SemiAnalysis report says delays in two key technologies trigger a sharp decline in "optoelectronic" stocks, netizens fiercely debate CPO
A report by AI industry chain star analysis agency SemiAnalysis, which directly points to delays in two core technological paths for AI data centers, triggered severe volatility in the optical communications sector on June 10, and sparked intense debate about future technological routes and investment opportunities among investors and industry insiders.
The report asserts that NVIDIA’s 800VDC power architecture shipments will be delayed until 2028, and scalable mass production of CPO (Co-packaged Optics) may be postponed to 2028 or even 2029, with both expectations downgraded simultaneously, catching the market off guard.
After the news was released, the US optical communications sector generally plunged. Applied Optoelectronics (AAOI) fell as much as 17% in a single day; Lumentum dropped roughly 8%; Himax Technologies (HIMX), Navitas Semiconductor Corp, Wolfspeed, and other companies cited in the report as warranting caution also came under obvious pressure.
At the time of the SemiAnalysis report's release, interviews with NVIDIA executives were also disclosed. According to veteran semiconductor and tech investment journalist Tae Kim, Gilad Shainer, Senior VP of Networking at NVIDIA, expressed sharply contrasting optimism about CPO at Computex 2026, stating “CPO is currently the most exciting technology,” and saying mass shipment would begin in the second half of the year, thus igniting a fierce debate about the CPO timeline on social platforms.
It is noteworthy that multiple market observers pointed out that a delay in CPO does not mean the disappearance of demand for optical interconnects, but more likely redirects capital flows to traditional pluggable optical modules and NPO (Near Package Optics) tracks—this logic led some investors to seek buying opportunities amid panic selling.
Core of SemiAnalysis Report: Dual Delay in Two Major Technologies
In this research note sent to institutional clients, SemiAnalysis delivered two core judgments with far-reaching market implications.
800VDC power architecture delayed until after 2028.
The report indicates that NVIDIA's original plan for large-scale adoption of single-ended 800VDC power design has been notably pushed back. Hyperscalers currently prefer using mature low-voltage solutions or gradually transitioning to 400VDC, rather than switching hastily to 800VDC.
The report argues that the marginal efficiency gains from 800VDC under current power grid conditions do not justify its engineering complexity. In contrast, 400VDC products are expected to start ramping up in Q2 2026, with notable growth in 2027.
CPO mass production markedly lags market expectations.
The report states that CPO shipment volumes in 2027 will be significantly lower than previous aggressive forecasts, with mass production possibly postponed to 2028 or even 2029. Major bottlenecks are concentrated on three levels:
Optical engine connection yield rate (optimistically about 95%, but CPO production per single ASIC remains extremely limited), ASIC integration difficulty, and overall cost economics.
Scale-out CPO switch shipment volumes face downgrading risks; the Sidecar shipments dependent on new platforms like Rubin Ultra/Kyber are also postponed to the 2028 window.
On individual stocks, SemiAnalysis maintains relatively positive views on Amphenol, Vertiv, Legrand, etc., while remaining cautious toward Lumentum, Himax Technologies, Navitas Semiconductor, and Wolfspeed.
However, the report itself also acknowledges that CPO as an important future direction for data center networks is not being denied; the key reason for the delay is engineering challenges that haven’t been fully overcome, rather than a disappearance of demand.
Meanwhile, the report also notes some NPO (Near Package Optics) projects may accelerate.
NVIDIA Executive Publicly Contradicts, Tae Kim Interview Draws Attention
While the SemiAnalysis report circulated widely among institutions, veteran semiconductor and tech journalist Tae Kim published a transcript of an exclusive interview with NVIDIA Networking SVP Gilad Shainer during Computex in his Substack column, offering a stark contrast to the SemiAnalysis positions.
Shainer said in the interview, “The most exciting thing today is co-packaged optics, which is the most advanced frontier in technology.”
He further revealed that NVIDIA is ready to start shipments; partner Lambda has published a blog confirming receipt of the CPO switch, ramping up shipments in the second half of the year, and extending from scale-out to scale-up scenarios. “If it were up to me, I’d want all places using optical networks to adopt CPO.”
Tae Kim added in his article that Shainer’s overall demeanor and body language in the interview displayed strong enthusiasm for both near-term and long-term ramp of CPO, stating that “This attitude seems to directly contradict the narrative of SemiAnalysis.”
This contrast plunged the market into an information war. User @qinbafrank on social platform X pointed out that Bernstein's report as early as mid-May stated explicitly that cloud vendors would not sacrifice system reliability for energy savings, and currently no cloud vendor plans large-scale deployment of CPO in 2026–2027, “If you looked at Bernstein’s report in detail, you wouldn’t be surprised by today's SemiAnalysis.”
Netizens Fiercely Debate: Is CPO Delay Bad News or Wrong Sell-Off?
The market turmoil triggered by the report quickly spilled onto social media, with sharply divided opinions on the investment logic around CPO delays.
Bearish: Yield rate and reliability are real bottlenecks.
The SemiAnalysis report emphasizes that under CPO architecture, optical engines are packaged together with large, valuable ASICs on the same substrate. If the optical engine fails due to laser aging or fiber damage, the entire board must often be removed and returned to the factory. Maintenance costs and downtime risks are much higher than traditional pluggable modules, regarded as the main barrier to large-scale CPO adoption in the near term.
Bullish: CPO delay instead benefits pluggable modules and NPO.
User @TomSzczypka on platform X analyzed, “If CPO comes late, data still needs to be transmitted—the AI clusters can’t wait two years. Hyperscalers will buy more pluggable modules and NPO for longer; the money doesn’t disappear, it just changes pockets.”
He also noted that AAOI’s single-day decline (17%) far exceeded Lumentum (8%), which shows the day’s sell-off was not rational analysis but rather a washout of the weakest holders.
User @michaelsikand added that any photonics company currently generates zero revenue from CPO, and today’s high growth comes from the huge, unmet NPO opportunities: “The timeline may be delayed, but TAM won’t be.”
Voices questioning the logic of the report also exist.
User @cherryPayment posted a long-form analysis, pointing out internal contradictions: the report claims the supply chain is unprepared for 2027, yet predicts Celestial AI (acquired by Marvell) will reach $1 billion in revenue run rate by the end of 2028, and Amazon has signed a contract for Trainium 4, “You can’t suddenly hit $1 billion at the end of 2028 if the supply chain is completely unprepared in 2027.”
He also noted the target audience of SemiAnalysis is the procurement decision-makers at hyperscalers, and their conclusion is “no need to all-in now,” rather than timing the capital markets: “They analyze deployment pace, not investment timing.”
X user @Herman Jin criticized the timeliness of US investment research information, suggesting the delay of CPO and 800VDC “is a matter of time,” and that this news has long circulated in institutional circles, with the SemiAnalysis report just putting known information on paper.
Unexpected Beneficiaries: Copper Connections and Pluggable Modules
With the market under broad pressure, some analysts turned their attention to potential beneficiaries of the CPO delay.
User @qinbafrank summarized that the most realistic revenue opportunity in 2026 is concentrated in 1.6T pluggable modules, LPO/NPO, light sources, testing, PCB, ABF, and CCL, etc. “Optics won’t eliminate copper overnight; copper won’t guard every scenario forever. Different distances and system levels will choose different solutions.”
Lumentum’s CEO recently said that non-NVIDIA customer interest in NPO has noticeably increased in the past two months.
User @RealNickMugalli explained that at 1.6T rates and 200G per channel, copper cables—even with retimer tech—have hit a physical limit, so optical schemes will become mandatory rather than optional at reasonable distances. The NPO potential market could even exceed CPO’s.
SemiAnalysis also pointed out that some NPO projects may accelerate, and 400VDC products will start ramping in Q2 2026. For Amphenol and Vertiv, the report is relatively positive, believing they will benefit from sustained demand during the 400VDC transition.
User @TomSzczypka cited this week’s supply chain data to confirm AI infrastructure demand hasn’t weakened:
Fujikura raised datacenter cable prices because nearly all US hyperscalers placed orders simultaneously; King Slide’s rack rail revenues rose 47% QoQ; Google ordered 6 million TPUs from Intel; SK Hynix signed a multi-year memory supply deal with NVIDIA, etc.
“The real AI bottleneck is power, storage, and GPU—none of these has gotten worse today.”
Meanwhile, @tuolaji2024 posted on X that memory (HBM/DRAM) as the real physical bottleneck, its supply shortage situation is totally unaffected by this technical delay event.
Analysis shows that, synthesizing various views, the market fluctuation caused by this SemiAnalysis report reflects more a recalibration of technology route timelines, rather than a fundamental reversal of overall AI data center demand.
Risk warning and disclaimerThe market has risks, investment must be cautious. This article does not constitute personal investment advice, nor does it take into account individual users’ particular investment objectives, financial situation, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article fit their specific situation. Investments based on this are at your own risk.