Apple announced on Monday that John Giannandrea, the executive leading their artificial intelligence efforts since 2018, is leaving the company. He will remain as an advisor through the spring season.

Taking his place is Amar Subramanya, a well-known leader from Microsoft, who previously dedicated 16 years to Google, most recently directing engineering for the Gemini Assistant. This appointment is considered a strategic move, as Subramanya possesses extensive knowledge of the competitive landscape.

This personnel change is being viewed as a significant restructuring. Looking back, it appears almost unavoidable, especially considering the challenges faced by Apple Intelligence, the company’s response to the growing prominence of technologies like ChatGPT, since its debut in October 2024. Initial assessments of Apple Intelligence have varied from disappointing to deeply concerning.

The initial period following its release was particularly difficult. A feature designed to summarize notifications into concise updates produced a number of inaccurate and embarrassing headlines in late 2024 and early 2025. For example, the BBC filed two complaints after Apple Intelligence incorrectly reported that Luigi Mangione, accused in the death of UnitedHealthcare CEO Brian Thompson, had died by suicide (which was not the case) and that Luke Littler, a darts player, had won a championship before the competition concluded.

The anticipated improvements to Siri also became a source of negative attention for Apple.

A Bloomberg report published in May detailed the difficulties Apple was experiencing with its AI initiatives. The report revealed that Craig Federighi, Apple’s head of software, discovered during testing weeks before the planned April launch that many of the advertised features were not functioning as expected. This led to an indefinite postponement of the launch and subsequent class-action lawsuits filed by iPhone 16 purchasers who had been promised an AI-powered assistant.

According to Bloomberg, Giannandrea had already been relieved of some responsibilities by that time. The news outlet reported that Tim Cook removed Siri from Giannandrea’s purview in March, assigning it to Mike Rockwell, the creator of Vision Pro. Apple also took control of its confidential robotics division away from Giannandrea.

Bloomberg’s investigation highlighted organizational issues, including poor communication between AI and marketing teams, budgetary discrepancies, and a leadership crisis so severe that some employees derisively referred to Giannandrea’s team as “AI/MLess.” The report also noted a growing number of AI researchers leaving Apple for competitors like OpenAI, Google, and Meta.

Apple is now reportedly utilizing Google’s Gemini to power the next iteration of Siri, a surprising development given the long-standing and intense rivalry between the two companies spanning over 15 years in areas such as mobile operating systems, app stores, browsers, maps, cloud services, smart home technology, and now, artificial intelligence.

Giannandrea joined Apple after a tenure at Google, where he led Machine Intelligence and Search. During his time at Apple, he was responsible for the overall AI strategy, machine learning infrastructure, and the development of Siri.

Subramanya will now assume these responsibilities, reporting to Federighi with the primary goal of accelerating Apple’s progress in the field of AI.

This represents a pivotal moment for the company. While other companies are investing heavily in large-scale AI data centers, Apple has prioritized processing AI tasks directly on users’ devices using its custom Apple Silicon chips, a privacy-focused strategy that avoids collecting user data. (When more demanding tasks require cloud processing, Apple uses Private Cloud Compute, servers designed to process data temporarily and then delete it.)

The ultimate success of this approach remains to be seen. Apple’s strategy involves inherent trade-offs. Specifically, on-device models are generally smaller and less powerful than the extensive models used by competitors’ data centers, and Apple’s reluctance to gather user data means its researchers are relying on licensed and synthetic data instead of the vast amounts of real-world information used by its rivals.