Why did Apple shift to licensing Gemini instead of building its own large language model?

The speaker suggests Apple may be prioritizing speed, cost, and reliability by licensing existing models rather than building a proprietary LLM from scratch. The idea is to power apps like Siri rather than owning a model outright, turning the AI race into a potential cost-saving approach.

What happened to Apple's 2024 Apple Intelligence and Siri promises?

The video describes the 2024 announcement as promise-heavy, with large swaths of promised functionality never arriving. Users saw the same basic Siri behavior and had their queries routed to Google or ChatGPT rather than seeing tight, in-app actions.

Who left Apple's AI leadership in 2025?

The video notes that in June 2025, Apple's head of foundational AI left for Meta, and other top engineers began departing as well, contributing to internal disruption and 'Pandemonium' around Siri and Apple Intelligence.

How does Gemini compare to Apple's internal AI model in terms of size?

Gemini is described as around 1.2 trillion parameters, significantly larger than Apple's internal model at about 150 billion parameters, highlighting the scale difference involved in the licensing deal.

What privacy rationale does Apple cite for running models on its own chips and cloud?

Apple reportedly argues that running models on its own chips in Apple-controlled cloud servers enhances user privacy by avoiding third-party infrastructure, a claim Bloomberg is cited as supporting.

What broader takeaway does the video offer about AI spending in the tech industry?

The video suggests that the trillions spent on AI by some companies may not be the most efficient or sustainable path. It hints at a shift toward leveraging external models or focusing on practical app-level AI rather than massive in-house LLMs.

Why Apple Just Gave Up on AI

ColdFusion

Science & Technology5 min read14 min video

Nov 11, 2025|935,210 views|28,146|4,805

Coldfusion TV Dagogo Altraide Technology Apple Google Samsung Facebook Tesla

Save to Pod

Key Moments

On this page

TL;DR

Apple grapples with AI lag, leans on Google Gemini for Siri amid internal turmoil.

Key Insights

Apple acknowledges a widening AI gap, with Siri lagging behind competitors.

Internal turmoil—leadership changes, fractured teams, and lawsuits—hampered progress.

Bloomberg reports Apple will license Google's Gemini to power Siri on a private Apple cloud.

The deal raises questions about the value of owning massive AI models vs licensing mature ones.

Consumer demand for AI features in smartphones remains modest, suggesting AI is not yet a must-have.

THE RISE AND FALL OF APPLE AI

Apple’s public stance on AI shifted dramatically from late 2023 to 2025. Executives at the time projected a long-term, integrated AI未来 for iOS, yet by 2025 observers labeled Apple as falling behind. Siri remained behind rivals, and delays around Apple Intelligence were described as ugly and embarrassing by a senior director in 2025. Features announced for the iPhone 16 failed to reach users, fueling questions about the value of AI in phones and whether Apple could ever meet its own high standards given the track record.

SIRI'S STRUGGLE AGAINST COMPETITORS

The transcript contrasts Siri with the capabilities of Gemini-powered assistants. Siri often delivered Google search results or failed to execute nuanced tasks, while Gemini demonstrated stronger context handling, multi-step planning, and even accurate media control. Comparisons in demonstrations illustrate a meaningful capability gap, not just a delay. This gap isn’t only about speed; it’s about reliability and intelligent behavior within apps. Apple’s challenge extends beyond voice responses to delivering integrated, context-aware actions across iOS, which competitors have managed to execute more consistently.

INTERNAL FISSURES AND LEADERSHIP SHIFTS

Behind the public narrative, Apple’s AI program suffered from internal strife. By mid-2025, the head of Apple’s foundational AI left for Meta, and other top engineers departed. The divided teams—one focused on Siri, another on broader Apple Intelligence—fueled a culture of delays and blame. Lawsuits over false advertising indicated a mismatch between marketing promises and engineering reality. This internal fragmentation slowed progress and undermined the credibility of Apple’s AI initiatives, contributing to the perception that the company could not deliver the promised capabilities.

PROMISES VS REALITY: APPLE INTELLIGENCE AND SIRI

Apple announced Apple Intelligence and a revamped Siri in 2024, but the rollout largely never materialized. In-app actions, personal-data context, and seamless automation remained elusive, while ChatGPT and other tools filled gaps for writing, image generation, and analysis. After further delays in 2025, a new version of Siri was teased for 2026, but the ongoing governance and execution issues suggested many promised features would remain disappointingly vaporware. The mismatch between promise and delivery highlighted a fundamental misalignment within Apple’s AI program.

THE GEMINI DEAL: A STOPGAP SOLUTION

Bloomberg reported that Apple is pursuing a strategic pivot: licensing Google’s Gemini to power Siri, with the model running on Apple’s private cloud for privacy and control. Gemini is projected to be around 1.2 trillion parameters, far larger than Apple’s internal 150 billion-parameter effort. Apple would pay roughly $1 billion per year for access and use Gemini to enable features like summaries and multi-step planning. This arrangement signals a pragmatic pivot from in-house build-out toward a high-performance, externally sourced core while continuing to emphasize privacy.

PRIVACY-FIRST COMPUTING: ON-DEVICE VS CLOUD

Apple defends its privacy-first stance by framing the Gemini integration as an on-device–friendly, privacy-preserving approach: models on Apple-provisioned hardware and in Apple-controlled cloud environments reduce data exposure. The plan is to blend external AI capabilities with Apple’s own privacy controls, aiming to reassure users who value data security. While this could placate privacy concerns, it also raises questions about the user experience and how seamlessly a third-party AI model can be woven into the iPhone’s native workflows without compromising brand trust.

THE ECONOMICS OF AI: COSTS, BENEFITS, AND THE ARMS RACE

The story reframes the AI arms race as a cost-benefit calculation. If LLMs become commodities, it may be cheaper to license top models and focus on app-level experiences rather than building bespoke models. The smartphone analogy helps: hardware (data centers), OS (models), and apps (applications). Licensing a mature AI for Siri could unlock meaningful capabilities quickly while preserving Apple’s control over privacy and user experience—arguably a smarter allocation of resources than sustaining a prolonged in-house build-out in a volatile AI market.

LICENSING VS BUILDING: APPLE'S STRATEGIC CALCULATION

Apple’s approach mirrors what Samsung has already done by basing Galaxy AI on Google’s Gemini. Licensing reduces risk, accelerates time-to-value, and protects capital. Yet it also introduces questions about long-term differentiation and control: does relying on external models dilute the Apple brand if the core AI experience isn’t uniquely theirs? The narrative suggests Apple is embracing a pragmatic path—prioritizing reliability and privacy through partnerships rather than pursuing aggressive, in-house dominance in AI research and development.

MARKET REACTIONS: APPLE'S HARDWARE PERFORMANCE AND AI BEHAVIOR

Despite AI struggles, Apple’s hardware performance remains a bright spot. iPhone 17 demand reportedly rose in its first year, with strong growth in China, while MacBook momentum grows as Windows frustrations persist. The broader consumer landscape shows AI features are not yet a decisive factor in upgrading; surveys indicate low willingness to switch devices for AI alone. The mixed market signal suggests Apple can weather AI missteps as long as core devices deliver compelling value, but AI remains a meaningful, not sole, differentiator.

LONG-TERM IMPLICATIONS FOR APPLE AND THE AI INDUSTRY

The Gemini licensing revelation could influence how tech giants approach AI strategy for years. If LLMs become commodity-like, the emphasis may shift toward apps, privacy-centric design, and seamless integration rather than owning the most powerful models. Apple’s pause-and-partner approach might set a precedent for balancing reliability and innovation with cost controls. This shift invites broader reflection on whether the AI investments of the last few years will yield durable competitive advantages or simply fuel a protracted arms race with incremental gains for many players.

TAKEAWAYS AND OPEN QUESTIONS

Key takeaways emphasize a potential pivot in AI strategy: licensing mature models can offer faster, safer, privacy-aligned improvements than building from scratch. Open questions remain about preserving Apple’s brand experience with third-party models, maintaining robust privacy protections, and determining how much consumers will value on-device AI versus cloud-enabled capabilities. Viewers should consider whether AI integration should be a core, differentiating feature or a peripheral enhancement that complements a premium hardware strategy.

Mentioned in This Episode

●Products

●Software & Apps

●Tools

●Concepts

●People Referenced

Common Questions

According to the video, Apple is reportedly paying Google about $1 billion per year for a custom Gemini model to power Siri, with that model running on Apple's private cloud for security. This deal is framed as a stopgap while Apple continues to work on its own AI capabilities.