Former Microsoft Executive on Apple’s Hidden China Problem
Summary
Apple’s China dependence is a capability moat that became a strategic vulnerability, not a cheap-labor trade. Torenberg relays the book’s estimate that Apple invests $55 billion annually in China, while Sinofsky argues the decisive capability came from Apple people swarming manufacturing lines and transferring knowledge by osmosis. The point of no return arrived roughly “two years into the iPhone,” when no other location could match the required skills or scale.
Apple’s muted AI showing looks more like a retreat from premature promises than an abandonment of the field. Sinofsky calls the earlier pre-announcement uncharacteristic and expects Apple to revert to being “not the first mover company” but “the first integrator company.” Torenberg says Apple still needs a model tuned for its unique hardware, edge operation, and privacy constraints; Sinofsky identifies Siri as a conspicuous perception hole.
AI is unlikely to become winner-take-all, making second- and third-place platforms potentially generational investments. Sinofsky’s arc runs from IBM’s 100% mainframe position and Microsoft’s 95% PC share toward phones split roughly 80/20 and cloud potentially settling at 40/40/20. Torenberg says a16z’s mistake was not investing even more aggressively because it overemphasized backing only the leader; privacy, security, and inference cost should also drive a large edge-AI wave.
Meta’s Scale transaction matters because AI needs several scaled competitors, not because one model has already won. Torenberg asks whether Meta’s 49% Scale AI stake could be viewed as something like a $15 billion acquihire of Alex Wang and top talent; Sinofsky’s verdict is that it is at least “great for AI,” while reserving judgment on how great it is for Meta. His larger fear is one government-sanctioned approach, one player claiming it should be the only one, or geographically isolated ecosystems resembling Japan’s advanced but domestically trapped i-mode market.
COVID converted supply-chain efficiency into national-security exposure by exposing “single points of failure all over the place.” A closed city, factory, or transit route could halt global output, while the United States discovered that drones—and even their components—came from the only maker, treated as “our enemy, so to speak.” Apple’s India buildout addresses its China concentration, but it does not necessarily satisfy a policy goal centered on restoring American production and employment.
Sinofsky rejects the claim that domestic production mechanically means a $5,000 iPhone. A hand-built iPhone might cost $100,000; its current roughly $1,000–$1,500 price reflects manufacturing innovation, so changed constraints should produce another step-function through robotics, packaging, fewer components, and new processes. “Innovation isn’t invention”: the problem is one engineers can solve against new constraints.
Apple can keep gaining device share, but dispersed manufacturing know-how and unresolved IP rules change its old defenses. Sinofsky expects today’s phone form factor to persist longer than enthusiasts assume, while stable use cases and comparable pricing have already taken Macs toward half of some U.S. laptop segments. The larger uncertainty is how to navigate intellectual property: treating China’s IP record as disqualifying and treating all knowledge as free for AI training are, in Sinofsky’s view, both unrealistic positions.
Deep dive
1. Apple is pausing AI theater while it finishes the platform
Sinofsky treats the Liquid Glass backlash as predictable for a redesign of a product used at enormous scale, while noting that his own prior redesigns were not used by a billion people. People immediately ask, “What problem does it solve?” Yet WWDC was a developer event and the software was unfinished; Apple is good at fixing blur, noise, legibility problems, and bugs. His position is deliberately provisional: “I’m reserving judgment and not joining in the hysteria on either side.”
Giving the iPad “Windows with a lowercase W” overturns years of toaster-refrigerator separation between tablets and PCs. Sinofsky worries it may be late because users have established habits, but stresses that Apple sells more iPads than the U.S. laptop run rate. Since iPads and Macs share the same underlying hardware, software, apps, and experience become the real boundaries.
WWDC’s major omission was AI, beginning with Craig saying, in Apple’s style, that features had “took more time than we planned.” Sinofsky thinks Apple’s earlier fanfare around unfinished Apple Intelligence features was out of character. It will likely wait for the technology to mature and return as a “first integrator,” although saying little about Siri leaves a perception hole.
2. AI’s market can support several giants—and migrate toward the edge
Sinofsky sees platform concentration weakening across eras: IBM had 100% in mainframes, Microsoft 95% in PCs, servers fell toward a roughly 50/50 split before becoming predominantly Linux, phones split roughly 80/20 globally depending on country, and cloud may become 40/40/20. One AI winner is therefore “almost inconceivable.” Japan’s i-mode shows the opposite risk: sophisticated technology can remain geographically trapped.
To prevent one company, one sanctioned approach, or regional fragmentation from dominating, Sinofsky argues for “a lot of players operating at huge scale.” He counts OpenAI and Anthropic in startups, Meta in open source, and Microsoft and Google as other major participants; Meta’s Scale move is, at minimum, “great for AI,” whatever its ultimate value to Meta. “There’s nothing better than more and open for AI right now.”
Torenberg’s investing lesson is unusually candid: a16z backed the category aggressively but regrets not being more aggressive because it dismissed second-place companies. This market may be large enough to produce several generational businesses. He also expects privacy, security, and cost to push a large wave of AI toward the edge rather than imposing variable cloud cost on “the entire world of software.”
Incumbents have three broad plays: Microsoft’s partnership dance with OpenAI, Amazon’s retail-like promise to carry “all the cereals, not just Froot Loops,” or vertically integrated paths such as Gemini, OpenAI, and Anthropic. Sinofsky says Apple needs to choose one; Torenberg says its real task is tuning a model for its hardware, edge deployment, and privacy constraints. Microsoft’s advantage is different: “You never have to be best. You just have to be included” in the enterprise bundle. Sinofsky also notes that Microsoft has been researching AI since 1993 and would not be surprised if it has a first-party effort, though he says he does not know what is happening.
3. Ambitious hardware pushed Apple into Asian manufacturing
Apple began with extreme internal control, including carefully designed motherboards and its own factory. When U.S. manufacturing constrained the PowerBook, Jean-Louis Gassée pushed a joint design with Sony in Japan—culturally brazen for Apple. The model was complicated and not the bestseller, but it demonstrated that an external partner could expand what Apple was capable of building.
The 1999 iMac G3 then made the new model visible: its translucent, gumdrop-shaped, toolless and fanless enclosure was not merely stylish; Sinofsky calls it “impossible” by the standards of gray stamped-metal computers. Apple achieved it through a China-based manufacturing model, with Apple people swarming the line and sending samples back and forth.
The iPod turned that process global: a disk drive invented in Japan and manufactured in Thailand joined the click wheel and other components in China. Competitors produced “big honking” players, while Apple delivered a solid deck of cards, glued together with almost no tolerances. This became the template: “designed in Cupertino, manufactured in China.”
At the time, that model was celebrated. Sinofsky places it beside the 1999 Seattle WTO conflict over admitting China and the management doctrine to “stick to your knitting”: outsource everything except the company’s core competency. Both U.S. parties were divided between free trade, labor concerns, and anti-communism, but business culture saw a new industrial era rather than a strategic trap.
4. Commodity PCs created the supplier ladder Apple later climbed
Dell, Compaq, and IBM relied on standardized Intel processors, Seagate drives, cables, and cases, leaving little product differentiation. Parts came from different countries and were assembled in the United States, while the companies competed through price, place, and promotion. As cases and then larger portions of production moved to China, assembly plants began serving multiple companies and became original design manufacturers, or ODMs, that could present ready-made computers and apply whichever customer logo won the order.
PC companies wanted “more gray boxes, more interchangeable parts, lower cost,” even saying they did not need engineers as volumes rose. ODMs shared the same labor pool and factories, so they could not win merely through cheaper workers. China’s system moved hundreds of millions of rural workers temporarily into factories and sent cash back to rural areas, helping lift a billion people out of poverty; the ODMs therefore moved up the stack into prototyping, engineering, and product design. Distributed knowledge then made possible 50 different music players and many other branded electronics products.
Living in China in 2004 to combat Windows and Office piracy, Sinofsky toured enormous ODM campuses. One owner proudly showed him the legal Windows sticker station, then privately declared, “The Windows people are losing,” while unveiling sophisticated prototypes Dell and Compaq would not buy. Nearby, Apple’s engineer-heavy operation was helping develop the capabilities that surfaced in MacBook Air production.
Microsoft created Surface partly because PC makers refused aluminum computers and all-in-ones. Its team built a metallurgy plant for injection-molded magnesium alloy for the original NVIDIA-based Surface; OEMs still treated the designs as a niche. Only later did Intel use pricing investments to induce Windows manufacturers to build Ultrabooks, adding more technical knowledge to the ecosystem Apple had helped create.
5. China’s accumulated skill became Apple’s dependency
Tim Cook’s crucial distinction is that China is not primarily about cheap manufacturing but “the skills they have.” Sinofsky’s examples include operating 500 aluminum prototyping machines simultaneously, controlling defects and waste, and mastering pressure-sensitive adhesives to hold components together in ultra-thin products. Apple did not intend to transfer a complete capability base, but “osmosis happens.”
The point of no return came about two years into the iPhone, when scale eliminated plausible alternatives; even low-volume Surface production could not have begun elsewhere. The annual $50 billion-plus investment was not the original cause. The cause was an unexpected hybrid of totalitarian controls and intensely competitive entrepreneurship—after experts predicted China would remain a “third-world dictatorship forever.”
Foreign automakers illustrate the leverage that followed: mandated joint ventures exposed IP, trapped capital, and ultimately left Volkswagen, Ford, and Mercedes in a painful position. Tesla avoided that formal structure, but Sinofsky says softer tools remain—from the Communist Party favoring a Huawei model to domestic 5G rules touching Qualcomm patents. His caveat is that U.S. defense procurement and American-only municipal fleets can also look like protectionism from outside.
6. COVID exposed fragility, but new constraints can drive reinvention
COVID was the global wake-up call. A system optimized for price and specialization proved fragile because “there are single points of failure all over the place”: shut a city, factory, or transit network, and production or components stop moving. Chips assembled in one place and packaging concentrated among a small set of skilled producers made those dependencies strategically consequential rather than merely inconvenient.
Defense had already localized guns and rockets—the Italian maker Beretta built a U.S. factory to win its military contract. Drones revealed the reverse: the United States made neither the critical product nor many of its parts, while the only maker was “our enemy, so to speak.” Sinofsky considers reducing that exposure unavoidable, even though China remains far more open than the Soviet Union was.
Apple’s expansion in India addresses corporate concentration risk but not necessarily an administration demanding U.S. jobs. Sinofsky connects that tension to NAFTA: free trade with Canada and Mexico looked obvious, yet crossing a border substantially changed production economics and employment. The reset therefore extends beyond Apple to cars assembled in Detroit or Canada that still rely on imported in-dash electronics and car computers.
Torenberg asks whether diversification means radically higher device prices; Sinofsky calls the $5,000-iPhone argument “incredibly silly.” Scale and process innovation made a potentially $100,000 hand-built object cost roughly $1,000–$1,500. Apple can now pursue giant factories of robots, new packaging, fewer assembly steps, and the manufacturing of components such as the lenses in Vision Pro: “Innovation isn’t invention”—it is engineering against changed constraints.
7. Durable devices face lower barriers and years of IP uncertainty
Sinofsky does not know which device comes next, but expects the phone form factor to last “much longer than even the most optimistic people think.” Macs approached half of some U.S. laptop segments after the use case stabilized around browsers and Microsoft Office. At a $1,000 Mac versus a $700 alternative, price will still decide some purchases; another segment will choose the phone that is free with a carrier plan.
Phones should become more competitive as their target stabilizes and manufacturing barriers fall through automation. Microsoft’s old warning still applies: someone can arrive with a better version of everything. Hardware retains more scale protection than software, but less than before; Samsung’s many capable phones exist because manufacturers “have all seen the movie.”
The concluding fault line is intellectual property. Declaring that China’s behavior excludes it from global markets is one extreme, while declaring that AI may freely train on and reuse all knowledge is the other. Sinofsky cites pharmaceuticals and the BYD-Tesla relationship from one direction and unrestricted AI training from the other: “Neither of those are realistic positions.”
Policy could fragment markets before a stable settlement emerges. The EU might legislate sharply in either direction, while Japan may go the other way because of its language challenge; it remains unclear whether a Japanese-only approach would create an English-content loophole. In the U.S., the issue is structured as a litigation problem, so Sinofsky expects years of market uncertainty—just as Apple’s advantage in maintaining the intellectual property of components, manufacturing, and assembly has dispersed.