Nomura Optical Communications Electronic Chip Experts Meeting: 1.6T supply is the main bottleneck, and the shortage will last until at least the second half of 2027.
The 1.6T optical module digital signal processor (DSP) is facing a structural supply gap, and this gap will last until at least the second half of 2027. This is the core judgment revealed by Nomura in a conference call with AI infrastructure service providers experts organized on September 22. The root cause of the supply-demand imbalance lies in the insufficient capacity and low yield of the 3-nanometer process, rather than short-term fluctuations in demand.
According to data from the trading platform Zhuifeng, Nomura, in the minutes of a conference call with experts in optical communication chips, pointed out that demand for 1.6T DSPs is projected to exceed 20 million units in 2026, but actual shipments will only reach 16 to 17 million units , leaving a gap of approximately 3 to 4 million units. The supply-demand imbalance will further widen in 2027 —demand will climb to 60 to 70 million units, while supply will only reach 40 to 50 million units, expanding the gap to over 20 million units. Experts clearly indicated that significant capacity release will not occur until the second half of 2027.
In terms of the competitive landscape, the high cost of advanced process tape-out, the approximately two-year customer verification cycle, and the complex signal processing algorithms have collectively built a moat in the high-end DSP market, causing market share to concentrate on existing suppliers and prices to decline moderately rather than engage in fierce price wars . Meanwhile, the evolution of the NPO/CPO route is reshaping the value distribution of the industry chain—DSPs will be removed, and equalization and error correction functions will shift to transimpedance amplifiers (TIAs) and driver chips, placing higher demands on the linearity of related devices.
Gap size: Expected to continue expanding from 2026 to 2027
The supply and demand trajectories of 800G and 1.6T DSPs show completely different trends.
The meeting minutes stated that, according to experts, shipments of 800G DSPs are projected to reach approximately 25 to 26 million units in 2025, approaching 50 million units in 2026, and further jumping to 120 to 150 million units in 2027. This increase will come not only from AI data centers but also from the telecommunications market, which is entering a replacement cycle. 800G DSPs utilize 7nm and 5nm processes, resulting in a relatively ample supply.
The situation is quite different for 1.6T DSPs. Shipments of 1.6T DSPs will only exceed 3 million units in 2025. While demand is projected to exceed 20 million units in 2026, actual shipments will only reach 16 to 17 million units. Demand is expected to expand further to 60 to 70 million units in 2027, while supply will only reach 40 to 50 million units. Experts predict that substantial capacity release will not occur until the second half of 2027, meaning that the supply-demand gap will be structural rather than cyclical for a considerable period.
The direct cause of this gap is the process bottleneck: 1.6T DSP requires a 3-nanometer process, but the current production capacity of this process is insufficient and the yield is low. In addition, some forward error correction (FEC) algorithms have vulnerabilities that need to be fixed, and there are also compatibility issues between high-speed DSP interfaces and network switches. The superposition of multiple technical obstacles further suppresses the release speed of effective supply.
Competitive barriers: New entrants find it difficult to enter the market with low prices.
The competitive landscape of the high-end DSP market has not unfolded according to the conventional logic of price wars.
According to experts, DSP technology has extremely high barriers to entry, mainly in three dimensions : First, the manufacturing process must use advanced processes below 7 nanometers, resulting in extremely high tape-out costs; second, the signal processing algorithms are complex, covering core capabilities such as FEC, bit error rate control, equalization technology, and dispersion compensation; and third, power management and heat dissipation control also constitute significant barriers to entry.
Customer verification cycles further strengthen the position of existing suppliers. DSP testing requires joint participation from optical module manufacturers and end customers, with a single testing cycle potentially lasting up to two years. Switching to a new supplier is extremely costly for end customers, leading them to prefer continuing their partnerships with existing suppliers. While some second-tier manufacturers possess the production capacity for 800G DSPs, they lack a strong customer ecosystem, making it difficult for them to substantially penetrate the market.
In the Chinese market, manufacturers such as Sitrus Technology and Joywell Semi are developing 800G DSPs, but have not yet achieved large-scale shipments.
Price trend: Moderate decline, with smaller price drop for the 1.6T model.
Tight supply and demand have supported the relative stability of DSP prices. According to experts, the current price of 800G DSP Long Term Agreement (LTA) is $45 to $70, and the price of 1.6T DSP LTA is $120 to $150, with retail prices 10% to 20% higher than contract prices.
Looking ahead to 2027, DSP prices are expected to decline slightly as production capacity gradually increases. Experts predict that contract prices for 800G DSPs will drop by about 10%, while the price decrease for 1.6T DSPs will be relatively smaller due to a larger supply gap. This price path aligns well with the competitive market landscape—existing suppliers are able to maintain relatively stable pricing power through technological barriers and customer loyalty, rather than engaging in fierce price competition.
TIA/Driver Chips: Value Shifts Partially, Chinese Manufacturers Still Lag Behind
The evolution of the NPO/CPO packaging route is reshaping the value distribution of the industry chain. According to experts, under the NPO/CPO architecture, the DSP will be eliminated to reduce power consumption. Functions such as equalization, bit error correction, and distortion correction will be partially undertaken by the TIA/driver chip and partially by the switching chip. This places higher technical requirements on the linearity of the TIA and driver chip.
From the current technological advancements, leading TIA/driver manufacturers have achieved mass production of 200G per channel, while 400G is still under development. The main technological barrier lies in simultaneously achieving high linearity, low noise, and low power consumption. Since TIA and driver chips can utilize the mature 28nm process, Chinese manufacturers have achieved mass production of 100G products, and 200G is in the sampling stage. However, there is still a significant gap between domestic and overseas manufacturers in terms of product stability and customer verification cycles.
Google has adopted 2.4T coherent-lite transceivers in its Tensor Processing Unit (TPU) and Optical Switch (OCS) networks. This solution uses only a single fixed wavelength O-band, covering transmission distances of 2 to 20 kilometers. Experts point out that the technological barrier for coherent-lite DSPs is lower than that for traditional coherent DSPs, but higher than that for standard PAM4 DSPs.
Verification window: The first half of 2027 is a key observation period.
The above supply and demand assessment has several verification conditions that require continuous monitoring.
In terms of results , the actual shipment volume of 1.6T optical modules and electrical chips, as well as the release schedule of 3nm DSP supply in the second half of 2027, are the most direct verification indicators. In terms of mechanisms , the progress of DSP yield improvement and FEC algorithm defect repair, the sampling and customer verification progress of 400G TIA/drivers, and the adoption rate of NPO/CPO on the switch side all need to be closely monitored.
Experts believe that the current gap assessment needs to be revised downward if the following situations occur : the supply of 1.6T in 2027 is significantly higher than 40 million to 50 million units; the price drop of DSPs significantly exceeds the assumption of about 10% of the contract price; or the accelerated substitution of NPO/CPOs leads to the premature compression of DSP demand.
The first half of 2027 (a critical observation period before capacity release) is a key verification window, and the quarterly performance releases of major manufacturers and optical module shipment data will be important observation points.
~~~~~~~~~~~~~~~~~~~~~~~
The above content is from Zhuifeng Trading Platform .
For more detailed analysis, including real-time updates and firsthand research, please join the [ Trading Channel Annual Membership ].
Risk warning and disclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.