Sources: Apple has asked Foxconn to produce AI servers in Taiwan using Apple silicon, but Foxconn's capacity is limited by demand for Nvidia's GB200 servers
Nikkei Asia :
Context & Ripple Effects
Foxconn had already signaled that GB200 server output would begin at small volume before a later ramp, making its Taiwan server lines a scarce manufacturing input rather than a neutral expansion pool. Its planned GB200 ramp gives the reported Apple request a concrete capacity trade-off.
Subsequent coverage shows how consequential that trade-off became: AI servers overtook Apple-related products in Foxconn’s revenue mix, while Taiwan-to-US AI-server shipments were expected to accelerate. The projected shipment increase situates Apple’s request within a broader pull on the same supply base.
First-order effects
- Apple’s server build plans using its own silicon must compete for Foxconn’s limited Taiwan capacity, potentially constraining how quickly those systems can be produced.
- Foxconn must allocate constrained assembly capacity between Apple’s requested work and demand for Nvidia GB200-based servers, with GB200 demand limiting immediate flexibility.
Second-order effects
- Apple may need to secure manufacturing capacity earlier or diversify production options if Nvidia-linked server orders continue to absorb Foxconn’s available lines.
- For Foxconn, AI-server demand strengthens the commercial importance of server production relative to its traditional Apple-device business, a shift later reflected in its AI-server revenue surpassing Apple-related revenue.
Third-order effects
- If hyperscalers and platform companies increasingly commission proprietary-silicon servers, contract manufacturers will allocate capacity across competing compute ecosystems rather than simply scale one vendor’s hardware.
- The constraint shifts AI competition beyond chip design: access to qualified server assembly capacity can become a product-timing and deployment constraint, especially in Taiwan’s concentrated supply chain.
The trend: AI infrastructure demand is turning contract-server capacity into a strategic bottleneck shared by Nvidia’s ecosystem and companies building custom silicon.