GlobalFoundries says demand for optical modules in China is "strong," according to a report by the South China Morning Post, offering one of the clearest signals yet that the AI infrastructure build-out is pulling China's optical transceiver supply chain firmly into the foundry's growth story.

The characterization matters because of who is making it. GlobalFoundries does not make optical modules itself. It fabricates the chips — silicon photonics components, drivers, transimpedance amplifiers, controllers and power management devices — that sit inside the transceivers moving data between AI servers in data centers. When a merchant foundry of GlobalFoundries' scale describes demand from Chinese optical module makers as strong, it is describing real wafer pull-through, not a forecast slide.

The demand driver is straightforward: AI. Training and inference clusters move enormous volumes of data between GPUs, and electrical interconnects hit distance and power limits beyond short rack-level runs. Optical modules — the pluggable transceivers that convert electrical signals to light and back — have become a structural bottleneck in AI data center construction. The bigger the clusters, the more modules each accelerator requires, and the more specialized silicon each module needs.

China's role in that supply chain is substantial. Chinese firms are among the world's highest-volume producers of optical transceivers, and the AI boom has turned their order books into a leading indicator for the broader photonics component ecosystem — including the foundries that supply them.

GlobalFoundries has invested specifically in this segment. The company operates dedicated silicon photonics process technology at its fabs, a platform aimed squarely at the co-packaged and pluggable optical module market, where integrating photonic functions onto silicon reduces cost and power compared with discrete optical assemblies. A strong demand signal from Chinese module makers translates directly into utilization for those process lines.

The geographic dimension is hard to ignore. Washington's export controls restrict the most advanced AI accelerators from reaching Chinese customers, yet the optical layer that connects compute clusters remains an active commercial arena. A U.S.-headquartered foundry reporting robust demand from Chinese optical module producers illustrates where trade policy has left the lanes open: the interconnect silicon that moves data, rather than the processors that generate it. For GlobalFoundries, which does not compete at the leading edge, that lane aligns neatly with its specialty process portfolio.

The timing of the comment also says something about mix. AI-related demand has been the standout growth category across the semiconductor industry for two years, but the benefit has concentrated heavily in leading-edge logic and high-bandwidth memory. Specialty foundries have needed their own AI exposure. Silicon photonics and optical module silicon are exactly that — AI-adjacent demand that flows through mature and specialty nodes rather than 3nm-class capacity.

For China's module industry, the foundry's assessment adds a supply-side confirmation to what shipment volumes have suggested: the AI boom is not a Western-only phenomenon in the optical layer. Chinese data center construction, domestic AI model development and export business all feed the same module demand curve.

The open question is durability. Optical module demand tracks AI capital expenditure cycles closely, and any pause in hyperscaler or Chinese cloud spending would flow through to transceiver orders — and from there to foundry wafer demand — with a lag measured in quarters, not years. For now, GlobalFoundries' read of the market is unambiguous: Chinese optical module customers are pulling hard, and the foundry expects that pull to shape its photonics business as data center interconnect requirements scale with each new generation of AI clusters.