UBS China AI Industry Chain Research: Optical Interconnect and SiC Heating Up, Shenghong Bets on Rubin PCB

TL;DR

• After UBS surveyed five Chinese tech hardware companies, AI infrastructure demand remains a common thread.
• Shengyi Technology plans fixed asset investment not exceeding 18 billion yuan in 2026, with the Rubin platform being the PCB growth focal point.
• Several key figures come from management discussions; realization depends on mass production, certification, and external constraints.

In a research note released on June 26, UBS provided a more detailed picture of the spreading demand within China's AI hardware chain: demand for AI infrastructure remains robust, and the pull factors now extend beyond GPUs to include PCBs, optical interconnects, liquid cooling, SiC substrates, and optical communication materials.

The key takeaway from this material isn't order changes at a single chip company, but several sectors simultaneously presenting more aggressive timelines and capacity targets. Biren Technology has mapped out its GPU roadmap through 2028, Shengyi Technology plans fixed asset investment not exceeding 18 billion yuan in 2026, and SICC management stated that orders for SiC are saturated, with selective price increases for rush orders from smaller clients.

These insights are primarily derived from management discussions and company guidance during UBS' research, not formal earnings forecasts. They serve more as a thermometer for AI hardware chain demand, reflecting industry heat but not directly equating to realized revenue and profit.

Domestic GPU Roadmap Reaches 2028, Shengyi Bets on Rubin PCB

The timeline provided by Biren Technology is one of the most direct signals for domestic AI computing power in the research material.

The company stated to Caijing that the BR20X is expected to commercialize in 2026, while the BR30X/BR31X is expected to commercialize in 2028. Management discussions within the UBS material further mentioned that the BR20X series GPUs are planned for launch in the second half of 2026, offering upgrades in computing power, memory capacity, and interconnect bandwidth compared to the previous generation.

Catching up in domestic GPUs doesn't just rely on single-chip computing power; it also requires advancements in memory, interconnects, packaging, and the software ecosystem to keep pace. For AI training and inference clusters, bottlenecks often occur between chips, servers, and at the system scheduling level, not just in theoretical computing power.

The limitations are also clear. Advanced process nodes, advanced packaging, customer mass production schedules, and software ecosystem adaptation will all impact the subsequent deployment of the BR20X and BR30X. The external export control environment continues to affect the pace of the domestic AI hardware supply chain.

Closer to near-term orders and capital expenditure is Shengyi Technology's PCB guidance.

Management stated that overall AI PCB raw material costs are stable, cost pass-through for the new generation of products is relatively smooth, and they expect the gross margin for fiscal year 2026 to be roughly flat with 2025. The AI revenue share is expected to increase from below 50% to 60%-70%.

The Rubin platform is the most anticipated growth driver. The research material states that Shengyi Technology maintains a majority share in NVIDIA's Rubin compute tray HDI, consistent with the GB300 platform situation, targeting an overall customer share of about 50%. Since this claim hasn't been verified in public announcements, it's better understood as a management target and sell-side research insight. If realized, the iteration of NVIDIA's new platforms will continue to stimulate China's high-end PCB supply chain.

The expansion figures also require more precise context. Public information indicates that Shengyi Technology's total investment for 2026 will not exceed 20 billion yuan, with fixed asset investment not exceeding 18 billion yuan, mainly directed towards Huizhou Factories 10-13. This scale shows that AI server PCBs have been incorporated by management into the core capacity direction for the coming years.

The material solution for the orthogonal backplane required by the Rubin Ultra has not been fully finalized. The company stated there are no delays related to this, but various materials and supplier combinations are still being evaluated, including Q-glass and PTFE. The order direction is relatively clear, but the process and material roadmap are still under selection.

Optical Interconnects and Liquid Cooling Become Essential AI Cluster Support

As AI clusters continue to scale, the importance of interconnects and cooling rises. Lightmatter and Lens Technology provide two supporting pieces of evidence.

The UBS research material states that Lightmatter holds approximately 88% market share in independent scale-up optical interconnect solutions, focusing on LPO, NPO, and similar solutions for large-scale GPU and ASIC clusters. Its optical computing processors use a 3D TSV vertical stacking method for electronic ICs and silicon photonics, offloading some computational tasks to the silicon photonics layer, thereby reducing latency and lessening reliance on advanced process nodes.

The key point of such solutions isn't "optical computing replacing GPUs," but rather that the larger the AI cluster, the more prominent the interconnect bottleneck becomes. If LPO, NPO, and silicon photonic packaging can achieve stable mass production, they could offer incremental improvements in bandwidth, power consumption, and latency for large-scale clusters.

Lens Technology's AI infrastructure layout comes from optical communications and liquid cooling. Public reports show that Lens Electronics strategically acquired a controlling stake in Shenzhen Tongsheng Optoelectronics Co., Ltd. in June 2026, entering the field of hollow-core fiber optic cables. Management targets from the UBS research indicate that the optical communication business is expected to make a more meaningful contribution to the income statement starting in 2027, with a revenue potential of approximately 10 billion yuan.

On the liquid cooling front, the management target is to become the largest shareholder of Yuans Tech, with liquid cooling business revenue potentially reaching several billion yuan in fiscal year 2026. The unit value for cold plates, manifolds, and UOD is approximately $40,000-$50,000. These figures lack verification from public announcements and should still be understood at the target and guidance level.

Historical stock price and target price trend chart for Lens Technology, along with historical rating and target price adjustment records.

SiC Orders Saturated, 8-Inch Shipment Target Raised to 50%

In the power device chain, signals from SICC are closer to "supply-demand tightness is already affecting some prices."

The company's press release, citing data from Fuji Keizai, stated that in 2025 it held the global number one share for SiC substrates, both 6-inch and 8-inch, with an 8-inch share of 51.3%. Management discussions within the UBS research material further indicated that current demand is strong and orders are saturated. Prices were generally stable in Q1 2026, but selective price increases have been implemented for rush orders from smaller clients.

Management's target is for 50% of shipments in 2026 to come from higher-margin 8-inch SiC substrates, driving gross margin expansion. If this target is achieved, it will directly impact the company's product mix and profitability.

SiC was previously primarily driven by electric vehicles, photovoltaics, and energy storage. Now, high-power AI servers and data center power systems are emerging as new sources of demand. HVDC, SST, AI glasses, and advanced packaging are also listed as potential drivers.

The key point here isn't that SiC demand has suddenly shifted from automotive to AI, but that AI data centers continue to expand the demand boundary for power semiconductors. Saturated orders and selective price increases suggest that supply tightness is starting to impact prices in some customer and rush order scenarios.

Whether prices can continue to rise depends on the ramp-up of 8-inch capacity, customer validation, and downstream demand pace. If capacity release outpaces order growth, pricing flexibility may be weakened. If data center demand materializes faster than expected, the potential for gross margin improvement at SICC will become clearer.

Order Direction Clear, Challenges Remain in Mass Production and Constraints

This set of research insights collectively points to one change: AI hardware demand is spreading outward along GPUs, PCBs, optical interconnects, liquid cooling, and SiC, with the Chinese supply chain providing more aggressive capacity and product roadmaps in multiple segments.

However, it's inappropriate to frame this as a "comprehensive breakthrough in domestic AI hardware." Biren Technology's GPU roadmap remains constrained by process nodes and ecosystem limitations. Shengyi Technology's Rubin-related share still awaits platform mass production and finalization of material solutions. Lightmatter's optical interconnects and optical computing require large-scale customer validation. The contribution of Lens Technology's optical communications and liquid cooling primarily depends on the 2026-2027 timeframe. Whether SICC's selective price increases continue depends on 8-inch capacity and data center demand realization.

The broader external constraints remain in areas related to advanced process nodes, advanced packaging, high-end GPUs, and cloud provider capital expenditure. The closer the AI hardware chain is to these positions, the more susceptible it is to policy and supply chain constraints.

The most valuable aspect of these figures is that they show AI infrastructure orders are diffusing into a wider array of hardware segments. The boundaries that need to be maintained are equally clear: most figures are still management targets and research insights, and true realization depends on customer mass production, material finalization, and the external environment.