One of the most significant developments that has recently caught my attention in the optics sector is the substantial upward revision in demand forecasts for 800G and 1.6T transceivers for 2027.
According to data shared during a BofA analyst call regarding the AI optical communications supply chain on September 4, 2026, demand for both 800G and 1.6T transceivers could individually approach 100 million units by 2027. Subsequent supply chain summaries pegged this expectation at 90–100 million units.
If these figures materialize, it implies a potential demand of approximately 180–200 million units for 800G and 1.6T transceivers alone.
However, the truly critical factor here is not demand, but supply.
BofA’s supply chain checks indicate that DSP capacity could pose a significant bottleneck. Additionally, there are capacity constraints regarding mSAP PCBs, CW lasers, and other critical components.
In other words, the issue facing the optics sector in the near future may not be whether there is demand driven by AI. The real question is whether there will be sufficient components and manufacturing capacity to meet that demand.
In my view, this is precisely the most crucial aspect of the investment thesis for the optics sector.
We often see a simplistic narrative in the market: "AI data center investments are rising, so optics companies stand to gain." Yet, the issue is no longer just about the total capital allocated to AI infrastructure. What matters more is how that spending translates into specific volumes of transceivers, operating at specific speeds and utilizing specific components. In BofA’s previous forecasts, demand for 2026 stood at 50–60 million units for 800G and approximately 30 million units for 1.6T. Consequently, the expectation of 90–100 million units for 2027 signals very strong growth, particularly in the 1.6T segment.
The transition to 1.6T is particularly significant here.
While 800G has become a more established product for many hyperscalers, the production and component processes for 1.6T are more complex. New capacity requirements are emerging in areas such as DSP, lasers, packaging, signal integrity, and thermal design.
$COHR reports that multiple customers are ramping up 1.6T production simultaneously and that some are accelerating the transition. The company also notes that 1.6T demand has strengthened faster than anticipated recently.
A similar picture emerges for $LITE. BofA views the company as a key player in the optical supply chain, particularly due to its position in EML and CW lasers. Long lead times for capacity expansion are also cited as a factor contributing to supply constraints.
$AAOI is particularly noteworthy regarding capacity expansion.
The company expects demand for 800G and 1.6T modules to exceed current production capacity through mid-2027. It aims to reach a monthly production capacity of over 650,000 units for 800G and 1.6T by the end of 2026, and over 930,000 units by the end of 2027. More than half of this output is planned to come from Texas. For this reason, the key question regarding optics, from my perspective, is this:
How much of the hundreds of billions of dollars spent on AI will translate into demand for optics, and which companies will be able to turn that demand into products in a timely manner?
This is because a significant supply-demand imbalance could emerge here.
If demand approaches 200 million units while capacity for a critical component falls well short of that figure, a substantial gap could open up between theoretical demand and actual shipments. In such a scenario, the focus must shift beyond the question of "which company will sell more products?" to "which company possesses the capacity and market share for the bottleneck components?"
In this regard, the narratives of the most discussed photonics companies—$COHR, $LITE, and $AAOI—differ from one another.
For $COHR, the focus on technologies such as silicon photonics, EML, and VCSEL for 1.6T—alongside their InP and laser ecosystems—is significant.
For $LITE, critical components like EML and CW lasers stand out.
Meanwhile, $AAOI is aggressively expanding its manufacturing capacity, particularly in Texas, USA.
However, the story does not end with 800G and 1.6T.
Previous analyses by BofA projected that NPO/XPO solutions would begin to gain volume from 2027 onwards, while the 3.2T pluggable segment would likely accelerate around 2028.
In other words, the 800G-to-1.6T transition we are discussing today is merely one phase of a much longer optical bandwidth cycle.
For me, the most important investment takeaway here is this:
In the optics sector, it is no longer sufficient to look solely at the "AI is growing" narrative. Key factors to monitor in tandem include the transition speed from 800G to 1.6T, transceiver density per hyperscaler, DSP capacity, laser supply, manufacturing capacity, silicon photonics penetration, and the timing of the shift to 3.2T/NPO.
This is because as the compute capacity of AI clusters expands, so does the bandwidth required for GPUs to communicate with one another and with different sections of the data center. Simply increasing the number of GPUs is not enough; these GPUs must be interconnected.
Therefore, regarding optics over the next few years, the primary issue I will be tracking is not so much the question of "Are AI investments continuing?" but rather:
“At what rate is bandwidth demand growing, and which companies possess sufficient capacity to meet it?”
If total demand for 800G and 1.6T technologies truly reaches 180–200 million units by 2027, the resulting bottleneck itself could become a significant investment theme for the sector.
At the same time, however, one must not overlook how emerging architectures—such as 3.2T, NPO, and CPO—will alter the market landscape for pluggable optics in the coming years.
In my view, regarding the optics sector, the key factor to monitor in the period ahead will not merely be the growth of AI-driven demand, but the speed at which that demand can physically translate into actual production and shipments.
Not investment advice.
