Approximately five years ago, the data center industry operated under an informal consensus: optical transceiver data rates would roughly double every four years. 400G began large-scale deployment around 2019, and 800G became mainstream in 2023. This relatively predictable cadence gave procurement teams sufficient time to plan budgets, validate products, and complete deployments.
That rhythm has now been disrupted. The transition from 800G to 1.6T has taken approximately two years, and the industry expects the window between 1.6T and 3.2T could be even shorter. According to forecasts from LightCounting, market demand for 1.6T DR8 modules is expected to exceed 125 million units over the five-year period from 2027 to 2031. CITIC Securities, meanwhile, predicts that 1.6T optical modules could succeed 800G as the mainstream product category in 2027–2028.
This acceleration in data rate evolution has created a difficult decision for procurement teams: When should you transition from 800G to 1.6T? Should you already be allocating budget for 3.2T?
To answer these questions, buyers first need to understand the current state of the market. 800G optical modules remain the dominant shipping category, with shipments expected to exceed 40 million units in 2026. Four major North American cloud providers—Meta, Google, Microsoft, and Amazon—are projected to account for combined demand exceeding 30 million units. Meanwhile, 1.6T optical modules are entering large-scale adoption. NVIDIA is among the most aggressive drivers of this transition, with demand projected to exceed five million units in 2026, while Google's demand is estimated at approximately four million units. 3.2T optical modules have already completed system validation with leading customers. Accelink Technologies' 3.2T NPO products reportedly have orders in hand and are being delivered. Innolight's 3.2T silicon photonics single-mode NPO modules have also completed full-system validation with leading domestic cloud service providers (CSPs), making the company the first optical module manufacturer in the industry to achieve this milestone.
These figures point to three important realities: 800G will not disappear in the near term, 1.6T is ramping up rapidly, and 3.2T is no longer merely a future technology—it is already becoming a present-day reality. As a result, procurement strategies need to move away from a linear transition model toward a multi-speed, parallel deployment approach.
A practical procurement framework can be built around three scenarios. For expansion projects in existing data centers, continuing to purchase 800G modules remains a highly cost-effective option, provided rack space and thermal conditions permit. The 800G supply chain is the most mature, and pricing has benefited from economies of scale and increasing market competition. For newly built AI training clusters, 1.6T modules should be considered the default option. AI clusters demand substantially higher bandwidth density than traditional cloud computing environments, and the port-density advantages of 1.6T can significantly reduce switch counts and cabling complexity. For cutting-edge hyperscale training clusters, buyers can begin evaluating the feasibility of 3.2T NPO solutions, particularly when GPU deployments exceed 10,000 units.
It is also worth noting that optical interconnect technology is moving from a single dominant architecture toward a more diversified landscape. Pluggable optical modules are expected to remain the primary solution over the next three to five years, while NPO (Near-Packaged Optics) and CPO (Co-Packaged Optics) are gaining momentum. The projected lifecycle of NPO has been revised upward to at least five years, significantly longer than the previously anticipated two to three years. Large-scale CPO adoption is expected to begin after 2027, although the technology is already demonstrating distinct advantages in switching systems that require extremely high bandwidth density.
This parallel development of multiple technology paths places new demands on procurement teams. Does your supplier have full-stack technology capabilities? A supplier focused exclusively on traditional pluggable modules may face technology transition risks over the next three years. By contrast, a supplier with capabilities spanning pluggable optics, NPO, and CPO can offer a smoother and more flexible upgrade path.
When evaluating the timing of a speed upgrade, one frequently overlooked factor is compatibility with existing fiber infrastructure. Moving from 800G to 1.6T generally does not require changing the fiber type, although links may need to be retested and optimized. The transition from 1.6T to 3.2T may require closer attention to fiber dispersion characteristics and connector cleanliness requirements. Although these factors do not directly determine which optical modules buyers purchase, they can significantly affect the total cost and deployment timeline of an upgrade.
Finally, regarding “bandwidth anxiety,” the real problem is not anxiety itself but the lack of planning. Accelerating data rate evolution is a clear and persistent trend. The AI computing race is unlikely to slow down, and demand for optical interconnect bandwidth will continue to grow.
As a buyer, your objective should not be to place a bet on a single technology. Instead, you should establish a procurement framework that can flexibly accommodate multiple technology paths. Prioritize partners with clearly defined technology roadmaps, proven large-scale manufacturing and delivery capabilities, and investments across several complementary technology directions.
Throughout the journey from 800G to 3.2T, the best procurement strategy is neither to buy the fastest product nor to buy the cheapest one. It is to buy the solution that best fits your requirements—and to ensure that when an upgrade becomes necessary, you can make the transition at the lowest practical cost.
