Why did the optical communication sector fall so badly? A breakdown of Rosenblatt’s latest report, plus my own take

Where do market concerns come from?
Mainly three factors compounding on top of each other:
First, the market worries that Chinese manufacturers are rapidly expanding capacity for phosphide indium (InP) lasers, fearing that future supply surges will trigger a price war.
Second, market rumors suggest CPO (co-packaged optics) may be delayed by about a year before large-scale deployment, causing many people to start questioning whether optical communication demand in the coming years might also be pushed back.
Third, many bears do not genuinely think the fundamentals of this sector are bad. Instead, they are using these news items and the pre-earnings “quiet period” to manufacture panic. Even more, a significant portion of the bears plan to cover short positions gradually in late July and early August before earnings are released. Some also plan to go long the entire sector again in 2027, positioning ahead for a CPO supercycle from 2028 to 2030.

Is China’s capacity expansion truly a real threat?
This is the question the market cares about most, but I think the concern is a bit overblown. First, even if Chinese manufacturers keep expanding, it is expected that by 2030 there will still be an approximate 40% supply gap for global InP lasers versus demand. The industry overall will remain short of supply, so capacity expansion will not cause oversupply in the near term.
Second, China currently mainly mass-produces low-power continuous-wave lasers and 100G EML. Applications are concentrated in relatively mature product lines such as silicon photonics transceivers and 800G modules. But for the 200G EML and high-power continuous-wave lasers that are truly most profitable in the coming few years, China does not yet have mature mass-production capability. And these two are exactly the core light-source components for CPO and NPO. This means Western companies will still hold a clear technology advantage in the high-end parts in the next few years.
Third, the China–US supply chain itself is gradually decoupling. For political reasons, supply-chain security, differences in technology routes, and different customer qualification systems, Western customers likely will not switch to purchasing products from Chinese manufacturers. Currently, many products from Lumentum and Coherent have production scheduled through 2027 or even 2028, and many orders also come from Nvidia’s early lock-ins.

Has CPO actually been delayed?
I tend to believe there is no substantive delay.
CPO can roughly be viewed in two stages:
The first stage is small-scale construction. Work starts in the second half of 2026, and shipments begin in 2027. The current progress is still moving forward as planned. It is expected that Nvidia and Broadcom’s production volumes in 2027 will each be around 30k to 350k units.
The second stage is the real large-scale ramp-up, with the timing in 2028. And during 2027, NPO (near-packaged optics) will serve as an interim solution to absorb part of the demand.

Will network architecture changes reduce demand for optical modules?
The market also worries that after new network technologies like MRC and RNG appear, the number of optical modules required per switch might drop. It is true that it could fall from the previous 6 to 7 down to 3 to 4, but the growth speed of GPU numbers far exceeds the magnitude of that decrease. Overall optical demand is still expected to keep rising, so this concern is also somewhat unnecessary.

Key takeaways from a few companies’ earnings
Lumentum’s earnings pressure is relatively the highest this time because it previously gave the most aggressive long-term guidance in the market. What investors are watching is whether it can further raise its 2027 targets—especially whether quarterly revenue can reach 2 billion USD (annualized 8 billion USD), and whether operating margin can exceed 40%.
Coherent is favored by many market institutions this time because the company’s in-house production share of 200G EML and high-power lasers continues to increase. In addition, it just received Google’s approximately 1.6T module order of about 1 billion USD per year, and also has Nvidia’s long-term purchase agreements from 2027 to 2030. Its profit margin is therefore expected to keep improving.
AAOI’s earnings this time could be a strong “beat” on both revenue and guidance. The main driver is 800G orders from Amazon and Oracle. It expects second-quarter revenue to exceed 200 million USD, third-quarter guidance close to 300 million USD, and gross margin to be clearly higher than the roughly 30% level in the first half.
ANET’s growth momentum remains strong. Its main customers include Meta, Microsoft, Oracle, OpenAI, Anthropic, and Google. Among them, Google could potentially become a customer with more than 10% of its revenue share in the future. A large amount of deferred revenue is expected to start being recognized from the second half of 2026, which will drive second-half revenue growth to exceed 30% and 2027 growth to exceed 40%.

The core conclusion of this report
To summarize, capacity expansion in China is indeed a real phenomenon, but in the short term it is not enough to change the global supply–demand landscape. The true technological barriers are in 200G EML, high-power lasers, and CPO-related products, where Western manufacturers still lead. 2028 remains the time when CPO truly ramps up at scale, while 2027 is covered as a transition by NPO.
The current round of pullback in the optical communication sector is driven more by sentiment and expectations; fundamentals have not deteriorated. The upcoming earnings season from late July to early August will be an important window to validate whether demand in this sector remains strong.

After reading this report, I want to add my own judgment
I think near-term shortages and future structural oversupply are not contradictory (very bullish in 2026–2027, but be careful after 2028). These two things can both be true at the same time—the key is to see which specific link oversupply appears first.
Right now, the whole industry has seen AI demand and is actively expanding capacity. Cloud providers also hope to bring in more suppliers to lower costs and diversify supply-chain risk. If, in the future, there are increasingly more suppliers for 800G and 1.6T modules and the rate of ASP decline starts to be faster than the rate of cost decline, then even if shipment volume and revenue continue to grow, gross margin may come under pressure first.
But based on current conditions, the industry has not reached that point yet. The information companies have disclosed still points to tight supply: production for high-end lasers scheduled out for the next few years, backlog orders staying at a high level, and export restrictions, yields, customer qualification and testing capabilities remaining the main bottlenecks. What is truly missing is qualified high-end capacity.
If the industry eventually enters a supply–demand rebalancing phase, I believe the first area likely to face price pressure is the optical module segment with higher standardization—such as 400G, progressively maturing 800G, and future 1.6T products entering the mainstream stage. For module-focused companies like AAOI, 中际旭创 (Eoptolink), 新易盛 (O-Net?), Source Photonics, more attention should be paid to changes in ASP and gross margin. However, this is a risk worth watching more after 2028. At least in my view, the industry is still in a supply–demand tight phase in 2026–2027, and companies like AAOI continue to benefit from order releases, capacity ramp-ups, and performance delivery.
By contrast, I like more the high-end component companies with core component barriers—for example, Lumentum (LITE) and Coherent (COHR); as well as system-level companies like Ciena (CIEN) and Nokia (NOK). These companies compete more through technology, network architecture, and system-integration capabilities. As for equipment and testing companies like Veeco (VECO), Aehr (AEHR), and Viavi (VIAV), as long as the industry continues expanding capacity, they are likely to keep benefiting from the capital expenditure cycle.

On the timeline, my view is basically consistent with this report: supply–demand tightness remains through the first half of 2027 (from 2026 to 2027). In the second half of 2027 through 2028, as new capacity is gradually released, the market will start repricing the supply–demand relationship.
It is still too early to discuss the optical communications industry as a whole experiencing comprehensive capacity oversupply. In the coming years, different products, different technology routes, and different companies will most likely enter completely different supply–demand cycles, so the entire industry will not all oversupply together.