Why Apple Is Restructuring Its AI Leadership Team
Apple is updating how its AI organization reports and executes. The shift clarifies ownership of core model work, day-to-day AI operations, and service delivery without signaling a pivot in technical direction.
John Giannandrea, Senior Vice President for Machine Learning and AI Strategy, is stepping down from his executive role and will serve as an advisor until his planned retirement in spring 2026. Apple has appointed Amar Subramanya as Vice President of AI, reporting to Craig Federighi, Senior Vice President of Software Engineering. Parts of John's former organization will now report to Sabih Khan (operations) and Eddy Cue (services).
The Moves That Matter
- Leadership continuity: John remains in an advisory capacity through 2026 to ensure a clean handoff of active AI programs and infrastructure.
- Clear ownership: Amar takes responsibility for Apple Foundation Models and related research, including ML research, AI safety, and evaluation across device, edge, and cloud.
- Stronger platform tie-in: Amar reports to Craig Federighi, bringing core model development closer to software engineering and client-side integration.
- Service and operations linkage: Teams moving to Eddy Cue and Sabih Khan connect AI delivery with media/services and supply chain execution.
What Changes, What Stays the Same
John's group historically covered Apple Foundation Models, Search, Knowledge ML research, and AI infrastructure-work that touches everything from on-device inference to cloud-assisted flows like Siri. Those functions continue, with new reporting lines and expanded oversight by Federighi.
Tim Cook has reiterated that AI has long been central to Apple's strategy, with a more personalized Siri expected next year. The message: continuity on product direction, tighter accountability on execution.
Why This Restructure Makes Strategic Sense
- Model-to-product velocity: Placing foundation models under Federighi shortens the path from research to iOS, macOS, and app experiences.
- On-device plus cloud balance: Apple Intelligence relies on low-latency, private on-device processing complemented by cloud assistance when needed. This structure supports both tracks.
- Governance and risk: Explicit mandates for AI safety and evaluation formalize guardrails as Apple scales model usage.
- Operational readiness: Routing parts of the team to operations and services leaders links AI capabilities with distribution, quality, and content/business systems.
- Continuity through transition: Keeping John involved reduces risk to critical infrastructure and ongoing model programs.
Who Is Amar Subramanya, And Why He Fits
Amar brings experience from Microsoft as Corporate VP of AI and 16 years at Google, where he was Head of Engineering for the Gemini Assistant. He has shipped AI into consumer products at scale and understands latency, reliability, and cost trade-offs across device and cloud.
That background is directly relevant to Apple Intelligence features that must coordinate on-device compute with server-side inference-especially for voice commands, context, and live network behavior.
What Executives Should Watch Next
- Apple Foundation Models roadmap: Which capabilities ship on-device vs. server, and how that split affects privacy, latency, and cost.
- Siri upgrades: Timing, scope, and third-party integrations across apps and services.
- Safety and evaluation: How Apple measures model quality, bias, and reliability before and after release.
- Compute strategy: Local silicon advantages vs. external compute needs; potential partnerships for training and inference capacity.
- Talent flow and M&A: Senior hires in research and infrastructure; targeted acquisitions to accelerate model or tooling gaps.
Signals For Your Own AI Org Design
- Separate model platform from applied product teams, but keep them tightly coupled through shared leadership and SLAs.
- Own the full stack: data, training, evaluation, deployment, and observability across device, edge, and cloud.
- Build a formal AI safety and evaluation function with authority to gate launches and monitor post-release performance.
- Plan transitions: For key leaders, put advisory bridges in place to protect delivery during handoffs.
- Tie AI to operations and services early so capacity, quality, and go-to-market stay in sync.
Context And Further Reading
Upskill Your Team
If you're aligning leadership, safety, and model ops in your organization, curated learning paths can speed up adoption and governance.
Bottom line: this is a structural clean-up to accelerate delivery and reduce risk, not a reset. Expect faster model-to-product cycles, a more capable Siri, and tighter coordination across device, cloud, and services.
Your membership also unlocks:
