UMC Pushes Silicon Photonics Into 12-Inch Production as AI Data Centers Demand Faster Optical Interconnects
The rapid rise of artificial intelligence is putting enormous pressure on hyperscale data centers. As AI models become larger and more complex, the amount of data moving between servers, accelerators, and networking equipment continues to surge. This growing demand is accelerating the shift toward next-generation optical modules, with the industry preparing for 1.6T and 3.2T interconnect speeds.
United Microelectronics Corporation, better known as UMC, is strengthening its position in the silicon photonics market by expanding development around thin-film lithium niobate, or TFLN. The company sees TFLN as a key technology for solving performance bottlenecks that could limit future high-speed data transmission.
Silicon photonics has become increasingly important for AI infrastructure because traditional electrical interconnects face challenges with bandwidth, power consumption, and signal loss at higher speeds. By using light to move data, silicon photonics can help data centers achieve faster communication while improving energy efficiency, an essential factor as AI workloads continue to scale.
UMC’s silicon photonics platform has now advanced into 12-inch wafer production, marking an important step for manufacturing efficiency and scalability. Moving to 12-inch wafers can support higher output and better cost control, both of which are critical as demand for optical networking components increases across cloud computing and AI data center markets.
Thin-film lithium niobate is gaining attention because of its strong electro-optic properties, which can enable high-speed modulation with lower power usage. In simple terms, it can help optical signals switch faster and more efficiently, making it well suited for future 1.6T and 3.2T optical modules.
For hyperscale operators, the need for higher-bandwidth optical links is becoming urgent. AI training clusters depend on extremely fast data movement between GPUs and other accelerators. Any delay or bottleneck in the network can reduce overall system performance. This is why advanced optical interconnect technologies are becoming a major focus for semiconductor manufacturers, cloud providers, and networking companies.
UMC’s investment in silicon photonics and TFLN reflects a broader industry trend: data center performance is no longer driven only by faster processors. The connections between chips, servers, and racks are now just as important. As AI computing expands, optical interconnects are expected to play a central role in keeping data moving at the speeds required by next-generation workloads.
With its move into 12-inch silicon photonics production and its focus on TFLN technology, UMC is positioning itself to support the next wave of high-speed optical networking. As the market transitions toward 1.6T and 3.2T modules, scalable manufacturing and efficient photonic technologies could become essential for overcoming the bandwidth and power challenges facing AI data centers.
