Apple briefly crossed $4 trillion. Its newest defense against an AI-hardware rival is a lawsuit over what departing engineers may have carried through the exit.

In late 2023, Tang Tan’s planned departure looked like executive turnover. Two and a half years later, Apple’s allegations against OpenAI recast it as the first visible move in an AI-hardware race turning defensive.

The departure was visible before the strategy was

In December 2023, sources said Tan, then Apple’s vice president for iPhone and Watch product design, planned to leave in February 2024. Later that month, reporting added a destination and a purpose: he would join Jony Ive’s LoveFrom to work on AI hardware.

The move was unusually specific: from responsibility for two of Apple’s defining devices to an effort to build AI hardware. What remained unclear was the scale.

Each report looked discrete when it arrived. First an executive left. Then a design organization formed. Then the departing executive became the hardware chief of an AI company. Then the former employer reorganized design and went to court. The sequence did not change direction; it changed meaning as capability accumulated around it.

The scarce component is organizational memory

The early AI race rewarded model capability. The hardware phase added another visible constraint: specialized silicon. But a processor roadmap and a product roadmap are not the same structure. A chip can increase the capacity available to a device; it cannot decide what the device should be, how its parts become one product or which compromises users will accept.

That conversion is the work of product-design organizations, and its output is only partly recorded in formal specifications. The rest is accumulated judgment: the repeated decisions through which teams learn what can be built, what should be rejected and how an idea survives contact with manufacturing and customers. When those people move, a company does not merely lose labor. It risks allowing a rival to shorten the distance between capability and product.

average stock-based compensation per OpenAI employee in 2025, according to Equilar

That figure was roughly seven times Google’s before its IPO and about 34 times the average among other pre-IPO peers. It was not reported as hardware-specific, so it does not establish what any individual designer received. It does establish the economic environment around the hardware contest: OpenAI has built a compensation structure capable of assigning an unusually high price to scarce people.

This creates a reinforcing loop. A well-funded entrant recruits experienced product builders; their arrival makes the hardware effort more credible; that credibility lowers the risk for the next recruit. Each additional hire increases the value of the knowledge already gathered. The incumbent responds with retention, organizational reform and legal boundaries. No participant has to intend the full outcome. The loop produces it.

A stronger chip does not settle the device question

Apple is reportedly already taping out an M7 with major upgrades to its neural processing unit. The M7 and M8 roadmap comes from sources, not a confirmed Apple announcement—a necessary distinction for a company whose unreleased products generate elaborate shadow roadmaps.

Even if the reporting is accurate, the M7 does not contradict the talent conflict. It explains it. Apple can improve the silicon beneath its products while defending the human system that determines what those products become. The processor is proprietary capability embodied in a design file. Product judgment is proprietary capability embodied in people who can leave.

Ternus’s elevation sharpens the same point. Apple named him as its next chief executive in April while identifying the retention of top talent as a challenge amid historic turnover. By June, sources said he was preparing to reshape the industrial design team and reassert its executive influence. The chip roadmap and the organization chart are moving together because the hardware race now requires both.

Apple did not suddenly discover that design matters. The question changed. Apple’s design organization was built to concentrate product judgment inside Apple. Once competitors began assembling AI devices with veterans of that system, the concentration that made execution powerful also made the relevant talent identifiable. A strength became a recruiting map.

The lawsuit is a boundary marker, not a verdict

Apple alleges that iPhone engineer Chang Liu took more than his experience when he left for OpenAI’s nascent hardware division. It also alleges that Tan directed Apple employees interviewing at OpenAI to share Apple secrets. Those are allegations, not adjudicated findings, and they should not be converted into facts merely because they fit the larger pattern.

The filing itself establishes one point before the merits are settled: Apple considers the boundary around product knowledge important enough to defend in court. Litigation can slow the movement of protected know-how, raise the cost of acquiring it and mark a line between portable experience and company property.

That line is difficult precisely because product execution cannot be reduced to a box of documents. An engineer is expected to carry skill from one employer to another; an employer is entitled to protect trade secrets. The conflict begins where those categories overlap, which is also where the most valuable organizational knowledge tends to live.

The abstraction called AI hardware has acquired a very ordinary address: the recruiting interview.

This is the defensive phase. The entrant is assembling the human structure that can make models, capital and chips cohere as a product; the incumbent is trying to prevent that structure from being reproduced elsewhere. Once teams become the target, personnel movement becomes competitive architecture.

Apple can briefly be worth $4 trillion and reportedly tape out an M7 with major NPU upgrades. Yet its hardest boundary now begins at the door, when an engineer’s badge stops working.