Ternus takes the helm at Apple this Tuesday: the Siri rebuilt on Gemini arrives this month, and it's all or nothing

Ternus takes over at Apple this Tuesday: the Siri rebuilt on Gemini arrives this month, and it's all or nothing

🔎 A hardware engineer facing Apple's biggest software challenge

On September 1, 2026, Apple enters an unprecedented era. John Ternus, until now SVP Hardware Engineering, becomes effective CEO. Tim Cook moves to executive chairman. The transition has been official since the announcement of the Apple April 2026 press release.

But the real symbol lies elsewhere. Ternus's first major test isn't a physical product. It's the deployment of the new Siri, entirely rebuilt on Google's Gemini model, expected this very month. A ~$1 billion a year bet, according to the analysis by Michael Parekh.

The timing is tight. WWDC 2026 unveiled the promises in June. The reality of the public rollout arrives now, under a new boss who has never managed a software puzzle of this magnitude.

The key points

• John Ternus becomes Apple CEO on September 1, 2026, with Tim Cook moving to executive chairman after 15 years at the helm of the company.
• The new Siri, rebuilt on Google's Gemini, is expected to roll out to the general public this month, following positive early reviews.
• The contract with Google is estimated at ~$1B/year by analysts, an unprecedented amount for a third-party AI partnership.
• Jefferies downgraded Apple to Underperform on August 10, 2026, citing risks related to the delay in generative AI.
• Ternus brought back Johny Srouji as VP Silicon, reunifying hardware and chip under a single vertical leadership.
• The September keynote is expected to introduce the first foldable iPhone, but it is Siri that will decide Apple's relevance in AI.

Recommended tools

Ternus CEO: an atypical profile for the moment Apple is experiencing

John Ternus is not a CEO with a traditional background for a $4.6T market cap tech group. An engineer by training, he spent his career at Apple in hardware: chips, cameras, industrial design. As Fortune points out in its analysis of the transition, he is a "low-key engineer" arriving in a world dominated by software and AI.

Under Cook, the market cap grew from around $350B to ~$4.6T. The track record is massive. But the AI legacy is thin. Apple Intelligence, unveiled in 2024, never truly convinced. The version initially planned for spring 2026 stumbled over accuracy issues, according to MacRumors.

Ternus therefore inherits a paradox: the richest company in the world is perceived as being behind on the most defining issue of the decade.

The timeline of Cook's fifteen years

StartupFortune lays out a detailed timeline of Cook's fifteen years: the Apple Watch, AirPods, the M1/M2/M3 transition, crossing the $4T mark. But also the hesitations regarding generative AI, the key departures from the AI division, and an Apple Intelligence that never reached the bar set by ChatGPT and then Gemini.

The report from TradingKey notes that Jefferies downgraded to Underperform on August 10, 2026. The signal is rare for Apple. It reflects a concrete concern from the markets about the group's ability to monetize AI.

Siri on Gemini: why Apple abandoned its own engine

The decision to rebuild Siri on Gemini is an admission. Apple failed to produce a competitive in-house LLM in time. Not against GPT-5.5, not against Gemini 3 Pro Deep Think, not even against Claude Sonnet 4.6 in terms of smooth mobile integration.

At WWDC 2026, Tim Cook unveiled the new Siri as the centerpiece of his final keynote as CEO. CNBC describes it as Cook's AI legacy: a gamble to catch up with ChatGPT and Gemini, even as he prepares to step down.

Quartz confirms that Apple is preparing a Siri rebuilt on Gemini, with a deployment planned "later this year" — which is to say now, September 2026.

The $1 billion per year contract

The partnership with Google to integrate Gemini into Siri represents about $1 billion per year, according to the analysis by Michael Parekh on Substack. This is a colossal amount for an AI supply agreement, but it needs to be put into perspective.

Apple sells more than 220 million iPhones per year. The cost per device comes out to about $4.50/year. For a feature considered the main differentiator of the next sales cycle, it's a reasonable price. Especially compared to the cost of an in-house LLM that did not exist.

Why not GPT-5.5 or Claude?

The question comes up often: why Google and not OpenAI or Anthropic? Several factors are at play. Google is already Apple's default search provider (a contract estimated at $15-20 billion/year). The distribution infrastructure exists.

But above all, Gemini 3.1 Pro achieves a score of 92 on the General benchmark, tied with GPT-5.5 and GPT-5.4 Pro. The model is competitive. And Google had an interest in locking down this partnership to prevent OpenAI from crashing into the Apple ecosystem.

What actually changes with the new Siri

The Siri rebuilt on Gemini is not a minor update. ABC7 reports that Apple is positioning it as a direct strategic response to ChatGPT and Gemini on mobile.

Concretely, the new Siri gains three major capabilities.

Multimodal contextual understanding

The current Siri does not handle chained queries well. The Gemini version allows for extended conversation memory, with visual context understanding (screen, photos, camera). This is the baseline level expected in 2026.

Action execution within the ecosystem

Not just answering, but acting. Booking a restaurant in the native app, moving an appointment, compiling a summary of emails and broadcasting it as a voice message. All via structured API calls between Gemini and iOS frameworks.

A dedicated app and the Dynamic Island

WWDC 2026 also announced a dedicated Siri app, breaking out of the Settings app's orbit, and an integration with the Dynamic Island for persistent interactions. This is a fundamental UX change: Siri becomes an entry point, not a secondary voice shortcut.

iOS 27 Extensions: the multi-provider game-changer hub

Beyond Gemini, Apple is preparing a deeper architectural shift with iOS 27 Extensions. Siri would no longer be tied to a single model provider, but would become a multi-provider AI hub.

The idea is simple but powerful: the user could choose or let Siri route to the best model based on the task. Claude Opus 4.7 for complex reasoning, Gemini 3.1 Pro for general queries, Grok 4.1 for real-time searches.

What this means for the Gemini contract

If Siri becomes a multi-provider hub, the exclusive contract with Google loses some of its strategic value. Gemini would remain the default engine, but not the only one. It's a delicate position for Ternus: he must roll out Gemini this month as an immediate win, while preparing the multi-provider architecture that will make this partnership less central.

Claude, Gemini, Grok in Siri

The announcement of iOS 27 Extensions clearly positions Apple as a model distributor rather than a creator. It's a technological admission but potentially a stroke of business genius: why build when you can aggregate and take a commission on each routing?

The Ternus/Srouji duo: Apple's vertical response

The most underestimated move in this transition is the return of Johny Srouji. Michael Parekh notes it in his analysis: Ternus brought Srouji back as VP Silicon, reunifying chip and hardware leadership under a single command.

This is a strong signal. Srouji is the architect of the A and M chips, the man who gave Apple its most enduring competitive advantage: complete mastery of silicon. By bringing him closest to Ternus, Apple is playing its vertical card.

Why the chip is the key to AI

Models like Gemini 3.1 Pro or GPT-5.5 run partly on device, partly in the cloud. The quality of the experience depends directly on the capabilities of the Neural Engine and the NPU integrated into the Apple Silicon chip.

By aligning Srouji with the AI strategy, Ternus is likely preparing the A20 or the chip for the next cycle to specifically optimize Gemini inference. This is the advantage that no Android competitor can replicate: Google controls the model, but not the chip. Apple controls both.

A hardware CEO is not a handicap if hardware is the strategy

The main criticism against Ternus — "he's a hardware guy in an AI software world" — forgets that mobile AI is won as much on the chip as on the model. A CEO who understands silicon can make integration decisions that a purely software CEO wouldn't see.

The competitive landscape: where AI stands in September 2026

To measure Ternus's challenge, one must look at the current model landscape. Apple's lag is not theoretical: it is measurable.

The agentic ranking

In agentic reasoning, OpenAI's GPT-5.5 dominates with a score of 98.2. Gemini 3 Pro Deep Think follows at 95.4. Claude Opus 4.7 Adaptive reaches 94.3. These three models are far ahead of what Apple could produce in-house.

Even when looking for less expensive alternatives, DeepSeek V4 Pro (Max) reaches 88 overall, and Kimi K2.6 climbs to 88.1 in agentic. The market's baseline level has risen considerably.

The overall ranking

In the general benchmark, the situation is similar. Gemini 3.1 Pro, GPT-5.5, and GPT-5.4 Pro are all at 90-92. Claude Opus 4.7 and Grok 4.1 follow at 90. The competitiveness threshold is now above 85, a level that no internal Apple effort has publicly reached.

Comparison of relevant models

Apple does not appear in any of these rankings. Hence the Gemini deal.

The foldable iPhone: a product, not a strategy

The September 2026 keynote is expected to present the first foldable iPhone, as Quartz reports. This is a major product event. But it is not what will define the success or failure of Ternus.

The foldable market represents about 3-4% of global smartphone sales in 2026 (IDC estimate, market data). It is a premium segment, not a volume driver. The foldable iPhone will be a symbol of hardware innovation capability — exactly Ternus's territory.

But the real question is: what does this foldable iPhone do differently with AI? If the new Siri works poorly, the foldable won't save anything. If it works well, the foldable becomes a premium form factor for a superior AI experience.

The pitfall of a product without a platform

Apple's history is full of beautiful products without a matching software ecosystem. The Vision Pro is the recent example: a technically impressive headset, held back by a lack of content and killer use cases.

The foldable iPhone risks the same dynamic if Siri-Gemini doesn't deliver. A screen that folds is cool. An assistant that actually understands what you want and executes it is strategic.

The AI division shake-up: what happened behind the scenes

Apple's delay in generative AI is no accident. It is the result of organizational decisions made between 2023 and 2025. The AI division experienced key departures, multiple reorganizations, and chronic tension between the hardware team and the software team.

The problem with the privacy-first approach

Apple long insisted on on-device processing for privacy reasons. This is an honorable stance, but technically limiting in 2026. Competitive models require cloud infrastructure. Gemini 3 Pro Deep Think does not run entirely on a mobile NPU.

The pivot to Gemini cloud is a partial abandonment of this doctrine. Apple had to concede that privacy washing did not compensate for a third-tier model.

The departures and the failed recruitment

Several lead researchers from the AI division left Apple between 2024 and 2026 to join OpenAI, Anthropic, or startups. The recruitment of top-level AI talent was complicated by Apple's reputation regarding scientific publications (historically low) and by its culture of secrecy.

Ternus inherits an AI division that must be rebuilt at the same time it is expected to deliver. The Gemini contract buys time, but it does not solve the underlying problem.

What "make or break" means for Ternus

The expression is no exaggeration. The Siri-Gemini rollout this month is a tipping point for three reasons.

Market credibility

Jefferies at Underperform is rare. If Siri-Gemini disappoints, other analysts will follow. The $4.6T market cap already incorporates an Apple premium. This premium relies on confidence in the capacity to innovate. An AI failure would call this confidence into question.

The iPhone sales cycle

The new Siri must be ready for the sales cycle that opens with the September keynote. It's the main selling point of the next iPhone. Not the foldable, not the camera, not the chip: AI. If the rollout is buggy or if early mainstream reviews are mixed, the sales cycle takes a direct hit.

The Google contract as a trap door

If Gemini in Siri works, Apple buys time to build or acquire an in-house model. If it doesn't work, the $1B/year contract becomes a burden: Apple pays Google for a feature that doesn't differentiate, all while depending on a strategic competitor.

❌ Common mistakes

Mistake 1: Confusing Ternus with a software CEO

The most widespread analysis since the announcement is that Ternus is "the wrong profile" because he comes from hardware. This ignores the fact that Apple's enduring advantage has always been hardware-software vertical integration. With Srouji back, Ternus precisely controls the part (the chip) that will determine the quality of on-device inference. The real risk is not his hardware background, it's his lack of experience in managing complex software partnerships on a global scale.

Mistake 2: Thinking that Gemini in Siri solves everything

A good engine doesn't make a good car. Gemini 3.1 Pro is an excellent model, but the Siri experience also depends on OS integration, API calls, network latency, and voice UX. Apple has historically been average at integrating third-party services (see Siri with Yelp, then with Wolfram Alpha). The integration risk is real.

Mistake 3: Ignoring the multi-provider hub

Focusing solely on Gemini means missing the most structuring announcement: iOS 27 Extensions. Ternus's strategy is probably not "Siri = Gemini" but "Siri = routing platform". The Gemini contract is a first step, not the destination.

❓ Frequently Asked Questions

Does John Ternus have the legitimacy to lead Apple?

Yes. Fifteen years at Apple, SVP Hardware Engineering since 2021, in charge of A/M chips, industrial design, and cameras. He oversaw the Intel→Apple Silicon transition, one of the most successful technical maneuvers in the industry. His deficit is in software/AI strategy, not in overall legitimacy.

Why doesn't Apple build its own LLM?

Apple tried, but internal results did not reach the competitiveness threshold. GPT-5.5 (98.2 agentic), Gemini 3 Pro Deep Think (95.4) and Claude Opus 4.7 (94.3) have widened a gap that Apple's AI team has not closed. The Gemini contract is a catch-up move, not an ideological choice. Srouji's return suggests Apple wants to build a silicon advantage for a next generation of models.

Will the new Siri be available on all iPhones?

The initial rollout will likely target the iPhone 16 Pro and later, equipped with chips that have an NPU powerful enough for partial on-device inference. Older models will rely more on the cloud, with potentially higher latency. Apple has not yet communicated the exact compatibility grid.

What is the risk if the rollout fails this month?

The risk is threefold: user disappointment, analyst pressure (Jefferies is already at Underperform), and increased reliance on Google without any compensation in terms of experience. Ternus would then need to accelerate iOS 27's multi-provider architecture and potentially acquire an AI lab to internalize the capability.

Is the foldable iPhone linked to the AI strategy?

Indirectly. The foldable form factor offers more screen surface for AI interactions (agentic interfaces, control panels, previews). But the AI added value does not come from the folding: it comes from the quality of Siri's comprehension and execution. The foldable is a multiplier, not a founder.

✅ Conclusion

John Ternus takes the helm at Apple with the hottest dossier in the company's recent history: deploying a rebuilt Siri on a third-party engine, at a scale of hundreds of millions of users, this very month. The Ternus/Srouji duo and the iOS 27 Extensions architecture are the real strategic levers — the Gemini contract is just the opening. To follow the evolution of the models that could integrate this hub, check out our comparison of the best free LLMs and the ChatGPT vs Gemini analysis.