By ByteBot
Apple’s AI chief John Giannandrea is stepping down after seven years marked by Siri stagnation, repeated delays, and a rare public loss of control over the virtual assistant he was hired to fix. The departure, announced December 1, comes months after CEO Tim Cook stripped Siri from Giannandrea’s responsibilities entirely and handed it to Vision Pro creator Mike Rockwell.
Amar Subramanya, who spent 16 years at Google leading Gemini engineering and recently served as Microsoft’s Corporate Vice President of AI, will replace Giannandrea as Apple’s new VP of AI. If the timing seems urgent, it is: Apple is finalizing a $1 billion annual deal with Google to power Siri with the same Gemini technology Subramanya helped build.
Seven Years of Siri Stagnation
Giannandrea joined Apple from Google in 2018 with a clear mandate: revive Siri, the voice assistant Apple shipped in 2011 and has seemingly forgotten about ever since. Seven years later, Siri remains weak compared to ChatGPT, Google Assistant, and even Alexa. The promised AI-powered upgrade has been delayed repeatedly—from iOS 18.4 in April 2025 to iOS 18.5 in May to the current target of March 2026. Internal reports suggest a fully conversational Siri won’t arrive until 2027.
Apple’s software chief Craig Federighi reportedly voiced “strong concerns” that Siri features didn’t work properly in testing, according to MacRumors. Internal testers warned the new Siri “doesn’t compete with today’s chatbots.” Apple executives called the delays “ugly” and “embarrassing” in meetings with staff.
Bloomberg reported that Cook lost confidence in Giannandrea’s ability to execute on product development. By March 2025, Siri was completely removed from his oversight. Apple removed Giannandrea from its executive leadership page two days after the departure announcement.
The Google Gemini Dependency
In what might be the most embarrassing admission yet, Apple is paying Google approximately $1 billion per year for access to a 1.2 trillion parameter Gemini model to power Siri’s core functions. That model is eight times larger than Apple’s current 150 billion parameter offering. Apple’s biggest mobile competitor will effectively run a core iOS feature starting in March 2026.
According to Bloomberg, Gemini will handle Siri’s “summarizer and planner” functions—the components that help the assistant synthesize information and execute complex tasks. Apple maintains that Google’s model will run on Apple’s own servers through its Private Cloud Compute infrastructure, so no user data reaches Google. However, the strategic dependency remains uncomfortable: what happens if Google changes terms or terminates the deal?
Apple says it plans to transition to in-house models “when capable”—a tacit admission that it can’t compete with Google’s LLM technology right now.
A Culture Problem, Not Just Technology
Former Apple employees told The Information that poor leadership and an overly relaxed culture contributed to execution failures. The AI/ML group has been dubbed “AIMLess” internally, while employees refer to Siri as a “hot potato” nobody wants to own. When your AI team is internally nicknamed “AIMLess,” you don’t have a technology problem. You have a culture problem.
The Apple Intelligence launch has been rocky. The feature rollout was delayed from iOS 18.0 to iOS 18.1 and arrived with stability issues. Most notoriously, Apple’s notification summary feature falsely reported that a suspect in a high-profile case had died by suicide, prompting a complaint from the BBC. The same feature incorrectly announced a darts championship winner hours before the final even took place.
Can Subramanya Turn It Around?
Subramanya brings deep expertise in large-scale AI systems. At Google, he rose from staff research scientist to vice president of engineering, leading the team behind Gemini. His five-month stint at Microsoft focused on foundation models powering Copilot. Now he inherits seven years of technical debt, a demoralized team culture, and a product roadmap dependent on a competitor’s technology.
The 46-year-old also knows intimately what Apple is buying. If Apple is going to lean on Gemini while building in-house capabilities, hiring the person who led Gemini’s engineering is strategic.
Nevertheless, leadership alone won’t fix Apple’s AI problem overnight. Analysts note Apple is one to two years behind OpenAI, Google, Meta, and Anthropic. Zeus Kerravala of ZK Research called Apple Intelligence “a debacle.” Forrester’s Dipanjan Chatterjee said “the silence surrounding Siri was deafening.” iPhone sales declined year-over-year despite the Apple Intelligence launch, dashing investor hopes for an “AI super cycle.”
What’s Next for Apple AI
Apple is targeting March 2026 for the Gemini-powered Siri 2.0 launch. Longer-term, the company insists it’s building its own LLMs to eventually replace Google’s technology. Whether Subramanya can fix the cultural issues, catch up technically, and navigate the Google dependency remains uncertain.
What is certain: Apple isn’t just behind in AI. It’s watching the race from the pit lane while competitors lap it. Giannandrea’s departure is less a retirement and more an acknowledgment that Apple’s AI strategy needs more than a tune-up—it needs a complete overhaul.
