Four exam papers of Apple's new CEO Ternus

Late at night on September 1, newly appointed Apple CEO John Ternus posted his first piece of content on Sina Weibo.

There were only two words: "Hello hello".

No popular emojis, grand visions, or the rigid formality like "it is my great honor". The post was sent from an iPhone 17 Pro Max, with the IP location showing Shanghai. There were only two people in his following list: his predecessor Tim Cook, and Isabel Ge Mahe, Managing Director of Apple Greater China.

Some netizens complained that his profile photo was too blurry, and the operation team quickly replaced it with a high-definition version. This detail is more interesting than the Weibo post itself. It shows how long Apple has prepared for this leadership transition, and how much it values the Chinese market.

On the same day, Apple updated its management page on the official website: Ternus's title became CEO, Cook became Executive Chairman, and the number of board members increased from 8 to 9.

Documents submitted by Apple to the U.S. Securities and Exchange Commission (SEC) show that Ternus has an annual base salary of 3 million U.S. dollars, with target equity awards of 55 million U.S. dollars for fiscal 2027, bringing his total target compensation to about 58 million U.S. dollars; 75% of which are performance-based restricted stock units, whose vesting conditions are directly linked to Apple's total shareholder return relative to S&P 500 constituents. After Cook transitions to Executive Chairman, his target compensation will be reduced to about 47 million U.S. dollars, only half of which is tied to performance.

The capital market also gave the new leader a nice greeting gift. When Apple's stock market opened on September 1, its share price dipped slightly, then rose more than 2.5% during the trading session, pushing its market capitalization to 4.74 trillion U.S. dollars, a sharp increase of 120 billion U.S. dollars (equivalent to over 8 trillion yuan) from the 4.62 trillion U.S. dollars close of the previous trading day. By the close of trading on September 1, Apple's stock price rose 2.61% to close at 325.13 U.S. dollars, with a market capitalization of 4.745 trillion U.S. dollars.

Eight days later, on September 9 (1 a.m. Beijing time on September 10), Apple will hold its autumn launch event at the Steve Jobs Theater in Apple Park, with the theme "Surprise and shine".

This is also Apple's first launch event after a leadership transition since 2011, and Ternus's first major public test as CEO.

According to multiple sources including Bloomberg, the first foldable iPhone Ultra will be unveiled on that day. In his internal letter on the first day of taking office, Ternus said: "We have a huge launch coming next week, and it is bound to amaze everyone."

8 days of preparation. Apple arranged the leadership transition right after Cook's last financial report and before the start of the new product cycle, directly pushing the new CEO into the spotlight. This is clearly a carefully designed arrangement.

But it is also a metaphor: Apple has been in a hurry for this transition for two years, but only left the new CEO 8 days of preparation time.

01 25 Years of an Engineer

Public information shows that Ternus is 51 years old this year, and graduated from the University of Pennsylvania in 1997 with a major in mechanical engineering. He was a member of the swimming team when he was in school, and won the university championships in 50-meter freestyle and 200-meter individual medley.

His graduation project was a mechanical feeding arm that allowed quadriplegic patients to control their food intake through head movements. After graduation, he worked as a mechanical engineer at Virtual Research Systems. This company is now only a footnote in the history of VR, but back in the 1990s, it made head-mounted displays and immersive technologies, which was pretty cool at the time.

In 2001, 26-year-old Ternus joined Apple's product design team, initially working on Mac external displays. He was promoted to manager three years later. In 2013, he was promoted to Vice President of Hardware Engineering, responsible for Mac, iPad and AirPods. He began to take over iPhone hardware engineering in 2020. In January 2021, he succeeded Dan Riccio as Senior Vice President of Hardware Engineering, joined the executive team, and took charge of all Apple hardware engineering. At the end of 2022, Apple Watch hardware was also placed under his management.

This career path seems almost without any twists and turns. For 25 years, an engineer stayed at the same company, working from displays to all categories of hardware.

The official Apple press release set the tone for him as follows:

He led the launch of multiple new product lines including iPad and AirPods, as well as the hardware engineering of multiple generations of iPhone, Mac and Apple Watch; his work on Mac "made the category stronger than ever in its 40-year history"; his team led the redefinition of the iPhone product line, including iPhone 17 Pro, the extremely thin iPhone Air and iPhone 17; he also promoted the use of recycled aluminum composite materials, 3D-printed titanium on Apple Watch Ultra 3, and innovations in repairability.

There are two key nodes here that are worth highlighting alone, because they define Ternus's capabilities.

The first is the Apple Silicon transition.

At WWDC 2020, Ternus was the one standing on the stage announcing that Mac would bid farewell to Intel. This was not just a chip replacement, but a major overhaul that took the performance, power consumption and cost structure of the entire Mac product line back from external suppliers. As a result, Mac achieved revenue of 10.352 billion U.S. dollars in the third quarter of fiscal 2026, a year-on-year increase of 28.7%, becoming Apple's second fastest growing business. This result is quite impressive.

The second is MacBook Neo.

In March 2026, Apple released the MacBook Neo starting at 599 U.S. dollars, with an education price of 499 U.S. dollars. It uses the A18 Pro chip instead of the M-series, starts with 8GB of memory, has an aluminum alloy body, a 13-inch Liquid Retina display, and comes in multiple color options.

IDC data shows that in the March quarter, the product was on the market for only about three weeks, and MacBook Neo shipped about 1.1 million units, exceeding the shipments of MacBook Air M5 (about 900,000 units) and MacBook Pro M5 (about 550,000 units) that were launched in the same quarter; the U.S. accounted for 44% of the total shipments, and India contributed nearly 18,000 units even with only a few weeks of supply.

Cook called customer response "off the charts" on the April earnings call, and acknowledged that supply was tight. According to estimates from Counterpoint Research, Neo may have raised Apple's share in the 400-699 U.S. dollar laptop market from about 2% to around 15%.

"Business Show" believes that MacBook Neo is the most critical sample to understand Ternus. It does not prove that Apple can make cheap computers, but that a manager with a hardware engineering background can restructure the cost structure without sacrificing product appearance, and accurately enter a price band that Apple has never taken seriously before. Using a mobile phone chip to power a computer, using a unified memory architecture to lower the threshold, and retaining brand premium through consistent design language, this can be said to be a textbook demonstration of "system integration" capabilities.

Cook's evaluation of him is: "He has the mind of an engineer, the soul of an innovator, and a leadership heart of integrity and honor." The general evaluation of him inside Apple is more plain: he does not seek the limelight, is friendly to others, and has a solid technical foundation.

The problem lies precisely here. All these strengths are the advantages of "continuity", not the advantages of "disruption". And what Apple needs now is obviously the latter.

Apple's reason for choosing him as the new CEO is almost irrefutable from the perspective of corporate logic: Cook came from operations and supply chain, and took Apple's market capitalization from about 350 billion U.S. dollars to 4 trillion U.S. dollars, annual revenue from 108 billion U.S. dollars in fiscal 2011 to more than 416 billion U.S. dollars in fiscal 2025, with more than 2.5 billion active devices, and the services business grew to a scale of over 100 billion U.S. dollars.

Cook has finished all the business homework. What Apple needs now is another kind of capability: product capability. So it must choose someone who understands products.

This judgment also holds true in business, but when we shift the perspective to the competitive landscape, it is no longer that convincing: the area where Apple most needs change is artificial intelligence, but Apple has chosen the person who is best at maintaining continuity.

This is not hindsight. Some external observations have put it very bluntly: the BBC's evaluation is that "Apple is no longer as innovative as it used to be"; the chief analyst of the research firm Forrester pointed out that Cook "did not bring the company a product like the iPhone that could allow Ternus to sustain 20 more years of success".

The more acute point is that Apple's current strategic shortcoming is AI, and Craig Federighi, Senior Vice President of Software Engineering, is considered by many to be more suitable for leading this transformation. But Apple finally chose Ternus. The subtext of this choice is very clear: in Apple's value ranking, hardware still comes first, and AI can only be an accessory to hardware.

Fighting an offensive war with a defensive posture. It may not be wrong, but it is a choice, not the only solution.

There is another arrangement that cannot be ignored: Cook did not leave.

According to Bloomberg, after stepping down as CEO, Cook "will continue to be deeply involved in Apple's affairs, he is eager to take on more responsibilities, and continue to maintain relationships with Donald Trump and China". As Executive Chairman, he continues to be responsible for communication with global policymakers. The Financial Times commented on this that Cook staying in office "is expected to ease investors' concerns about the company's ability to influence global policymakers".

Translated into plain language: Ternus does not have this capability yet, so Cook cannot leave.

According to Bloomberg, Apple internally compares this pair to Bezos and Jassy of Amazon. But there is an obvious flaw in this analogy: Jassy took over a cloud computing engine that has already been fully operational, while Ternus took over a machine with a fully functional commercial engine but a technical engine that needs to be replaced.

02 Four Hot Potato Exams

Now let's look at the plate that Ternus took over, which I would like to call the "four exam papers" he is facing.

The first one is naturally AI, the debt Apple has owed for two years.

Let's lay out the facts first, numbers are more convincing than adjectives.

At WWDC in June 2024, Apple launched Apple Intelligence, showing a Siri that can understand personal context, know what is on the screen, and perform tasks across apps. This set of features was used as the core selling point of the iPhone 16 for large-scale advertising.

But by March 2025, Apple admitted the delay, pulled down the ads, and said the feature would be launched "in the coming year". Between March and April 2025, the Siri management team began to reshuffle: former Vision Pro head Mike Rockwell replaced John Giannandrea to take charge of Siri, reporting to Federighi; Giannandrea officially retired in April 2026, ending his eight-year tenure at Apple. Rockwell promoted the technical route to shift from "pure self-developed closed" to "self-development + third-party cooperation".

On January 12, 2026, according to reports from CNBC and other media, Apple reached an agreement with Google to use a custom Gemini model with about 1.2 trillion parameters at a cost of about 1 billion U.S. dollars per year. On May 5, Apple reached a 250 million U.S. dollar settlement in the class-action lawsuit over Siri false publicity: eligible U.S. users can get 25 U.S. dollars in compensation per device, and up to 95 U.S. dollars when the number of claims is low, while Apple does not admit any wrongdoing.

By WWDC on June 8, Apple continued to showcase the revamped Siri AI. It is worth noting that this demonstration used pre-recorded real-device footage, rather than the polished concept videos used in 2024.

On July 15, Chinese national cyberspace authorities announced that Apple Technology Development (Shanghai) Co., Ltd.'s "Apple Intelligence" has completed the filing, with the filing date being July 8, about 22 months after the global debut of Apple Intelligence. In September, the new Siri AI was officially launched with iOS 27.

It took about two years and three months from commitment to delivery.

We tend to believe that this is not a simple progress delay, but a paradigm conflict. Apple's most successful methodology in the past two decades is to "keep all key links under full control", designing its own chips, writing its own systems, defining its own interfaces, and managing its own distribution.

This methodology had a crushing advantage in the era of transition from feature phones to smartphones, and from Intel to Apple Silicon. But generative AI does not follow this methodology: the capabilities of large models are emergent, and you cannot "release" a model's capability through a strict schedule. It requires long-term, failure-tolerant investment that does not count short-term returns, and also requires putting user data, computing power, model training and other things that Apple is most reluctant to outsource into an open R&D loop.

Apple is only willing to spend money on things it can control, and AI is exactly a capability that you cannot fully control.

Now look at the capital investment. According to HSBC's calculation, among Apple's predicted sales in 2026, the proportion of capital expenditure used for AI data centers and other related items is only about 2.5%, while this proportion is as high as 39% for the four major cloud vendors Meta, Google, Amazon and Microsoft. Apple's total capital expenditure in fiscal 2025 was 12.7 billion U.S. dollars, while the total capital expenditure of these four companies in 2026 exceeded 700 billion U.S. dollars.

Apple chose another path: it does not build 10,000-GPU clusters, does not hoard GPUs, pays Google about 1 billion U.S. dollars in license fees every year, and spends money on self-developed end-side chips and R&D. In the first quarter of 2026, Apple's R&D expenditure as a percentage of revenue exceeded 10% for the first time, hitting a 30-year high.

This strategy was financially successful, and even once re-priced by the capital market: in July 2026, Apple's market capitalization once rose to 4.95 trillion U.S. dollars, briefly surpassing Nvidia to become the world's number one.

A report from UBS in mid-July pointed out that the capital expenditure of hyperscale cloud vendors is expected to grow by 76% to 673 billion U.S. dollars this year, but the growth rate will plummet to 25% next year. The market began to price for "the peak of capital expenditure growth", and Apple's "asset-light AI" became a safe haven.

So this 2.5% is not stinginess, it is path dependence. Apple is accustomed to the closed loop of "investment - controllability - output", while large models are an open loop of "investment - emergence - re-investment". Using the organization built for the former to fight the war of the latter will always be half a beat slower no matter how much money you spend. Apple bought two years of time, but did not buy two years of experience.

A specific consequence is that, according to reports from multiple foreign media, Apple's most powerful server model AFM 3 Cloud Pro is "customized in cooperation with Google" and distilled from Google's cutting-edge Gemini model, where Gemini acts as the "teacher signal" during the training process rather than the runtime model during deployment.

Apple emphasizes that "when you interact with Apple Foundation Models, you never touch a single line of Google code". This arrangement is completely technically valid and very clever in engineering. But its strategic implication is that the smartest part of Apple's AI system is not made by Apple. For a company that prides itself on "designing even every screw on its own", this is an unimaginable compromise that no one could have foreseen before.

In fact, the problem that is harder to solve than money is people.

By the time Apple filed the lawsuit in July 2026, more than 400 former Apple employees had joined OpenAI or its hardware subsidiary io Products,