Since the second half of 2026, one of the most significant legal filings in Silicon Valley has been Apple's lawsuit against OpenAI for trade secret misappropriation and breach of contract. The initial partnership for developing Siri was once so harmonious, but the current situation has become deeply awkward. One month after filing the lawsuit, Apple petitioned the court for a preliminary injunction, seeking to prohibit OpenAI and former Apple employees from using or disclosing its trade secrets, and to bar them from further accessing confidential information. Additionally, Apple has demanded that OpenAI return all of its proprietary materials. Hours later, OpenAI retaliated by publishing an article on its website titled "Apple is getting this wrong," publicly responding to Apple's trade secret lawsuit. The article described Apple's lawsuit as "hasty, aggressive, and oddly personal," and included screenshots of multiple emails and iMessages as evidence. The piece contained several pieces of concrete evidence. For instance, Apple's external lawyer, Gross, confused two Asian surnames, mistakenly sending an email intended for someone else to OpenAI's general counsel, claiming in the email that he had already spoken with them. After OpenAI pointed out the error, Apple's law firm acknowledged the mistake. Furthermore, OpenAI stated in the article that Apple alleged a former senior executive had stolen trade secrets after moving to OpenAI, but evidence showed that it was current Apple employees who proactively sought documents, project information, and contact details from the executive. The executive only downloaded and sent the confidential files to facilitate a proper work handover. However, the essence of this lawsuit goes far beyond trade secrets. In its complaint filed on July 10, Apple revealed a fact: over 400 former Apple employees now work at OpenAI. These individuals cover nearly all critical areas, including product design, hardware development, screen technology, antenna engineering, supply chain management, and hardware procurement, essentially forming a complete hardware development chain. This is not just ordinary talent flow. It represents OpenAI using Apple's people to rebuild a new Apple. The lawsuit is merely the surface of this conflict. What truly deserves attention is: why would an AI company, founded on software and algorithms, systematically poach talent from a hardware giant? What does it intend to do? And how will Apple fight back?
Everything stems from an acquisition. In 2025, OpenAI acquired io Products, a hardware design firm founded by former Apple chief designer Jony Ive, in a $6.5 billion all-stock deal. However, the core of this transaction goes far beyond simply acquiring a design team. Ive worked at Apple for nearly 30 years, leading the design of almost all iconic products, from the iPod and iPhone to the Apple Watch. Under his influence, io Products' core team consisted almost entirely of Apple veterans. This effectively gave OpenAI a ready-made Apple hardware design team. After the acquisition, Ive and his team began planning consumer hardware product lines for OpenAI. With the funds, Ive's first thought was to expand the team, naturally drawing talent from his former employer. Yet, the person who truly angered Apple was not Ive, but a senior executive named Tang Tan. Tan worked at Apple for 24 years, serving as the Vice President of Product Design for iPhone and Apple Watch before leaving, making him a key figure in Apple's hardware design system. In early 2026, he joined OpenAI as Chief Hardware Officer, directly overseeing all of OpenAI's hardware operations. Tan not only understood Apple's products but also knew the people working there. He was aware of which teams were working on what projects, which engineers were critical, and even knew their salary levels and career development bottlenecks. Apple's complaint alleges that Tan systematically identified and recruited Apple employees, using internal Apple project codenames during interviews to extract information about unreleased products. Apple also claimed that Tan would directly ask Apple employees still in the interview process to bring physical Apple parts with them to the interview for "demonstration and discussion." These parts included batteries, logic boards, system-in-package (SiP) modules, and casings. Many interviewees later reported being startled by this request, with one stating, "I didn't even know these things could be taken out of the office." Another key figure mentioned in the complaint is Chang Liu, who worked at Apple for eight years as a senior system electrical engineer, contributing to iPhone hardware development. He also moved to OpenAI in January 2026 but failed to return the company-issued MacBook laptop upon leaving. Crucially, he discovered a previously unnoticed authentication vulnerability in Apple's enterprise system. Using the unreturned computer, he continued accessing Apple's cloud storage, downloading dozens of confidential files, including technical presentations and specifications for unreleased products. Reports indicate he even messaged a former colleague still at Apple, boasting, "Haha, I found I can still access the network storage, hilarious." Apple also accused OpenAI of coaching employees to bypass exit security checks. For example, when HR pushes for a quick departure, they are advised to stall and not sign any resignation documents, thereby avoiding the immediate security clearance trigger. Because Apple's internal file synchronization mechanism continues to push updates to the personal iCloud accounts of former employees, some people have retained confidential Apple documents in their iCloud accounts years after leaving. This irony stems from Apple's internal collaboration relying on Pages, Keynote, iMessage, and iCloud. New employees receive 2TB of iCloud space, with a suggestion to merge it with their personal iCloud plan. However, since an iPhone can only log into one primary Apple ID at a time, the vast majority of employees simply use their personal Apple accounts to log into the company's iCloud, mixing work and personal data. Apple was not entirely unprepared; it created a company-managed "Apple Work" folder that would automatically disappear when an employee's badge was returned. The problem is that people inevitably save files in different locations during work, such as collaborative Pages documents or shared Keynote files, and they might still save them to their personal iCloud directory for convenience. If not placed in the "Apple Work" folder, these files are not automatically deleted upon departure. These management gaps have provided a breeding ground for information leaks during talent mobility. But the core question remains: why are so many Apple employees willing to leave? The answer lies in culture. For hardware engineers, working at Apple means iterating on the next iPhone, with each change requiring layers of approval, often resulting in a product that looks similar to last year's. In contrast, at OpenAI, they get to build something entirely new. What hardware in the AI era should look like, every OpenAI employee can participate in defining. Coupled with OpenAI's compensation packages and equity incentives, it proves highly attractive to mid-level engineers who haven't yet reached financial independence. Reports suggest that seeing employees leave in droves, Apple has started issuing retention bonuses, up to $400,000 in restricted stock, to core employees targeted by OpenAI.
OpenAI's Ambition: Rebuilding an Apple
OpenAI's hardware head, Richard Ho, previously stated in a speech at Stanford, "Phones are not designed for agents." His core judgment is that today's smartphones operate on "app-based" and "session-based" interaction logic. Users must actively seek out functions, like opening apps or a web browser, but the phone does not proactively come to you. Richard believes the logic of an AI agent is completely different. It should be always-on, understanding your context, and proactively handling tasks for you without waiting for you to open it. "If AI is the future, then the hardware designed to carry AI should not look like it does today," Richard said. OpenAI needs its own hardware, one designed from the ground up for AI agents. Based on leaked information, OpenAI's hardware roadmap is already quite clear: a speaker, a phone, followed by glasses, a lamp, and possibly earbuds. The speaker, as the earliest launch product, is expected to debut in early 2027. The mass production of the phone has been moved up to the first half of 2027. OpenAI's speaker is not a product like the HomePod or Amazon Echo. According to descriptions from the supply chain, it can be charged, moved between rooms, has a camera for understanding its surroundings, and includes moving mechanical parts to make it seem "alive." OpenAI positions it as an AI companion that lives in your home, like a person. Powered by ChatGPT, its conversation part uses GPT-Live, has its own personality, and becomes more personalized and proactive over time. It has no screen, no apps, so there is no "opening" action. Although it sounds a bit eerie, OpenAI's design philosophy is for the speaker to constantly listen, watch, and respond to commands. The second product is the phone. Analyst Ming-Chi Kuo first leaked it in April, describing it as an "AI agent phone," built around a continuous, context-aware interface. Kuo's view is that the phone is the only device capable of capturing a user's complete real-time status, including their location, activity, communication, and context. This is the most important context input for real-time agent reasoning. To deliver a complete agent service, one must control both the operating system and the hardware. On the same day, Sam Altman posted, "It's time to rethink how the operating system and user interface should be designed." More specifications for this phone have emerged. For example, its processor uses the same manufacturing process as Apple's next-generation A20 chip. The manufacturer is Luxshare Precision, and the camera module supplier is Sunny Optical. Interestingly, the most prominent feature in the entire spec sheet is its ISP with an enhanced HDR pipeline. Its purpose is to improve the perception of the real world through the camera. It is also accompanied by two dedicated AI processors, one for visual processing and one for language processing. Another point is clear from the supply chain: OpenAI has fully adopted Apple's approach. Luxshare is one of Apple's most important manufacturers, and the entire production line was facilitated by the aforementioned io Products. OpenAI is not just learning from Apple in product design; it is fundamentally trying to replicate Apple. The difference is only in the final device, where the default entry point has been swapped from the App Store to the Agent. Even at the chip level, OpenAI is copying Apple's path. In June 2026, OpenAI and Broadcom jointly released their first custom AI inference chip, Jalapeño, designed specifically for large language model inference, completed from architecture design to manufacturing tape-out in just nine months. This chip is not a modified general-purpose GPU; it is deeply customized for the core functions, memory movement, networking, and service models of large model inference. Apple built a hardware differentiation barrier with its self-developed A-series and M-series chips, and OpenAI clearly intends to follow the same path. Kuo predicts that if development goes smoothly, total shipments of OpenAI's AI agent phone for 2027 and 2028 will be around 30 million units, or about 10 million annually, placing it in the tier of a second-tier brand in the smartphone industry. Therefore, in the early stages, OpenAI cannot shake the foundations of iOS and Android. However, OpenAI's original goal might not have been market share. Richard has stated that for an agent to access the user's current state, it needs full permissions at the operating system level. This means OpenAI wants a phone where it calls the shots. In terms of sales volume, 10 million units is indeed a bit small for OpenAI, let alone Apple or Samsung. But if sales can truly reach 10 million, it would prove that OpenAI's agent phone theory is viable.
Apple's Counterattack
Faced with OpenAI's relentless pressure, Apple is not about to sit idly by. The lawsuit is the most direct counterattack, but it is only the surface. Apple's real trump card lies in its products. To completely sever ties with OpenAI, Apple decisively dropped ChatGPT for its AI models and turned to Google. In January 2026, Apple and Google signed a multi-year agreement, paying approximately $1 billion annually, to use Google's Gemini model to power the new Siri and Apple Intelligence. Furthermore, Apple is more concerned about OpenAI than Google. Google already has its own phone, but OpenAI does not. And it is not making just any phone; it is creating an AI agent device that could redefine human-computer interaction. For Apple, Google is far less dangerous than OpenAI. The threat from Google is known, but the threat from OpenAI is unknown. So Apple would rather pay Google $1 billion a year than continue a deep partnership with OpenAI. It cannot allow a competitor sharpening its tools to enter the hardware space to use its platform to accumulate users and refine its products. Dropping OpenAI is just the first step. Ultimately, Apple needs its own AI capabilities. This is the strategic significance of Apple Intelligence. Apple is taking a different path from both OpenAI and Google. Apple's approach prioritizes on-device processing, cloud collaboration, and privacy. Specifically, Apple's Foundation Models (AFM) use a layered architecture. The lightweight AFM Core has only 3 billion parameters, runs entirely on the device, supports text understanding and simple image recognition, consumes less than 1 watt of power, and has a response latency of under 200 milliseconds. Daily interactions like message summaries, email drafts, and photo searches are handled locally, with data never leaving the device. More complex tasks are routed to the "private cloud." This is Apple's own cloud model, but it uses encrypted channels, and Apple guarantees that user data will not be used for training. The underlying logic of this strategy is to amplify Apple's own strengths. In the large model field, Apple knows it cannot catch up to OpenAI and Google, nor does it need to. Apple's strengths lie in hardware, on-device processing, its ecosystem, and a user base of 2 billion active devices. Its goal is to deeply embed AI capabilities into the hardware layer, using on-device computing and privacy security to build a differentiated moat. On the hardware front, Apple's roadmap actually looks quite similar to OpenAI's: smart glasses, AirPods with cameras, an AI pendant, and a home hub with a screen and camera. Both Apple and OpenAI aim to use hardware to allow the agent to continuously see the world you are in. The design philosophy is also similar: a screen is optional, but a camera is required. In April 2026, Apple officially announced that Senior Vice President of Hardware Engineering, John Ternus, will succeed Tim Cook as CEO on September 1st. Ternus has spent 25 years at Apple, rising from a junior engineer to head of hardware, leading the Mac's transition from Intel chips to the self-developed M-series. He has also been responsible for the hardware engineering of almost all core product lines, including the iPhone, iPad, and AirPods. Cook came from a supply chain background, while Ternus comes from hardware engineering. Two consecutive CEOs with hardware backgrounds is no coincidence. Apple's foundation will always be hardware. Software and services are extensions of hardware, the ecosystem is a moat for hardware, and AI is a new capability for hardware. No matter how AI develops, Apple will ultimately make money by selling devices. So choosing someone who understands hardware to lead is the best choice. So, what are Apple's chances of success in this counterattack? In the short term, Apple's advantages are overwhelming. With 2 billion active devices, a complete ecosystem, strong supply chain control, and unparalleled brand premium, these are things OpenAI cannot catch up to in the short term. In the long term, it is uncertain. This is not a battle for existing market share, but an incremental race for the "agent phone." If AI agents truly redefine human-computer interaction, Apple's current advantages could become a burden. Just as Nokia's advantages in the feature phone era became a drag in the smartphone era. This is why OpenAI is starting with a screenless speaker. It is betting that users will accept a new way of interaction, and that AI agents will become a more natural entry point than apps. If it bets correctly, all the rules of today's smartphone industry will have to be rewritten. Hardware is Apple's lifeline. If the hardware entry point of the AI era is defined by someone else, Apple will no longer be Apple. So this war is essentially two wars layered on top of each other. On the surface, it is a legal battle over trade secrets. Deep down, it is a struggle for dominance in the hardware era of AI. Four hundred Apple employees are rebuilding an Apple inside OpenAI, and what Apple must do is defend its empire while not falling behind on the new track.
Comments