The optical module industry is undergoing an unusual “three-track parallel” phase: 400G remains the mainstay of shipments, 800G is rapidly ramping up to become the new mainstream, and 1.6T has entered early mass production. For North American data center operators, this generational overlap is both an opportunity and a challenge—choose the right cadence, and you can gain leading bandwidth capability at a reasonable cost; choose the wrong timing, and you may face the awkward situation of encountering upgrade pressure right after deployment.
First, clarify the maturity of each generation. After years of iteration, 400G technology and its supply chain are already very mature, price competition is full, and it remains the most cost-effective choice for general computing and storage traffic. But in AI training scenarios, the 400G bottleneck is already very obvious. For 800G, industry data shows that from 2021 to 2025, the CAGR of 800G modules reached 155.3%, making it one of the fastest-adopted optical module generations in history. In 2026, the 800G supply chain has become fully mature, and the combined monthly capacity of major suppliers exceeds the million-unit level.
The situation for 1.6T is more subtle. 2026 is indeed the turning year for 1.6T to move from the lab to the production line, and the global market size is expected to exceed US$50 billion, a year-on-year increase of more than 60%. However, the pace of large-scale deployment needs rational assessment. The consensus among industry analysts is that large-scale mass production of 1.6T is expected to be achieved from the end of 2026 to early 2027, and substantial compression of the price system will take another 12 to 18 months. This means that if your project is delivered within 2026, 800G remains the lowest-risk and most mature-ecosystem choice; if the project cycle extends into the second half of 2027 and beyond, 1.6T planning needs to be laid out in advance.
Another key dimension of procurement cadence is network architecture compatibility. 800G modules were designed with backward compatibility in mind—QSFP-DD800 ports can accept QSFP-DD 400G modules, which provides flexibility for phased deployment. But the situation for 1.6T is different: the OSFP-XD package has limited compatibility with existing ports, and the transition from 800G to 1.6T will likely require simultaneous upgrades of switches and line cards. Therefore, when planning procurement, “optical module upgrade” and “switching equipment upgrade” need to be considered as one overall project, rather than tendered separately.
From the perspective of price trends, the market environment in 2026 is favorable to buyers. The average price of 1.6T modules in 2025 was about US$1,000 to US$1,200 per unit. As capacity ramps up and yields improve, downward price movement is a certain direction. The price of 800G modules has already entered a relatively stable range, with the premium of high-end products mainly supported by technological differentiation and supply assurance capability.
For North American data center operators, the recommended procurement strategy is: for AI training clusters delivered currently, use 800G as the main configuration to ensure certainty of performance and delivery time; at the same time, reserve 1.6T upgrade interfaces and power headroom in the architecture design, and place the 1.6T procurement decision in the 2027 budget cycle. This “use one generation, prepare one generation” cadence can maintain architectural competitiveness and investment safety during a window of rapid technological iteration.
