An Apple supplier has rejected a request to increase orders for the upcoming iPhone 18 Pro, pivoting resources to the high-demand market for Co-Packaged Optics (CPO) that powers AI data centers. The move by optical lens maker Largan Precision shows the growing strategic importance of the AI hardware supply chain over even Apple’s high-volume manufacturing ecosystem.
According to a report from Western Securities, CPO technology is seeing accelerated commercialization driven by AI leaders like Nvidia and Broadcom. CPO offers significant advantages, cutting power consumption by over 50% compared to traditional modules. For instance, Nvidia's latest CPO architecture claims energy savings as high as 73 percent.
Largan’s decision, a rare refusal in Apple's tightly controlled supply chain, suggests the financial incentive for producing specialized AI components now outweighs the benefits of securing larger orders for consumer electronics. This could trigger a wider strategic re-evaluation among suppliers, channeling investment and capacity toward the AI infrastructure sector. The Wind CPO index has already rallied over 26% since April 1, with domestic leader Zhongji Innolight surging nearly 30 percent in the same period.
CPO’s Role in Next-Gen Data Centers
Co-Packaged Optics represents a fundamental shift in data center architecture. By integrating optical transceivers directly with switch and processor chips on the same substrate, CPO drastically shortens electrical pathways. This design reduces data transmission latency and heat, improving signal integrity to a mean-time-between-failures (MTBF) of 2.6 million hours—far exceeding the 500,000 to 1 million hours of current pluggable modules.
While Largan’s refusal may not disrupt Apple’s iPhone 18 Pro production, as it is reportedly a secondary supplier to Sunny Optical for the variable aperture lens, the move is a strong indicator of a market shift. As AI models become more complex, the demand for bandwidth and energy efficiency in data centers is exploding. CPO is viewed as a key enabling technology to manage these demands, breaking through the expansion limits of traditional copper wiring and allowing for massive GPU clusters, such as the 131,072 GPUs supported by Nvidia's Spectrum 6800 platform.
This article is for informational purposes only and does not constitute investment advice.