Hardware Veteran Takes Over Software Crisis: On Ternus's First Day, Apple's AI Outsourcing Bill Is Already on the Table

On September 1, 2026, Apple officially announced: John Ternus, 51, with a mechanical engineering background and 25 years of hardware experience, officially assumed the role of Apple CEO. Tim Cook transitioned to Executive Chairman, ending a 15-year leadership cycle that began in 2011.

This is a personnel arrangement widely read by the outside world as a "signal" — in an era of the most intense software competition, Apple chose a hardware engineer.

Outsourced Core, Outsourced Risk

On Ternus's first day, the bill on his desk was already clear: according to multiple financial media reports, Apple signed an agreement with Google this year to pay approximately $1 billion annually for a license to use Google's custom Gemini model — a system with 1.2 trillion parameters — to rebuild Siri's reasoning layer. The deal was officially announced at WWDC on June 8, 2026, with the new Siri shipping alongside iOS 27.

The logic behind this arrangement is not hard to understand: Apple has fallen years behind in natural language understanding and in-house large model development. ChatGPT and Claude have long since gone from followers to the de facto standard in voice interaction, while Siri — the product that nearly invented the voice assistant category in 2011 — has gradually become a laughingstock in users' eyes. When self-development cannot close the gap in the short term, external procurement is a commercially defensible choice.

But this choice carries a structural suspense: the AI experience Apple promises its users is actually determined by another company's model.

Apple introduced the "Private Cloud Compute" architecture, claiming that data between users and Gemini servers is anonymized and will not be used for Google's model training. This privacy narrative carries considerable weight within Apple's ecosystem, but it solves the problem of "who sees your data," not "who controls the quality of your AI experience." When Gemini hallucinates, responds incorrectly, or degrades in service, Apple's ability to explain and correct errors for end users is extremely limited — it has no permission to modify the model, only to wait for Google to update it.

This is not a hypothetical risk. Since Apple Intelligence launched in 2024, the frequency of feature cuts and delayed releases has already accumulated a substantial negative impression. After outsourcing the core reasoning layer to a partner, this uncertainty did not disappear — it simply migrated from inside to outside.

Two Readings of the Hardware DNA

The market's assessment of Ternus shows a clear divide, and the divide itself is worth a closer look.

The bullish case is most directly articulated by Dan Ives, partner at investment firm Yorkville Ives & Co.: "Tim Cook left behind a house in excellent condition. Now, the key is how Ternus defines the AI chapter." Ives predicts Ternus will focus on hardware innovation while preserving the supply chain system Cook built — in other words, let what is already running well keep running, and direct his energy into what he knows best.

Creative Strategies analyst Carolina Milanesi added a frequently overlooked perspective from user behavior: "Consumers still buy hardware first, and that will be the case for a long time. If you're not interested in hardware, you won't discover the value of AI." This assertion targets another camp — some analysts have argued that software chief Craig Federighi would have been better suited for the CEO role, because Apple's most pressing need right now is software catch-up capability.

Milanesi's rebuttal has some merit, but it also has its limits. Hardware is the carrier of the AI experience, but it is not the AI experience itself. A chip with sufficient compute power cannot automatically produce a sufficiently intelligent assistant. After Apple replaced Siri's reasoning core with Google's model, the relationship between Ternus's hardware expertise and the quality of AI intelligence has become more indirect than ever.

Bank of America analyst Wamsi Mohan offers a more actionable expectation: Ternus has the opportunity to be "more assertive" in device design and user interface. Apple's recent increases in R&D and capital expenditure also hint at how seriously management takes the AI race. But between investment signals and outcome, there is still a long way to go.

Nine Days: A Launch Event That Puts Both Hardware and AI to the Test

Ternus does not even have the luxury of a gradual transition period. On September 9, just eight days after taking office, he will host Apple's annual product launch, where he must deliver on two promises on stage at the same time.

The first is Apple's first foldable iPhone. This is Ternus's home turf — foldable phones demand exceptionally high standards in materials science, hinge engineering, and structural design, precisely the core areas he has led for decades. Rival Samsung has been cultivating the foldable market for years, and Meta's Ray-Ban smart glasses have already built consumer goodwill. Apple's foldable iPhone arrives late, but expectations for its craftsmanship have been pushed to an extremely high bar.

The second is the public debut of the new Siri — the first consumer-facing demonstration of the Gemini-powered version. This is an area where Ternus's hardware DNA cannot directly move the needle. Its performance will depend on Google's model quality, Apple's engineering execution at the integration layer, and the real-world balance between privacy architecture and response speed.

One launch event, simultaneously testing his greatest strength and his biggest structural weakness.

A Bigger Question: Who Backs Apple's AI Promise?

Apple has a durable brand asset throughout its history: the things it promises, it tends to actually deliver. From "1,000 songs in your pocket" to "the fastest chip in any smartphone," Apple's product claims rest on hard, engineering-verifiable metrics. That credibility was built up over decades.

But the AI experience is a promise of a different nature. Its quality cannot be captured in a single number, its stability depends on the continuous evolution of the reasoning layer — and now that reasoning layer is controlled by Google. When Apple says "a smarter Siri" at a launch event, consumers are putting their trust in Apple, but the intelligence quality they actually get is being decided by Google.

This misalignment was not created by Ternus, but he will be the first Apple CEO called upon to answer for it.

Meanwhile, Apple faces another potential competitive front in AI — according to Fortune, former Apple designer Jony Ive is collaborating with OpenAI on "a range of hardware products," which means Apple's aesthetic DNA and hardware ecosystem may, for the first time, face direct competition from a "former insider." Apple also filed a lawsuit against OpenAI in July 2026 over trade secret theft, which OpenAI has disputed — the trajectory of this legal battle will continue to shape both sides' public standing in the AI arena.

Assessment

John Ternus taking over as Apple CEO makes sense on both timing and personnel logic: he is the person at Apple with the deepest understanding of hardware, and Apple's revenue core remains hardware, just as the primary reason consumers choose Apple remains hardware.

But the biggest test he faces lies precisely outside his home turf. Apple's current AI strategy — filling an internal capability gap with an external model — is a pragmatic short-term choice, but it mortgages Apple's AI reputation to a supplier it cannot fully control. If a Gemini update makes the new Siri underperform expectations, Apple's brand endorsement will bear the brunt, while the engineering fix authority sits in another building.

The true measure of Ternus's tenure is not whether that launch event nine days from now goes smoothly, but whether Apple can build an AI reasoning capability that does not fully depend on an external partner within the next two to three years. If Apple's in-house AI roadmap takes shape within a visible timeframe, this $1 billion annual Gemini bill will become a footnote to a transition period; if in-house development continues to lag, the bill will stand as a long-term symbol of Apple forfeiting its AI sovereignty.