Table of Contents
Example
The growth in demand for computing power driven by AI continues unabated. Recently, industry insiders reported that Nvidia will increase its order volume for 800G optical modules in 2026, with an expected increase of 35%.
During the National Day holiday, the mutual investment among American tech giants, with a spiraling ascent on both feet, has also brought tangible order expectations for AI infrastructure. This includes OpenAI placing a five-year, 300 billion USD cloud service order with Oracle; Oracle purchasing 300 billion USD worth of Nvidia GPUs for data center construction; Nvidia and OpenAI reaching a 100 billion USD investment agreement; and OpenAI forming a strategic partnership with AMD, planning to deploy 6GW of AMD GPU capacity. As part of the agreement, AMD will also issue up to 160 million shares of AMD common stock warrants to OpenAI, accounting for approximately 10% of AMD’s shares.
These expectations for GPU procurement, often amounting to hundreds of billions of dollars, will undoubtedly boost order demand across various upstream and downstream segments of data centers.
InnoLight and Eoptolink secure 60% of the orders
According to sources, the additional 800G optical module orders from Nvidia are primarily captured by InnoLight and Eoptolink, accounting for 60%, with the remaining 40% divided among American optical communication vendors such as Coherent, Lumentum, and Broadcom.
Currently, Nvidia uses 800G optical modules in GB200 servers. For example, in a GB200 cluster with 576 GPUs, 18,432 single-end 800G DR8 optical modules are required. Under the current massive data center cluster demand, the consumption and value of optical modules are significant.
Thanks to the explosive growth in demand from AI data centers, all optical module manufacturers have experienced a breakout in their performance this year. Eoptolink achieved an operating revenue of 10.437 billion RMB in the first half of this year, representing a substantial year-on-year increase of 282.64%; net profit attributable to shareholders reached 3.942 billion RMB, soaring by 355.68% year-on-year. The rapid growth rate far exceeds the industry average, and the company stated that this was mainly driven by the high-speed optical modules’ volume increase under the influence of global AI computing power demand.
Among them, the shipment volume of 800G optical modules accounts for over 60% of Eoptolink’s total shipments (approximately 4.17 million units), becoming the core driver of revenue growth. Additionally, the company’s growth is also attributed to deep binding with North American cloud giants such as Nvidia, Meta, and Microsoft, as well as holding a 20% share of the 800G LPO module orders in Nvidia’s GB200 servers, which brings stable growth expectations.
InnoLight is similarly focused on overseas markets. In the first half of this year, it achieved an operating revenue of 14.789 billion RMB, a year-on-year increase of 36.95%; net profit attributable to shareholders was 3.995 billion RMB, an increase of 69.4% year-on-year; gross margin for the first half year reached 39.33%, up 6.2 percentage points compared to the same period last year.
Within this, InnoLight’s high-end products of 800G and above account for over 60% of revenue, with approximately 4.7 million units of 800G optical modules shipped. The share of silicon photonics solutions increased from 35% in 2024 to 60%. The increase in the proportion of high-end product shipments has significantly driven up gross profit margins.
At the same time, InnoLight has established deep binding relationships with North American cloud giants like Nvidia, Google, and Meta. Specifically, orders for 800G modules in Nvidia’s GB200 servers account for over 50%, while Google’s OCS customized modules account for 10%.
It is foreseeable that with the follow-up orders from Nvidia and AMD’s AI servers gradually coming to fruition, the performance of optical module manufacturers will continue to be driven upward.
GB300 begins shipments, ushering in the 1.6T optical modules as successors
In recent years, the explosive growth in AI computing power demand has directly propelled the rapid expansion of the 800G optical module market. According to industry forecast data, the global 800G optical module market size is expected to reach 8.6 billion USD in 2024, increasing to 12 billion USD in 2025, a year-on-year growth of 40%. By 2028, the overall market size for 800G and 1.6T optical modules will exceed 32 billion USD, with a compound annual growth rate (CAGR) of over 25%.
Following the 800G optical modules, as AI server clusters demand higher interconnection speeds, Nvidia has chosen to fully transition to 1.6T optical modules in GB300 servers and has also provided upgrade options to 1.6T modules on GB200. Meanwhile, products of 1.6T from Eoptolink and InnoLight have completed verification.
InnoLight previously stated during investor relations activities that in the second quarter of this year, 1.6T optical modules had begun phased shipments, with continuous mass production or scaled shipments expected in the second half of the year. At the 2025 OFC Optical Communications Conference, InnoLight’s subsidiary showcased products including the 1.6T-DR8 and 2xFR4 OSFP optical modules equipped with a new 3nm DSP chip, as well as 1.6T-2xLR4 optical modules based on TFLN MZM technology.
Eoptolink has also launched 1.6T DR8 series and 2FR4/4FR2 series products based on the OSFP-XD package, which have notable advantages in power consumption compared to competitors. The 1.6T OSFP 2VR4 optical module based on VCSEL technology plans to enter mass production in the fourth quarter of this year, and the 1.6T LRO optical modules have already secured orders.
The 1.6T optical modules are currently in a golden period of “demand explosion—technological maturity—capacity release,” with the years 2025-2026 being critical to shaping the industry landscape. Leading manufacturers, leveraging technological barriers, capacity advantages, and customer bonding, are expected to dominate this hundred-billion-level market contest. Future manufacturers should focus on increasing the proportion of 1.6T products, advancing independent development of optical chips, and breakthroughs in CPO technology—grasping the historic opportunities brought by the upgrade of AI computing power.