Amid a deepening crisis over its delayed Siri overhaul, Apple has replaced its artificial intelligence chief, John Giannandrea, with former Google and Microsoft executive Amar Subramanya.
Reporting directly to Senior Vice President of Software Engineering Craig Federighi, Subramanya joins the company after a tenure of less than five months at Microsoft. The aggressive hire signals a strategic pivot to consolidate AI development under the core operating system team.
Giannandrea will transition to an advisory role until spring 2026, aligning his departure with the targeted release of the company’s rebuilt “Siri V2” architecture.
Promo
A Strategic Consolidation Under Software Engineering
Apple officially announced the immediate transition of John Giannandrea from his role as Senior Vice President of Machine Learning and AI Strategy via a press release. Dissolving the standalone AI organization he built since joining from Google in 2018, the move marks a significant structural shift.
Giannandrea will remain as an advisor until Spring 2026, a timeline that conspicuously aligns with the projected release of the delayed until 2026 Siri V2 update.
His departure marks the end of an era defined by cautious, privacy-first development that ultimately left the company vulnerable to the generative AI boom.
Amar Subramanya has been appointed as the new Vice President of AI, tasked with leading Foundation Models and AI Safety. His mandate is to accelerate the development of the underlying models that power Apple Intelligence features.
Significantly, the reporting line has shifted: Subramanya reports directly to Craig Federighi, Senior Vice President of Software Engineering. Signaling a major shift, this structural change ends AI’s status as a standalone silo, integrating it directly into the core operating system teams responsible for iOS and macOS.
Extending beyond software, the reorganization sees Sabih Khan (Operations) and Eddy Cue (Services) absorb the remaining hardware and services components of Giannandrea’s former organization.
Distributing responsibilities in this manner suggests a desire to embed AI capabilities across every facet of the company’s operations. The official press release details the transition:
“Apple today announced John Giannandrea, Apple’s senior vice president for Machine Learning and AI Strategy, is stepping down from his position and will serve as an advisor to the company before retiring in the spring of 2026.”
“Apple also announced that renowned AI researcher Amar Subramanya has joined Apple as vice president of AI, reporting to Craig Federighi.”
“The balance of Giannandrea’s organization will shift to Sabih Khan and Eddy Cue to align closer with similar organizations.”
Speculation regarding Giannandrea’s diminishing influence had mounted for months, particularly after product development duties were reportedly shifted to Mike Rockwell in March. That internal shift effectively stripped the AI chief of control over the consumer-facing product roadmap.
Back then Mark Gurman of Bloomberg captured the internal sentiment surrounding the difficult role, suggesting a sense of relief that Giannandrea was likely “relieved” Siri since then was “someone else’s problem.” Reflecting the immense pressure of the role, this sentiment mirrors high turnover across the industry as companies race to ship products.
Anatomy of a ‘Wreck’: The Technical Debt Crisis
Driving the leadership change is a failed engineering strategy that attempted to graft generative AI onto Siri’s aging, database-driven architecture. This approach proved fundamentally flawed, creating a brittle system unable to handle the complexity of modern LLMs.
Internal sources previously described the resulting hybrid code as “a ‘wreck’”, where fixing one bug would frequently spawn multiple new issues. Hamstrung by technical debt, engineers were trapped in a cycle of patching legacy code rather than iterating quickly.
Real-world consequences of this instability included “phantom” restaurant reservations and a $95 million class-action settlement regarding privacy violations. Damaging the brand’s reputation for reliability, these public failures forced a strategic rethink.
To address these foundational flaws, Apple has initiated a complete “V2” architectural rebuild, necessitating a delay of advanced features until iOS 26.4 in Spring 2026. Pushing the most transformative capabilities of Apple Intelligence nearly two years behind competitors, this timeline represents a significant setback.
Engineers are currently using a secret internal chatbot named ‘Veritas’ to stress-test these new capabilities in a sandbox environment, isolating them from the live production code. Validating the new architecture before it faces the public is critical.
Allowing for the testing of complex, multi-turn agentic workflows without risking the stability of the current public-facing Siri, the “Veritas” tool represents a shift toward more rigorous, isolated testing protocols designed to prevent the embarrassing glitches of the past.
The Talent War: Poaching for Production Experience
Subramanya’s hiring represents a significant escalation in Silicon Valley’s talent war, as he joins Apple after a tenure of less than five months at Microsoft. His rapid departure from Redmond highlights the intense competition for executives who can actually ship product.
Prior to his brief stint at Microsoft, he spent 16 years at Google, giving him deep institutional knowledge of the architectures Apple is now trying to replicate. His background includes work on the very foundation models that Apple is now struggling to build.
Paying a premium for executives with recent, tangible experience in shipping large-scale AI products, the aggressive recruitment marks a departure from promoting internal candidates. It signals a need for fresh perspective and proven execution.
Driving this urgency is a widening competitive gap; while Apple reorganizes, competitors are launching production-grade agents. Google has already rolled out Gemini Live to millions of devices, and Amazon is expanding its Alexa’s user milestones.
Just today, ByteDance and ZTE unveiled an agentic AI smartphone prototype with an LLM integrated at the OS level, beating Apple to a key innovation milestone. Demonstrating the exact kind of deep system integration that Apple has promised but not yet delivered, this device highlights the stakes.
Gan Lin, a member of the Doubao Mobile Assistant team, highlighted how the industry has accelerated while Apple remained static, noting that “It’s been more than a decade since Apple launched Siri in 2011. In 2022, the release of ChatGPT inspired the entire industry to reimagine the potential of mobile voice assistants.”
Illustrating the external pressure facing Cupertino, this quote underscores how Chinese competitors move rapidly to fill the void left by Apple Intelligence’s absence in the region.
To bridge the gap until the internal rebuild is complete, Apple has been forced into a pragmatic agreement to test Google’s Gemini models for immediate feature needs. Providing a stopgap solution, this partnership covers features like summarization and image generation.
Simultaneously, a new ‘Answers’ team is working to build long-term internal search capacity to eventually reduce this dependency on rivals. Aiming to survive the current crisis while building the foundation for future independence, this dual-track strategy is critical.