Industry News
AI CapEx Boosts Next-Gen PAM4 Chipsets
Views : 118
Author : JIUZHOU
Update time : 2026-06-11 13:46:52
Currently, there is an explosive growth in generative AI and large language models worldwide. This is driving revolutionary changes in the physical layer architecture of data centers. The massive capital expenditures by hyperscale companies on AI infrastructure are flowing steadily through the supply chain toward the most critical underlying communication assets. These critical underlying communication assets are precisely the suppliers of high-bandwidth digital signal processors and optoelectronic chipsets.
The latest industry white papers and market research data indicate. The transition from 800G to 1.6T PAM4 architectures is proceeding at a much faster pace than previously anticipated. This will fundamentally reshape the global market landscape for photonic integration and data center interconnects between 2026 and 2031.
I. 800G Shipments Double Again; 1.6T Chipset Sales Set to Exceed $2 Billion
Driven by the large-scale deployment of AI computing clusters. High-bandwidth, low-latency optical interconnects have become a critical requirement. Industry statistics show that this market is experiencing exponential growth:
800G PAM4 Chipsets:
Following a nearly threefold increase in shipments and a doubling of sales revenue year-over-year. Current total shipments are expected to more than double again compared to last year.
1.6T Chipsets:
Rapidly transitioning from small-scale applications to large-scale deployments involving tens of millions of ports.
Market Size:
Global sales of 1.6T-related chipsets are projected to exceed $2 billion and maintain strong, rapid growth over the next few years.
In terms of application segments, traditional Ethernet and WDM continue to hold a significant share of the chipset market. However, the unique topologies within AI clusters are driving explosive growth in new connectivity forms:
Key Trends: While full-re-timing Ethernet transceivers will continue to contribute the largest absolute dollar growth. Sales of chipsets for active cables and on-board re-timers are experiencing a comprehensive surge. With linear optical devices following closely behind.
II. PAM4 vs. Coherent DSP: Diverging Technology Paths in AI Interconnects
In short-range AI data center clusters, PAM4 has demonstrated absolute dominance. In contrast, coherent DSP chipsets—traditionally used in long-haul backbone. And metropolitan area networks—have shown more moderate growth. Driven primarily by demand for traditional DWDM modules.
Due to the accelerated adoption of 1.6T PAM4 high-density optical components in AI backhaul networks. The sales gap between the two is widening further.
1. Technological Shifts in the PAM4 Market from 2027 to 2031
Industry analysis indicates that the large-scale deployment of linear driver solutions. And CPOs may displace sales of certain full-featured DSPs. Leading to a slight slowdown in the growth rate of PAM4 chipset sales in the later stages.
By eliminating or simplifying the DSP, LPO and CPO can significantly reduce system power consumption and latency. Metrics that are of paramount importance to AI supercomputing clusters.
2. The “New Battleground” for Coherent DSPs: Coherent-Lite and Inter-Building Interconnects
By 2027, coherent communication technology is penetrating data centers:
The Rise of Coherent-Lite Technology:
Lightweight coherent technology tailored for data center campuses will significantly drive sales of coherent DSPs.
Shipment Forecast: Global shipments of coherent DSPs are projected to exceed 8 million units by 2031.
On-Campus Cross-Facility Training Demand:
As the scale of distributed AI training expands, the demand for high-bandwidth. High-density interconnects spanning multiple data center facilities. It is likely to further boost market expectations for low-power coherent solutions.
III. Supply Chain Dynamics: The Time Lag Between Chip Demand and Transceiver Shipments
For global B2B buyers and supply chain managers:
This requires a rational perspective on the statistical discrepancy between chipset demand and actual optical module shipments.
Forecasts for high-end chipsets are typically based on long-term deployment projections. These projections are grounded in the anticipated long-term deployment of optical transceivers and active cables.
However, this model cannot fully capture, in real time, the short-term supply-demand mismatches. Caused by fluctuations in inventory levels across different stages of the supply chain. Therefore, establishing deep partnerships with underlying chip suppliers that possess vertical integration capabilities and stable production capacity is essential. This is the key barrier to ensuring the on-time delivery of AI computing infrastructure projects.
IV. Comparison of Key Technical Specifications
The latest industry white papers and market research data indicate. The transition from 800G to 1.6T PAM4 architectures is proceeding at a much faster pace than previously anticipated. This will fundamentally reshape the global market landscape for photonic integration and data center interconnects between 2026 and 2031.
I. 800G Shipments Double Again; 1.6T Chipset Sales Set to Exceed $2 Billion
Driven by the large-scale deployment of AI computing clusters. High-bandwidth, low-latency optical interconnects have become a critical requirement. Industry statistics show that this market is experiencing exponential growth:
800G PAM4 Chipsets:
Following a nearly threefold increase in shipments and a doubling of sales revenue year-over-year. Current total shipments are expected to more than double again compared to last year.
1.6T Chipsets:
Rapidly transitioning from small-scale applications to large-scale deployments involving tens of millions of ports.
Market Size:
Global sales of 1.6T-related chipsets are projected to exceed $2 billion and maintain strong, rapid growth over the next few years.
In terms of application segments, traditional Ethernet and WDM continue to hold a significant share of the chipset market. However, the unique topologies within AI clusters are driving explosive growth in new connectivity forms:
Key Trends: While full-re-timing Ethernet transceivers will continue to contribute the largest absolute dollar growth. Sales of chipsets for active cables and on-board re-timers are experiencing a comprehensive surge. With linear optical devices following closely behind.
II. PAM4 vs. Coherent DSP: Diverging Technology Paths in AI Interconnects
In short-range AI data center clusters, PAM4 has demonstrated absolute dominance. In contrast, coherent DSP chipsets—traditionally used in long-haul backbone. And metropolitan area networks—have shown more moderate growth. Driven primarily by demand for traditional DWDM modules.
Due to the accelerated adoption of 1.6T PAM4 high-density optical components in AI backhaul networks. The sales gap between the two is widening further.
1. Technological Shifts in the PAM4 Market from 2027 to 2031
Industry analysis indicates that the large-scale deployment of linear driver solutions. And CPOs may displace sales of certain full-featured DSPs. Leading to a slight slowdown in the growth rate of PAM4 chipset sales in the later stages.
By eliminating or simplifying the DSP, LPO and CPO can significantly reduce system power consumption and latency. Metrics that are of paramount importance to AI supercomputing clusters.
2. The “New Battleground” for Coherent DSPs: Coherent-Lite and Inter-Building Interconnects
By 2027, coherent communication technology is penetrating data centers:
The Rise of Coherent-Lite Technology:
Lightweight coherent technology tailored for data center campuses will significantly drive sales of coherent DSPs.
Shipment Forecast: Global shipments of coherent DSPs are projected to exceed 8 million units by 2031.
On-Campus Cross-Facility Training Demand:
As the scale of distributed AI training expands, the demand for high-bandwidth. High-density interconnects spanning multiple data center facilities. It is likely to further boost market expectations for low-power coherent solutions.
III. Supply Chain Dynamics: The Time Lag Between Chip Demand and Transceiver Shipments
For global B2B buyers and supply chain managers:
This requires a rational perspective on the statistical discrepancy between chipset demand and actual optical module shipments.
Forecasts for high-end chipsets are typically based on long-term deployment projections. These projections are grounded in the anticipated long-term deployment of optical transceivers and active cables.
However, this model cannot fully capture, in real time, the short-term supply-demand mismatches. Caused by fluctuations in inventory levels across different stages of the supply chain. Therefore, establishing deep partnerships with underlying chip suppliers that possess vertical integration capabilities and stable production capacity is essential. This is the key barrier to ensuring the on-time delivery of AI computing infrastructure projects.
IV. Comparison of Key Technical Specifications
| Technical Specifications | 800G Optical Modules | 1.6T Optical Modules | Linear Drive Technology (LPO/LRO) |
| Primary Modulation Methods | PAM4 (Single-Channel 100G/200G) |
PAM4 (Single-channel 200G) |
Direct Analog Signal Drive |
| DSP Dependency | Full-weight timing (requires high-performance DSP) |
High-density full-weighting (high chip power consumption) |
DSP-Free / Partial Retiming |
| Core Application Scenarios | Leaf-Spine Switches, AI Clusters |
AI supercomputing back-end networks, core switching layer |
Ultra-low-latency, low-power AI computing network |
| Core Technical Advantages | High technical maturity and excellent compatibility |
Doubled bandwidth, low cost per bit | 20%–30% reduction in power consumption, reduced thermal management demands |
Related News
Read More >>
AI Scale-up Switch Chip Market Experiences Explosive Growth
Jun .26.2026
Explore the rapid growth of AI scale-up switch chips powering hyperscale data centers, Ethernet ASICs, and CPO-driven next-gen networking architecture.
How Is AI Reshaping the Optical Communications Industry?
May .29.2026
With the rapid development of AI infrastructure, optical technology has become a cornerstone for scaling AI computing power.
Optical Connectivity: The Engine of AI Scaling
May .08.2026
With the rapid growth of generative AI and large language models, data centers are quickly becoming intelligent computing centers. The growing demand for computing power from AI models is not just about performance.
5G Core & MEC: AI-Driven Growth Through 2030
Apr .17.2026
Global communications infrastructure is currently witnessing a long-awaited "second wave of explosive growth."