“Largan's CPO Reality Check: 2 Million Units a Month… but Where Do 40,000 Workers Come From?”

The world's leading smartphone lens maker, Largan Precision, made its debut at Computex—not for mobile camera lenses, but to showcase one of the hottest topics in AI data centers: Co-Packaged Optics (CPO).

The appearance was low-key. Largan secured a last-minute mini booth to present its early technological progress in optical communication components.

What truly caught attention, however, was not the exhibit itself, but a number revealed by an executive on-site: If current semi-automated production methods are used, reaching a monthly output of 2 million optical communication components could require as many as 40,000 workers.

This is almost an impossible equation.

At present, Largan employs only around 8,000 people in total. In other words, this company—renowned for precision manufacturing and for pushing yield and efficiency to the extreme—has barely stepped into the CPO arena, only to encounter a challenge even more daunting than technology itself: Demand for AI optical communication is massive, but how can it actually be produced at scale?

Stock Prices Move First, Reality Follows

Since the beginning of the year, the CPO theme has swept through Taiwan's stock market. With TSMC announcing its silicon photonics mass production timeline for 2026, and NVIDIA and Broadcom positioning themselves aggressively, the entire optical communication sector has surged.

Amid rumors of Largan's active CPO expansion, its stock price has climbed steadily, hitting an intraday high of NT$4,225 on the 3rd—its highest level in nearly five years—while related optical stocks also rallied.

Previously, at an earnings call, Largan CEO Lin En-ping—nicknamed the “honest tree” for his candid remarks—stated bluntly that CPO mass production is still “a long way off,” requiring at least one to two more years. The stock promptly dropped over 7% in a single day, exposing the gap between market expectations and industrial reality in the most direct way possible.

A Late Start Due to Underestimating Complexity

Largan revealed that it initially misjudged CPO as a low-precision product similar to traditional fiber optics. It wasn't until the second quarter of last year, after visiting companies such as TSMC, that it realized the technical barriers were far higher than expected, prompting a rapid reallocation of R&D resources.

But the late start was compounded by delays in equipment procurement. High-precision machining equipment required for CPO—such as V-groove machines ordered from Japanese suppliers—now carries lead times of up to 12 months, with German equipment taking even longer. Currently, Largan has acquired only one such machine.

Even if samples are delivered to TSMC in the future, customer qualification alone could take at least another six months. Largan estimates that equipment will arrive next year, with mass production unlikely before the following year.

The Semi-Automation Bottleneck: Like Operating a Da Vinci Surgical Robot

The more critical issue lies in precision. The tolerance for traditional fiber optic assembly is about 2–3 microns, whereas CPO requires tolerances below 0.5 microns, with some processes demanding as tight as 0.3 microns—roughly one three-hundredth the diameter of a human hair.

Fully automated equipment capable of achieving such precision remains largely unavailable worldwide.

As a result, the industry relies on semi-automation: operators work alongside machines, manually adjusting and aligning optical fibers one by one, then fixing them with adhesive. The process is highly dependent on human intervention, functioning much like a surgeon operating a Da Vinci Surgical System—the machine cannot operate independently, and the human cannot step away.

Under this model, capacity expansion is constrained not by technology or capital, but by the ability to recruit labor. This is why the estimate of 40,000 workers for a monthly output of 2 million units emerges.

Worse still, even with such massive manpower, yields under sub-0.5-micron precision requirements hover at only around 30%.

The Real Race: Who Can Achieve True Automation First

Industry insiders point out that the entire sector is now racing toward one goal: Who can develop truly fully automated precision manufacturing the fastest?

Largan possesses deep expertise in optical R&D—its greatest strength in entering the CPO field. However, its late start introduces significant uncertainty into the outcome of this race.

Source of Information: Business Next