The recent volatility in optical sector equities stems not from a rejection of Co-Packaged Optics (CPO) technology, but from a critical reassessment of its commercialization timeline. The market is currently recalibrating whether the 2027-2028 window represents a definitive execution phase or merely an extended period of technical validation. While the fundamental drivers for CPO—rising bandwidth demands, power consumption constraints, and switching density pressures in AI data centers—remain intact, the physical limitations of copper cabling and traditional pluggable optics have not vanished. The core friction lies in the aggressive pricing assumptions embedded in recent stock run-ups, which anticipated a rapid volume production ramp for optical engines, lasers, silicon photonics, and switch chips immediately following NVIDIA's push into the commercialization window.

A pivotal moment occurred on June 9 when a report from SemiAnalysis challenged these aggressive timelines. The analysis suggested that NVIDIA's 800V DC and large-scale CPO production could be delayed until 2028-2029, with 400V DC remaining on track for 2026, while some Near-Packaged Optics (NPO) projects might accelerate. This assessment triggered immediate market turbulence, causing high-single to double-digit pullbacks in tickers such as AAOI, LITE, COHR, GLW, and MRVL. Data compiled by Woofun AI indicates that the market is no longer trading on the binary validity of the CPO direction but is instead pricing the velocity at which CPO can convert into tangible orders. The divergence highlights a shift from an endgame narrative to a timeline validation framework.

Conversely, the bearish sentiment faced a direct rebuttal from Serenity, an AI supply chain analyst known as the 'White-Haired Stock God.' Serenity argued that the SemiAnalysis report relies too heavily on conservative engineering models and underestimates NVIDIA's capacity to compress hardware cycles. Drawing on signals from NVIDIA, Lumentum, and Foxconn, Serenity maintains that CPO remains on track for deployment in the second half of 2026, the second half of 2027, and 2028. Woofun AI notes that this debate does not seek a definitive winner but rather shifts the valuation anchor of the optical chain, determining how value distributes among NPO, pluggable modules, light sources, and switch chips based on the arrival slope.

The market's recent capital allocation has been driven by expectations of a capital expenditure shift in next-gen AI data center network architecture rather than current revenue. As model training and inference clusters expand, communication pressure between GPUs, racks, and within data centers has intensified, turning the network into a potential efficiency bottleneck. CPO offers a theoretical solution by placing the optical engine closer to the ASIC, shortening high-speed electrical signal paths, and reducing power consumption and signal integrity issues associated with serial-to-parallel conversion circuits.

However, the market frequently mistakes the 'right direction' for 'volume certainty,' overlooking the significant valuation gap between 'going into production,' 'shipping to select partners,' and 'massive production.'

NVIDIA's official communications confirm that the Vera Rubin platform will introduce Spectrum-X Ethernet Photonics, with CPO switches already in production for lateral expansion and cross-cluster deployment. Networking executives mentioned in a June 3 report that Spectrum-X CPO switches have been shipped to some partners, with capacity expected to ramp in the second half of 2026. While these signals validate technological progress, they do not guarantee risk-free realization of large-scale orders. SemiAnalysis posits that the scaling implementation is slower than imagined because CPO compresses complexity from dispersed modules into a deeply coupled system, raising pressures on manufacturing, testing, maintenance, and supply chain flexibility.

The maintenance radius of CPO presents a distinct challenge compared to traditional pluggable modules. In a pluggable architecture, a failed optical module can be easily replaced, and vendor switches are straightforward. In CPO, the optical engine is integrated into the same packaging system as the ASIC; a failure may impact higher-value switching chips and entire machine systems. SemiAnalysis emphasizes that for hyperscale cloud providers, performance is not the sole metric; reliability, serviceability, yield, and supply chain maturity are critical. Woofun AI analysis suggests that if failure rates and replacement costs in production environments remain uncontrollable, even superior power consumption models will face deployment delays.

Serenity's counter-argument rests on the premise that NVIDIA is not a regular customer but the definer of GPU cluster architecture and a core driver of networking and supply chain cadence. With strong economic motivation to overcome network bottlenecks, NVIDIA possesses the industry discourse power to compress traditional adoption cycles. Evidence cited includes Lumentum's announcement of a multi-billion dollar CPO incremental order with deliveries planned for the first half of 2027, alongside expectations for a broader ramp-up in the second half of 2026.

Additionally, supply chain signals such as Foxconn delivering optical switches ahead of schedule support the view that execution capabilities may steepen the adoption curve.

The resolution of this debate may hinge on the role of NPO, which sits between the conservative engineering model and the aggressive execution thesis. NPO places the optical engine on a pluggable pedestal near the ASIC, shortening electrical paths while retaining better testability and replaceability. If SemiAnalysis's concerns regarding yield and maintenance prove accurate, NPO could become a viable intermediate solution for a longer period, allowing operators to accumulate experience without assuming full CPO risk. Alternatively, if Serenity's assessment holds, NPO, CPO, pluggable, and copper interconnects may coexist across different network levels, with CPO prioritizing horizontal expansion while single-rack scenarios rely on hybrid architectures.

Investors must now abandon the 'CPO wins, others lose' binary pricing model. Different technological paths favor distinct beneficiary segments: CPO benefits high-integration optical engines and silicon photonics, while an extended NPO and pluggable window supports existing module vendors and connector suppliers. The market is reopening the valuation space of the middle track, recognizing that the deployment pace will be layered across different customer reliability tolerances and cost models. The true divergence between the two assessments will rely on production-level data from the second half of 2026 to 2028.

The most crucial next step involves analyzing the caliber of shipments rather than mere existence. Delivering to select partners, customer evaluations, initial production, scale-up, and mass deployment represent entirely different phases. Future validation will depend on NVIDIA's descriptions of mass production for Spectrum-X and Quantum-X Photonics, as well as earnings reports from optical suppliers like Lumentum and Coherent regarding orders, capacity, and gross margins. On-site reliability and maintenance data will be equally critical; if CPO failure rates and total cost of ownership prove stable, the conservative model will adjust. If data remains stuck at laboratory levels, the NPO and pluggable module window will continue to expand.

Despite concerns over CPO technology over the next two years, SemiAnalysis maintains bullish views on five semiconductor subsectors: Copper/AEC/ACC, Pluggable Optics/DSP, CPO Testing Equipment, Power Grey Space/UPS Extension, and Board-Level VRM/Silicon-Based Power/Passives. The optics chain is now trading the slope of the timetable rather than the life or death of CPO. The next verification point will determine whether 'entering production' transitions into sustainable volume production and whether this volume reflects in the speed of orders, gross margins, and customer deployment caliber.