Apple announces departure of AI chief John Giannandrea and hires Amar Subramanya to lead restructuring

Apple announced on Monday, December 1, 2025, the departure of John Giannandrea as Senior Vice President of Machine Learning and Estratégia of IA. The executive, who reported directly to CEO Tim Cook, will act as an advisor until spring 2026, when he will retire. The change occurs in Cupertino, Califórnia, during a restructuring in the company’s artificial intelligence sector, motivated by lower-than-expected results from
Giannandrea joined Apple in 2018 from Google and built teams that developed key technologies in machine learning.
However, the launch of Apple Intelligence, announced in June 2024, faced criticism for limited functionality, such as inaccurate notification summaries, which were disabled shortly after debut.
- Key delays identified: Enhanced version of Siri postponed to June 2026.
- Common criticisms: slow integration with apps and lack of advanced customization.
- Comparison with rivals: OpenAI and Google have advanced faster in generative models.
Trajectory from John Giannandrea to Apple
Giannandrea has led advances in privacy and on-device processing since 2018.
It has integrated AI into products like Fotos and Mapas, prioritizing data security.
The transition reflects adjustments to align AI teams with software and operations.
Profile of Amar Subramanya, the new AI leader
Amar Subramanya, a 46-year-old researcher of Indian origin, takes over as vice president of AI, reporting to Craig Federighi, head of software.
With 16 years at Google, he directed engineering for the Gemini assistant, integrating text, image and video in search and YouTube.
In July 2025, he spent six months at Microsoft as Corporate Vice President of AI, focusing on models for Copilot.
Subramanya graduated in engineering from Universidade of Bangalore in 2001 and obtained a PhD at Universidade of Wisconsin, starting his career at IBM.
Areas under the responsibility of Subramanya
Subramanya will lead Apple Foundation Models, machine learning research and AI security assessment.
These sectors include developing generative models for iOS and macOS devices.
The goal is to accelerate innovations for a more contextual Siri in 2026.
Other teams from Giannandrea will migrate to Sabih Khan, COO, and Eddy Cue, services.
- Foundation Models: Foundation for tools like Genmoji and email digests.
- ML Research: studies in natural language processing.
- AI Safety: protocols to avoid bias and errors in responses.
Apple Intelligence challenges since launch
Apple Intelligence debuted in October 2024 with features like Writing Tools and Image Playground.
Users reported slowdowns on older devices, limiting adoption.
In 2025, the company expanded support for languages such as Portuguese and Spanish, but maintained a focus on English.
Delays in Siri generated a 16% loss in the market value of Apple in relation to peers like Microsoft.
Tim Cook’s statements about the transition
Tim Cook praised Giannandrea for building strong foundations in AI.
He highlighted Federighi as pivotal for future customizations on Siri.
Cook emphasized that the changes position Apple for smarter, more reliable experiences.
The statement came in an official statement, without details about immediate succession.
Initial reactions in the technology market
Analysts see the hiring of Subramanya as a sign of urgency to compete with ChatGPT and Gemini.
Shares of Apple rose 0.5% after the announcement, but analysts predict volatility.
AI experts highlight Subramanya’s expertise in model scalability.
The move comes amid talent departures to Meta and OpenAI.
History of restructuring in the AI division
In March 2025, Cook relocated Siri to Mike Rockwell, creator of Vision Pro.
The company canceled self-driving car project to redirect engineers to AI.
These adjustments aim for more fluid integration between hardware and software.
In 2024, Apple invested billions in chips for on-device AI.
Expectations for updates in 2026
The advanced version of Siri will include context understanding in conversations.
New features will involve image editing via prompts and priority notifications.
Subramanya should prioritize security testing to avoid past incidents.
Apple plans to support 10 more languages, expanding its global reach.
AI integration into Apple ecosystem
Tools like Genmoji are coming to Mac in January 2026 via macOS Sequoia 15.3.
iOS 18.4 will bring improvements to searches and integrated knowledge.
The strategy emphasizes privacy by processing data locally on A18 chips.
Rivals like Google use cloud more aggressively, but Apple prioritizes user control.
Impact on global competition in AI
The move reinforces the war for talent, with Microsoft losing Subramanya after six months.
OpenAI advances with updates to ChatGPT, focusing on speed and customization.
Google maintains leadership with Gemini, integrated with Android and search.
Apple seeks to balance innovation with regulatory compliance in Europa and Ásia.
Advances in AI Security and Assessment
Subramanya inherits protocols to mitigate risks in generative models.
The team will evaluate biases in virtual assistant responses.
In 2025, Apple disabled features due to inaccuracies in news summaries.
Focus now on benchmarks to measure accuracy in everyday tasks.
Preparation for language expansion
Apple Intelligence will support French, German and Japanese in 2026.
This targets markets like Europa and Ásia, where rivals already dominate.
Subramanya’s Gemini multilingual expertise accelerates adaptations.
Internal tests have been taking place in Cupertino laboratories since November.
Redistribution of teams post-Giannandrea
AI Infrastructure goes to Sabih Khan, optimizing data centers.
Search and Knowledge migrates to Eddy Cue, integrating with services.
These moves align AI with Apple’s daily operations.
The objective is to reduce silos and speed up product deployments.
Context of investments in AI at Apple
The company has allocated US$5 billion to chips and research in 2025.
Compared to $10 billion for Microsoft, the focus is on-device efficiency.
Subramanya must optimize costs in model training.
Expected results include quarterly reports starting in March 2026.






