sources (2)

  1. [1]
    bloomberg.com(opens in a new tab)Originalbloomberg.com
  2. [2]

Best of AI, daily