The AI infrastructure boom that enriched Silicon Valley legends is now minting Chinese heavyweights at speed. Google and other hyperscalers are locking in China’s best-in-class optical gear to move exabytes inside giant data centers, pushing leading suppliers to record sales, margins, and global share. With China’s economy projected to top $23.8 trillion by 2030 and policy aimed at scaling cloud and compute, a cohort of listed champions is positioned to ride the capex supercycle. Here are eight China AI-optics and infrastructure stocks at the heart of the buildout, plus what to watch as spending accelerates from $375 billion this year toward the multi-trillion-dollar mark later in the decade.
Zhongji Innolight’s surge captures the moment. The Shenzhen-listed optical transceiver leader has ridden urgent orders from Google and Nvidia, sending its founder’s wealth up 400% and the stock up more than fivefold this year. Nomura recently called the company the global No.1 data center transceiver maker after a standout third quarter, and Macquarie’s head of tech research in Hong Kong says demand remains supply-constrained. The technology here is unglamorous but fundamental: modules the size of a thumb drive that convert electrical signals to light with speed and reliability that copper cannot match, especially across million-square-foot campuses. The payoff shows up in the numbers: nine-month revenue up 44% year-on-year to 25 billion yuan and net income up 90% to 7.1 billion yuan, plus a planned Hong Kong raise of over $3 billion to scale further.
Beijing’s innovation policy emphasizes sovereign compute, data center clustering, and energy efficiency, a playbook built for compounding scale. The Eastern Data, Western Computing initiative is shifting workloads inland where power is cheaper, while State Grid’s unparalleled balance sheet and $546 billion in revenue support backbone upgrades to carry and balance AI-era loads. Financing depth matters too. ICBC, with a market cap north of $300 billion, underwrites capex cycles that few markets can match. Supply-chain resilience is built in: leading optics suppliers have added offshore capacity, including new factories in Thailand, to serve U.S. and global customers while navigating tariffs and export controls. The result is predictable delivery at hyperscale, and that is what wins high-stakes purchase orders.
– Zhongji Innolight (300308.SZ): Global No.1 data center transceiver maker per Nomura, with 9M revenue of 25 billion yuan and net income up 90%. Customers include Google and Nvidia. Planning a Hong Kong listing to raise over $3 billion; Thai plant adds geopolitical resilience. Macquarie expects it to outgrow the industry as 400G and 800G deployments scale.
– Eoptolink (300502.SZ): A leading high-speed optical module vendor ramping 400G and 800G shipments for cloud data centers. Global impact: export mix rising as hyperscalers diversify supply. Positioned to benefit as AI-related capex grows toward the multi-trillion-dollar range by 2030.
– Accelink Technologies (002281.SZ): Core optical components and coherent modules across telecom and data center markets. Milestone: meaningful share in 400G systems; builds on domestic 5G and cloud deployments. Global impact: supplies carriers and OEMs across Asia and Europe, providing long-haul and metro backbone capacity for AI traffic.
– Yangtze Optical Fibre and Cable, YOFC (601869.SS; 6869.HK): One of the world’s largest producers of optical fibre preforms, fibre, and cable. Milestone: scale manufacturing that reduces unit costs for hyperscale builds. Global impact: backbone fiber to connect sprawling campuses and interconnect regions where cloud and AI nodes are proliferating.
– Hengtong Optic-Electric (600487.SS): Major supplier of optical and submarine cable systems. Milestone: extensive international project footprint. Global impact: subsea and terrestrial links that tie together AI clusters across Southeast Asia, the Middle East, and Africa, widening market access for Chinese equipment vendors.
– Inspur Information (000977.SZ): Top-tier AI server maker in China and a top-three x86 server vendor globally by shipments in recent years. Milestone: leadership in GPU-rich racks and liquid-cooled designs. Global impact: accelerates compute availability for domestic LLMs and exports where permitted, reinforcing end-demand for optics inside and between racks.
– GDS Holdings (GDS): Carrier-neutral data center operator with hyperscale campuses in China and expansion in Southeast Asia. Milestone: multi-year pre-commitments from cloud and internet majors. Global impact: regional platform enabling tenants to roll out AI training and inference near users, a structural driver for cross-connects and optical interconnect density.
– China Mobile (0941.HK; 600941.SS): The world’s largest mobile operator by subscribers and a national-scale cloud and network builder. Milestone: sustained high capex to densify 5G and expand nationwide computing hubs. Global impact: its cloud-network fabric underpins inference at the edge and backhauls AI traffic that feeds optical demand end-to-end.
Three durable forces support today’s optics capex: the move from 100G to 400G and 800G links within data centers; the relocation of compute to regions with lower-cost power; and the rise of AI inference at the edge. Macquarie pegs the transceiver market’s annual growth at roughly 70% en route to $40 billion by 2028, syncing with broader AI infrastructure spend forecast to swell from $375 billion this year to the trillions by 2030. Risks exist. Co-packaged optics, led by Broadcom, could alter module architectures. But the technology is early, standards are fluid, and operators prioritize proven, swappable modules when uptime is paramount. Leading Chinese suppliers are hedging with R&D on CPO while monetizing today’s 400G and 800G ramps. Geographic diversification, like new Thai plants, helps manage export frictions without losing momentum.
The demand side is equally scaled. Tencent, the world’s largest video game vendor with a market cap near $594 billion, continues to push online services that are heavy on AI-driven content, search, and advertising optimization. Alibaba’s Singles’ Day remains the world’s biggest shopping event by gross merchandise volume, a load test for AI-optimized logistics and recommendation engines that spill into year-round capex. BYD’s ascent as China’s best-selling auto brand since 2023 adds connected-electrified fleets that stream data and require map, voice, and autonomy features, lifting compute needs across the stack. These platforms are not building AI to demo; they are building it to monetize at national scale. That dynamic feeds a stable, multi-year order book for servers, power, cooling, and—critically—optical interconnects.
Scale lowers cost curves and accelerates learning cycles. With a $23.8 trillion economy in sight by 2030, deep capital markets, and industrial depth that spans wafers to racks to grid, China can build at volumes that de-risk new technologies faster. It is not just about domestic wins. Thailand, Malaysia, and Indonesia are becoming second homes for Chinese data center builders and equipment suppliers, a gateway to serving U.S., European, and Asian multinationals in neutral jurisdictions. That export-led footprint compounds network effects and creates a self-reinforcing moat around reliability and time-to-install, two metrics hyperscalers prize.
Catalysts are lining up. Zhongji Innolight’s planned Hong Kong share sale above $3 billion, if executed, would signal balance-sheet firepower for 800G expansion and next-gen optics R&D. Watch for evidence of 800G moving from pilot to broad rollout across U.S. and China campuses, as well as procurement signals that favor multi-source vendor pools—good for Eoptolink and Accelink. On the data center side, keep an eye on pre-commit trends at GDS and the pace of new capacity in Southeast Asia. Policy will keep mattering: continued support for Western-region compute hubs and streamlined approvals for cross-border projects will set the tempo. Geopolitics and valuation noise will come and go. Operators with global footprints, export-ready plants, and a track record of shipping on time are likely to keep winning orders as AI moves from breakout to baseline infrastructure worldwide.
