Abstract: As 800G, 1.6T, and even CPO evolve rapidly, the optical module MCU is upgrading from a basic controller to a critical management chip for complex optoelectronic systems. GigaDevice is also leveraging its dedicated MCUs, Flash, analog capabilities, and supply chain synergy to further strengthen its competitiveness in the AI optical interconnect market.
In recent years, the rapid expansion of AI computing infrastructure has been transforming data transmission within data centers. 800G optical modules have entered the stage of large-scale deployment, and 1.6T is also accelerating its rollout. Amid this round of high-speed optical interconnect upgrades, the market has traditionally focused more on core components such as DSPs, lasers, modulators, and optical chips, but the MCU is equally indispensable.
The MCU in an optical module undertakes tasks such as current, voltage, temperature, and optical power monitoring, status reporting, and system management. A high-speed optical module can be valued at hundreds of US dollars, while the MCU within it typically costs only a few US dollars or even less, accounting for a very small proportion of the overall BOM cost. However, once this chip fails, the monitoring and protection mechanisms of the entire module could be compromised.
This is also one of the reasons why GigaDevice has continuously increased its investment in the optical communication niche market over the past few years. It is reported that GigaDevice's cumulative shipments of dedicated optical module MCUs have now approached 100 million units. Since 2020, its high-speed and low-speed products have been shipped consecutively for six years, establishing stable and in-depth cooperative relationships with leading module manufacturers.
As the data rate of optical modules continues to increase and internal components become increasingly dense, the challenges faced by MCUs are no longer just about 'whether the performance is sufficient.' Interface bandwidth, analog resources, package size, power consumption, reliability, and the next-generation CPO architecture are all redefining the role of this control chip. Around these changes, EEFOCUS interviewed Jeff Cui, Market Manager of the MCU Business Unit at GigaDevice, attempting to answer a more specific question: Under the rapid evolution of AI optical interconnects, what exactly has changed for the MCU inside optical modules?
Jeff Cui, Market Manager of the MCU Business Unit at GigaDevice
From Low-Speed PON to 1.6T: How Does GigaDevice Redefine Its Optical Module MCU Product Line?
It is reported that GigaDevice actually entered the optical module market in 2018 and subsequently launched its first dedicated optical module MCU. By 2022, the annual shipments of related products had reached the tens of millions level, gradually covering mainstream optical module and equipment customers both domestically and internationally. During the recent optical expo, GigaDevice disclosed that its cumulative shipments of dedicated optical module MCUs have approached 100 million units.
Jeff Cui revealed that the primary tasks of the MCU in optical modules are relatively clear: collecting parameters such as temperature, voltage, current, and optical power; executing digital diagnostics, status monitoring, and configuration management; and communicating with the host system via interfaces such as I²C. As 400G, 800G, and even 1.6T high-speed optical modules enter data centers, the workload undertaken by the MCU has begun to increase significantly. Higher data rates mean more complex DSPs, higher-density optoelectronic components, and stricter power and thermal management requirements, while the module form factor has not scaled up synchronously, leaving even tighter PCB space for the control chip and peripheral components.
This change is directly reflected in the definition of GigaDevice's latest two generations of products. The GD32E511/512 series, which is a key focus for promotion in 2026, is mainly targeted at 400G to 1.6T high-speed optical modules. It adopts the Arm Cortex-M33 core and adds new features for high-speed optical modules such as large-capacity dual-bank Flash and I3C. The GD32E251/252 series continues to cover applications such as PON, fronthaul, and industrial optical communication. It adopts the Cortex-M23 core, placing more emphasis on analog performance, cost-effectiveness, and the continuity of existing customer designs.
GD32E511/512 Series MCU
In terms of parameters, the GD32E512 supports an ultra-compact 3×3mm package and integrates multiple resources including I²C, I3C, MDIO, ADC, DAC, comparators, and Op-Amps. The considerations behind these configurations are not simply to pursue a 'higher-performance' MCU. The internal space of high-speed optical modules is limited; if functions such as DAC, comparators, and level shifters all rely on external chips, it will not only increase the BOM but also occupy PCB area. Meanwhile, as module management functions increase, Flash and RAM capacity, interface speed, and in-system upgrade capabilities must also be improved accordingly. Therefore, an MCU designed for high-speed optical modules needs to strike a new balance among digital performance, analog resources, power consumption, package size, and cost.
Jeff Cui told EEFOCUS reporters that GigaDevice's product iteration pace has also accelerated significantly in recent years. The first-generation products went through a long cycle from early layout to implementation, the second-generation products were launched around 2024, and by 2025 and 2026, new products have begun to iterate continuously. The reason is not complicated: the optical module itself is in a stage of rapid evolution, and the MCU cannot continue to be updated at the pace of traditional general-purpose controllers.
'Whether it is the increase in data rates or the evolution of form factors in this industry, the specification requirements for MCUs are always different, and the demands are extremely rigorous. A general-purpose MCU cannot meet all application scenarios and the ever-changing market demands,' said Jeff Cui.
This 'dedicated' approach does not mean that GigaDevice is only focusing on high-speed products. In fact, the GD32E251/252 still undertakes the task of upgrading low-speed optical module products. For mature applications such as PON and industrial optical communication, customers do not need to blindly increase the main frequency; they care more about whether existing designs can be migrated smoothly, whether analog accuracy is improved, whether the supply is stable, and whether costs can continue to drop. Jeff mentioned that when defining the new generation of low-speed products, the first consideration is compatibility with the previous generation to minimize customer switching costs, while improving temperature sensing, analog performance, and supply chain resilience.
This actually constitutes two parallel logics for GigaDevice's optical module MCU product line: one follows the rapid evolution of high-speed optical interconnects such as 800G and 1.6T, continuously adding I3C, dual-bank Flash, analog peripherals, and smaller packages; the other continues to serve mature markets such as PON and fronthaul, extending the product lifecycle through compatible upgrades and cost optimization. The former is more driven by AI data centers, while the latter still relies on the massive demand for communication infrastructure.
From the product evolution of GigaDevice, the optical module MCU is gradually transforming from a relatively general-purpose 'management chip' in the past into a type of device that needs to be redefined around specific optical module architectures. This also brings up another more critical question: since the CPU core itself is not the most difficult part of this market, what exactly is the true competitive threshold for optical module MCUs?
What Enables GigaDevice to Enter the Supply Chain of Leading Optical Module Manufacturers?
As mentioned earlier, the MCU accounts for less than 1% of the overall BOM of an optical module, but the cost of failure is extremely high. This also determines that the competitive logic of the optical module MCU market is different from that of ordinary consumer MCUs.
Jeff believes that the true threshold for optical module MCUs largely stems from long-term engineering know-how. The Arm Cortex-M series cores themselves are not scarce resources; many MCU manufacturers use CPU cores and basic IPs that are not fundamentally different. The differences are more reflected in how these IPs are combined, and whether they have undergone sufficient reliability verification and exception scenario testing during mass production.
Take the commonly used management interface in optical modules as an example. I²C itself is not a new technology, but optical modules have significantly higher requirements for interface stability than many ordinary applications. The MCU must not only continuously collect and report data but also face environments with high temperatures, strong interference, and long-term continuous operation. Two products on the chip parameter table might both state 'supports I²C,' but whether they can work stably under different data rates, different loads, different temperatures, and abnormal states often requires long-term verification to reveal the differences.
Besides reliability, another key competitive point is how analog resources are integrated. High-speed optical modules are not purely digital systems. In addition to completing logic control, the MCU needs to collect analog signals such as temperature, voltage, current, and optical power, and control some components. Therefore, analog peripherals such as ADCs, DACs, comparators, and Op-Amps will actually directly affect the BOM, PCB area, and system complexity of the optical module.
Jeff stated that there are currently roughly two types of product approaches in the industry: one integrates more analog IPs, has strong specificity, but relatively limited flexibility; the other leans more towards digital MCUs, delegating analog functions to external devices, which offers flexible design but increases the number of peripheral chips and PCB footprint. GigaDevice has chosen a middle path, integrating 'moderate' analog resources within the MCU to find a balance among integration, flexibility, cost, and power consumption.
This approach is quite evident in the GD32E512. This product supports multiple ADCs, DACs, comparators, and Op-Amps, while adopting a small 3×3mm package. For high-speed optical modules, eliminating even one external analog chip or level shifter brings not just a BOM change of a few cents or even a few US dollars, but more importantly, it frees up PCB area. As the component density inside 800G and 1.6T modules continues to increase, the value of this space is rising.
Jeff also mentioned that as the operating voltages of components like DSPs and Flash gradually decrease, traditional 3.3V MCU interfaces sometimes require additional level shifter chips. GigaDevice's new generation of products supports different voltage domain configurations such as 3.3V, 1.8V, and 1.2V, aiming to directly reduce such peripheral components. For high-speed optical modules that need to pack an increasing number of components within a fixed size, such engineering details often directly affect customer selection.
The third competitive factor is supply chain capability. Once an optical module MCU passes the certification of leading customers, it is usually not easily replaced. On the one hand, the cost proportion of the MCU itself is not high, so the savings from replacement are limited; on the other hand, re-validation requires time and engineering costs. Therefore, when choosing a supplier, customers will consider long-term supply, capacity elasticity, and compatibility between different product generations, in addition to performance and price.
Jeff stated in the interview that GigaDevice considers hardware compatibility in the design of both new and old products, and backs this up with different supply resources; optical modules have also become one of the company's higher-priority strategic directions. For customers, such capabilities may not be directly reflected in a specific chip parameter, but they will influence whether they are willing to put a seemingly inexpensive MCU into a long-term mass production project.
In addition, GigaDevice's simultaneous possession of MCU, Flash, and analog product lines brings it another supply chain synergy advantage. For customers, they can complete the selection of control, storage, and some PMICs within the same supplier system, reducing the adaptation work between different chips and lowering supplier management complexity. GigaDevice summarizes this as shortening the R&D cycle and reducing supply chain management costs. In summary, this explains why an MCU with a relatively low price proportion can still form strong customer stickiness.
After 1.6T, What Is the Biggest Change for Optical Module MCUs?
If the evolution of optical modules from 400G, 800G to 1.6T is simply understood as 'transmission speeds getting faster and faster,' it is easy to underestimate the changes it brings to peripheral chips.
For the MCU, the system management complexity of the entire optical module is increasing. There are more and more objects to be monitored inside the optical module, protocols are constantly upgrading, and component density, power consumption, and thermal density continue to rise, while the module size has not scaled up synchronously. This means the MCU must undertake more monitoring, communication, configuration, and fault handling tasks within a limited area and power budget.
As the data rate of optical modules increases, the supporting DSPs change, and the Flash, RAM, and related firmware resources that the MCU needs to manage will also change; if the module architecture further shifts to CPO or NPO in the future, some traditional functions may even be redistributed.
A very typical change is that the management interface is beginning to evolve from I²C to I3C.
For a long time, I²C has been one of the most common management communication interfaces inside optical modules. It features mature technology, low cost, and has formed a massive industry ecosystem. However, as the data that high-speed optical modules need to monitor and manage increases rapidly, traditional I²C has gradually exposed its limitations in bandwidth and response efficiency.
Jeff mentioned that as the CMIS protocol continues to upgrade, the parameters that optical modules need to monitor and report have increased from a dozen or so in the past to hundreds. Traditional I²C mainly adopts a polling method, and each command and data read requires a certain amount of time. As the number of monitored items increases, the entire status refresh cycle will also be lengthened. More importantly, high-speed optical modules contain higher-value and higher-power optoelectronic components. Once there is an abnormality in optical power, temperature, or other faults, the management system needs to respond more quickly, and the delay caused by the traditional polling mechanism could become a problem.
This is also one of the important backgrounds for GigaDevice to add I3C in the GD32E511/512. It is reported that the GD32E512 introduces the I3C high-performance communication interface, mainly to meet the new demands of high-speed optical modules in terms of high bandwidth, low latency, and high-density communication.
However, Jeff does not believe that the value of I3C is just 'running faster.' 'It must be about speed, but the reasons behind the speed are these.' In his view, the interface upgrade involves at least three layers of changes: first, the increase in monitoring parameters requires greater data throughput capacity; second, abnormal events require faster responses; third, the power and thermal design of high-speed modules are becoming increasingly strict, and the communication interface itself must also consider energy efficiency.
Of course, this upgrade will not be completed in a short time. The optical communication industry has a large number of mature designs and customer platforms, and I²C still has a strong compatibility advantage. Therefore, the new generation of optical module MCUs must support both old and new interfaces. Jeff mentioned that the industry is actually in a transition phase from I²C to I3C, and products cannot just lean to one side; they also need to accommodate different internal communication methods such as SPI and MDIO.
Besides communication interfaces, another increasingly prominent contradiction is that functions are constantly increasing, but the space left for the MCU has not grown.
High-speed optical modules still widely adopt compact packages such as QSFP-DD. After the data rate increases from 400G to 800G and 1.6T, the requirements for internal DSPs, optical devices, power supplies, and other auxiliary chips continue to increase, but the PCB area will not double accordingly. For the MCU, this means that on the one hand, it needs larger Flash, richer interfaces, and analog resources, and on the other hand, it must continue to shrink the package and control power consumption.
GigaDevice's GD32E512 achieves a 3mm×3mm package while integrating various analog peripherals such as ADC, DAC, comparators, and Op-Amps, precisely to cope with this engineering constraint. Jeff talked about how a small package is not simply about making the chip 'smaller,' but about constantly making trade-offs among Flash capacity, analog IP, power consumption, cost, and layout. As the data rate of optical modules increases, the MCU requires more and more storage and functional resources, but the physical size is required to be further compressed, which poses higher requirements for chip definition and layout design.
This also explains why optical module MCUs are not suitable for simply 'stacking performance' according to the idea of general-purpose MCUs. Therefore, in the future, the true competition for optical module MCUs will likely be about how much system management function can be completed per unit area and per unit of power consumption.
Finally, if optical interconnects gradually shift from pluggable optical modules to NPO and CPO in the future, and the form factor of traditional optical modules changes, will the role of today's optical module MCUs also change accordingly?
Jeff's judgment is that the functions undertaken by the MCU are likely to be redistributed. In traditional high-speed optical modules, the MCU needs to work in conjunction with the DSP, managing firmware, status information, and a large amount of digital resources. In some CPO architectures, components such as DSPs and built-in light sources may be re-laid out or even moved externally, and some large-capacity Flash requirements related to DSPs may also decrease. In this case, the pressure on the MCU's digital resources may decrease, but the requirements for system monitoring, analog detection, communication interfaces, and security functions will instead increase.
In other words, CPO is not simply 'moving' today's optical modules 'next to the chip,' but rather redefining the boundaries between optical, electrical, control, and management functions. For MCU manufacturers, the future challenge is not just to upgrade products along with the data rate of optical modules, but to anticipate in advance which functions will remain, which will be moved externally, and what new control requirements will be added in the new system architecture.
From a static perspective, if a 400G module requires one MCU, then theoretically, when achieving the same total bandwidth, the number of 1.6T modules only needs to be a quarter of the original, and the MCU demand seems to decrease as well. However, Jeff believes that this calculation ignores the fact that the total bandwidth of AI data centers is itself growing rapidly. While the data rate of a single module increases, the scale of GPU clusters, interconnect density, and east-west traffic are also expanding. Therefore, higher data rates do not necessarily mean a reduction in the total volume of MCUs. In the interview, he explicitly stated that, based on current customer demands and industry evolution, the overall demand for high-speed optical interconnect MCUs is still expanding.
Conclusion: Who Can Participate in Defining the Next-Generation Optical Modules?
From early PON and fronthaul, to 400G, 800G, and 1.6T, and then to the CPO currently being promoted, the control requirements inside optical modules have undergone obvious changes. For MCU manufacturers, future competition will no longer just revolve around performance parameters, but must follow or even anticipate the changes in optical module architecture.
GigaDevice's path over the past few years reflects this trend to some extent: on the one hand, it continues to cover the mature low-speed optical communication market through the GD32E251/252; on the other hand, it uses the GD32E511/512 to tap into the 400G to 1.6T high-speed optical module market, while simultaneously expanding the market coverage of Flash and analog products.
The author believes that optical module MCUs are not a type of product that can 'meet all changes with constancy.' Data rates are changing, module form factors are changing, DSPs, light sources, and communication protocols are also changing, and the corresponding MCU resource configurations need to be continuously adjusted. For domestic MCU manufacturers, the past was more about solving the supply problem. After the rapid evolution of AI optical interconnects, the competition is entering a new stage: whether they can anticipate the system requirements under 1.6T, 3.2T, or even CPO architectures, and write these requirements into the chip definition in advance, is becoming the new watershed.