TL;DR
The recent downturn in optical stocks is not just about the CPO narrative being tempered, but rather the market reassessing a more sensitive issue: whether the 2027-2028 ramp is a period of execution or merely continued validation.
CPO (Co-Packaged Optics) itself has not been invalidated. The bandwidth, power consumption, and switching density pressures in AI data centers continue to rise, and the physical constraints of copper cables and traditional pluggable optics have not disappeared. The issue lies in the fact that the run-up in related stocks had already implied an aggressive timeline: after NVIDIA pushed CPO into the commercialization window, the optical engine, laser, silicon photonics, and switch chips are expected to ramp into volume production rapidly in 2027-2028.
The June 9th report from SemiAnalysis precisely hit on this pricing assumption. According to public accounts, the report suggests that NVIDIA's 800V DC and large-scale CPO production may be delayed to around 2028-2029, with 400V DC still on track for 2026, while some NPO (Near-Packaged Optics) projects could see acceleration. Following the report-induced market turbulence, optical and related chain tickers such as AAOI, LITE, COHR, GLW, MRVL experienced high-single to double-digit pullbacks. The market is not trading on the "validity of the CPO direction" but on "how quickly CPO can turn into orders."
However, this is not a one-sided bearish view. The "White-Haired Stock God," AI supply chain analyst Serenity (@aleabitoreddit), subsequently refuted SemiAnalysis, suggesting it overly relies on a conservative engineering model and underestimates NVIDIA's ability to compress the hardware cycle. Based on their interpretation of signals from NVIDIA, Lumentum, Foxconn, and others, they emphasize that CPO is still on track for the second half of 2026, the second half of 2027, and 2028.
The value of this debate is not in determining who "won," but in shifting the valuation anchor of the optical chain from an endgame narrative back to a timeline validation: CPO is coming, but the slope at which it arrives will determine the value distribution among NPO, pluggable modules, light sources, and switch chips.
Timeline Reassessment Behind Optics Stock Slide
Over the past few months, the market has been buying into the optical chain not for current revenue but for the capital expenditure shift in AI data center next-gen network architecture.
As model training and inference clusters grow, the communication pressure between GPUs, racks, and within data centers continues to rise. The network is no longer just an ancillary system outside the servers but is increasingly resembling an efficiency bottleneck in the AI factory. The higher the bandwidth density and the lower the power consumption, the higher the upper limit of scalability for a unit of compute cluster, which is why CPO has been brought to the forefront.
The theoretical attraction of CPO is quite straightforward: getting the optical engine as close to the ASIC as possible, shortening the high-speed electrical signal path, reducing the power consumption, loss, and signal integrity pressure brought by serial-to-parallel conversion circuits and copper traces. Compared to traditional pluggable modules, CPO has a better power consumption and density imagination space in the high-bandwidth era.
The market's issue is that it easily trades the "right direction" into "volume certainty" prematurely. Nvidia's official press release stated that the Vera Rubin platform would introduce Spectrum-X Ethernet Photonics, with CPO switches already in production for lateral expansion and cross-cluster deployment in the AI factory. In a June 3 report, Nvidia's networking executives mentioned that Spectrum-X CPO switches had been shipped to some partners, and capacity is expected to ramp up in the second half of 2026.
These signals are sufficient to prove CPO is advancing, but they do not directly equate to risk-free realization of production-level large-scale orders. For the capital markets, there is a significant valuation gap between "going into production," "shipping to some partners," "customer evaluations," and "massive production." The pullback triggered by the SemiAnalysis report is essentially the market beginning to reevaluate these milestones.
SemiAnalysis's Conservative Model: CPO Challenges in Systems Engineering
SemiAnalysis is not saying CPO has no future. Its core assessment is more like: the theoretical advantages of CPO are clear, but the scaling implementation is slower than the market imagines.
The reason is not just one or two devices not being ready but that CPO has compressed the complexity that was originally dispersed in modules, boards, and whole machines into a more deeply coupled system. The higher the integration, the prettier the single-point performance, and the pressures on manufacturing, testing, maintenance, and supply chain flexibility will also rise together.
The advantage of traditional pluggable modules is modularity. If a certain optical module fails, it can be replaced, and vendor switches are relatively easier. CPO is different; the optical engine is closer to the ASIC, even entering the same packaging system. Power consumption and density benefits come from this close coupling, but the maintenance radius also expands. Once an optical component fails, the impact is no longer just a easily swappable module but may involve higher-value switching chips and whole machine systems.
Prior to this, SemiAnalysis' CPO Book has repeatedly emphasized serviceability, reliability, yield, and supply chain maturity. Especially in the hyperscale cloud provider scenario, performance is not the sole metric. Large customers have high requirements for reliability and serviceability. If the failure rate, repair processes, and replacement costs in the production environment are uncontrollable, even the best power consumption model may be delayed in deployment.
InP lasers are also a point of contention. Lab-level port uptime data can prove the technology's feasibility, but it does not equate to covering long-term operation, mass production, on-site maintenance, and supply chain redundancy in large-scale data centers. For investors, such differences are critical: lab validation indicates the direction, while on-site reliability determines scalability.
Within SemiAnalysis' framework, NPO and pluggable modules are not a backward route but a more realistic intermediate layer until engineering risks are fully mitigated. While CPO is theoretically superior, if comprehensive adoption takes longer, the market must reprice these "not so end-game but easier to mass-produce and maintain" solutions.
Serenity's Rebuttal: NVIDIA May Accelerate Hardware Cycle
Serenity's rebuttal does not deny the engineering challenges of CPO but believes SemiAnalysis underestimated NVIDIA's organizational capability in the AI hardware cycle.
His logic is clear: While regular hardware adoption is indeed hindered by yield, reliability, and customer validation, NVIDIA is not a regular customer. It is both the GPU cluster architecture definer and a core driver of networking, switches, system integration, and supply chain cadence. When the AI factory's expansion is stalled by network power consumption and bandwidth limitations, NVIDIA has strong enough economic motivation and industry discourse power to compress the traditional adoption cycle.
Serenity's referenced evidence is two-fold. The first layer consists of publicly verifiable company statements, including NVIDIA's official information on Spectrum-X Photonics entering production and Lumentum's mention of CPO orders and delivery cadence in Q2 FY26. Lumentum stated it received a multi-billion dollar CPO incremental order with deliveries planned for the first half of 2027, and the company material also mentions expectations for a broader ramp-up of CPO-related businesses in the second half of 2026.
The second layer involves interpreting supply chain signals, such as Foxconn delivering optical switches ahead of schedule to NVIDIA. However, further public information is needed to confirm the specific scale of these signals, whether they are prototype testing or production-level orders.
This is also a key point of contention between Serenity and SemiAnalysis: SemiAnalysis believes more in the natural elongation of the system engineering variables, while Serenity believes more in NVIDIA's supply chain execution capability to steepen this curve.
These two assessments are not completely contradictory. NVIDIA can allow the CPO to enter production and customer validation earlier, and may also drive some horizontal expansion scenarios ahead, but this does not automatically mean that all AI data center networks will rapidly switch to CPO by 2027. Horizontal expansion, single rack expansion, intra-rack, inter-rack, different customer reliability tolerances, and cost models are all different, and the deployment pace may also be layered.
Serenity's rebuttal is not to prove that "CPO will be significantly delayed," but rather to argue against the overly conservative conclusion, not to demonstrate that "CPO is already completely risk-free." For the market, this is enough to support the oversold rebound logic, but not enough to directly rewrite the aggressive revenue curve for 2027-2028 back to determinism.
Why NPO Suddenly Matters
In this controversy, NPO has suddenly become important because it happens to stand between the logics of SemiAnalysis and Serenity.
It is neither the opposite of CPO, nor is it a simple extension of traditional pluggable modules. The fundamental idea of NPO is to place the optical engine on a pluggable pedestal-like board near the ASIC, shorten the electrical signal path, gain some power and density benefits, while retaining better testability, replaceability, and supply chain resilience.
If SemiAnalysis's conservative model is closer to reality, the deep packaging of CPO may slow down due to yield, maintenance, and reliability issues, and NPO will become a more viable option for a longer period. It allows hyperscale cloud operators to gradually accumulate experience in photonic interconnect operations without overly assuming the full CPO risk, and also provides a longer window for existing optical module and optical engine suppliers.
If Serenity's assessment of NVIDIA's execution capability is more accurate, NPO may not necessarily disappear. The more likely scenario is that NPO, CPO, pluggable, and copper interconnect coexist at different network levels. NVIDIA's roadmap itself also indicates that the horizontal expansion of CPO may take precedence, while some single rack expansion scenarios may still rely on copper cabling or hybrid architecture in 2027-2028.
The impact on investors is that they cannot simply price the photonic chain by using the "CPO wins, others lose" approach. Different technological paths correspond to different beneficiary segments: CPO is more favorable for the high-integration optical engine, laser source, silicon photonics, and switching chip ecosystem. The extension of the NPO and pluggable window may allow existing optical module vendors, connectors, materials, and some light source suppliers to continue enjoying order and margin support.
The market's previous issue was translating technological endpoint prematurely into a single-track performance slope. What is now being reopened is the valuation space of the middle track.
Production-Level Data is the Next Verification Point
This debate will not be resolved in the short term by a single report or a set of posts. SemiAnalysis reminds the market that the Chief Product Officer's (CPO) challenge lies in system engineering. Serenity reminds the market that NVIDIA's supply chain organizational capability may alter the traditional hardware adoption cadence. The true divergence between the two will rely on production-level data from the second half of 2026 to 2028 for validation.
The most crucial next step is not whether there is a shipment but the caliber of the shipment. Delivering to select partners, customer evaluations, initial production, scale-up, and mass deployment are entirely different phases. NVIDIA's subsequent descriptions of mass production for Spectrum-X / Quantum-X Photonics, and optical suppliers such as Lumentum and Coherent in their earnings reports regarding orders, capacity, and gross margin, will be more critical than one-time conference rhetoric.
Also crucial to observe are on-site reliability and maintenance data. If the CPO's failure rate in a production environment, replacement processes, yield curve, and total cost of ownership prove to be stable enough, SemiAnalysis's conservative model will be adjusted. If this data remains stuck at the laboratory or small-scale validation level, the NPO and pluggable module window will continue to be adjusted in the market's favor.
The optics chain is now trading not the life or death of CPO but the slope of the timetable. The next verification point will settle on whether "entering production" can transition into sustainable volume production and whether this volume will ultimately be reflected in the speed of orders, gross margins, and customer deployment caliber.
While SemiAnalysis has raised concerns about CPO technology over the next two years, they have still provided bullish views on five semiconductor subsectors:
Copper / AEC / ACC;
Pluggable Optics / DSP;
CPO Testing Equipment;
Power Grey Space / UPS Extension;
Board-Level VRM / Silicon-Based Power / Passives
The specific targets are integrated in the image below for readers to refer to
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia
