Leading Optical Module Player Shows Divergence Again! What's the Outlook for the Sector?

Published on 2026-08-10Last updated on 2026-08-10

Abstract

①U.S. stock optical module giant Lumentum announced strong Q3 fiscal year earnings, with revenue of $808.4 million, a year-on-year increase of 90.1%, slightly exceeding analyst expectations;<br>③The company's CEO stated that the demand for its optical module products far exceeds its supply capacity, and the company's order backlog extends to 2028.

On Friday (August 7th), the ChiNext AI index continued its rebound and recovery. AI application concept stock Chinese Online led gains with an increase of nearly 7%, while memory leaders Jiangbolong and Beijing Junzheng both rose over 5%. Leading optical module/CPO stocks saw significant intraday gains, with TFC Optical Communication closing up over 2%. However, Zhongji Innolight experienced a sudden downturn late in the session, closing down nearly 4%, with a daily fluctuation exceeding 10%.

Regarding popular ETFs, the ChiNext Artificial Intelligence ETF Huabao (159363), heavily weighted in leading optical module stocks, surged intraday by nearly 3%. It was briefly dragged into negative territory by the late-session plunge of the leading stock but quickly rebounded, ultimately closing up nearly 1%, successfully achieving five consecutive days of gains.

The afternoon downturn of Zhongji Innolight may be related to a piece of news. It is reported that U.S. optical module manufacturer AAOI (Applied Optoelectronics) revealed plans for large-scale capacity expansion in its financial report and conference call, sparking market concerns about industry competition.

During AAOI's conference call, it was indicated that the goal for this year is to increase monthly production capacity by threefold, requiring more automated equipment for expansion. AAOI plans to achieve a monthly output of 650,000 units of 800G/1.6T products by year-end, with a tenfold increase in capacity next year. This has raised concerns about its long-term potential to capture market share from domestic manufacturers. Although some analysis suggests its short-term impact is limited and it relies on technology from leading domestic companies, this negative rumor still triggered the stock price decline.

Focusing on the overall optical module sector, three key logics support mid-to-long-term prosperity:

1. Upward revisions in capital expenditures confirm industry prosperity; high growth in cloud revenue validates commercial implementation. CITIC Securities believes that against the backdrop of cloud companies' performance continuously confirming AI's driving role in business growth, AI cluster scale will further expand. As a crucial part of cluster networks, optical interconnection will sustain high-speed growth driven by three factors: increased GPU allocation ratio, port speed upgrades, and "optical replacing copper." They are optimistic about the mid-to-long-term prosperity of the optical communication sector.*

2. Continued emphasis on optical interconnection; CPO mass production accelerates cluster restructuring. A Kaiyuan Securities research report notes that NVIDIA's Rubin Ultra shifts focus towards large-scale interconnection at the NVL576 level, with the system configuration being the interconnection of eight 72-GPU racks using NPO for cross-rack connectivity. Meanwhile, CPO has entered mass production and will be heavily integrated into global AI factories in the second half of the year, expected to accelerate the penetration of optical interconnection technology and drive a structural revaluation of optical communication's value.*

3. Sufficient adjustments and undervalued valuations; clearing of positions lays a solid foundation for recovery. Guosheng Securities believes the previously overly crowded trading structure in the optical module sector is improving, with short-term risks largely released. Sufficient turnover and rebalancing of positions in the bottoming phase still require time. Subsequently, we need to wait for the gradual accumulation of favorable factors and the progressive restoration of market confidence.*

Within the main theme of AI trading, besides compute power positioning sectors like optical modules, AI applications are also a key focus. Referring to the U.S. SaaS benchmark, Palantir's quarterly report exceeded expectations, leading to a significant rally, indicating the market places a high premium on the ability to implement AI applications. The ChiNext Artificial Intelligence Index gathers many "software + hardware" combination stocks (e.g., vertical industry applications). Compared to pure hardware communications, these stocks benefit more from the dual logic of performance realization at the application layer and the reshaping of valuation systems.

The ChiNext Artificial Intelligence ETF Huabao (159363) and its off-exchange feeder funds (Class A: 023407, Class C: 023408) primarily focus on leading optical module/CPO players while also covering AI applications. The underlying index holds approximately 40% in "Zhongji Innolight + Eoptolink + TFC Optical Communication," making it a core standard-bearer for AI compute power. Furthermore, the latest AUM of the ChiNext Artificial Intelligence ETF Huabao (159363) exceeds 7.4 billion RMB, with an average daily turnover over the past six months exceeding 1 billion RMB, leading the eight ETFs tracking the same underlying index in terms of size and liquidity.

Data source: SSE, SZSE, etc.

*Reference sources for institutional views: CITIC Securities 'AI Investment Return Loop Appears, Firmly Optimistic About Optical Communication Sector'; Kaiyuan Securities 'Time for Optical Communication Configuration Has Arrived'; Guosheng Securities 'Optics: Panic Release, Adversity Leads to Good Fortune'

Explanation of related ETF fund fees: When subscribing for or redeeming fund shares, subscription/redemption agents may charge a commission not exceeding 0.5%. Fees for on-exchange trading are subject to the actual charges by securities companies; no sales service fee is charged.

Explanation of related feeder fund fees: Class C of the ChiNext Artificial Intelligence ETF Feeder Fund charges no subscription fee; redemption fee is 1.5% for holdings less than 7 days, and 0% for holdings 7 days (inclusive) or more; sales service fee is 0.3%. For Class A of the ChiNext Artificial Intelligence ETF Feeder Fund, subscription fee is 1% for amounts below 1 million RMB, 0.6% for 1 million (inclusive) to 2 million RMB, and 1000 RMB per transaction for 2 million RMB (inclusive) or more; redemption fee is 1.5% for holdings less than 7 days, and 0% for holdings 7 days (inclusive) or more; no sales service fee is charged.

Related Reads

Bill CLARITY Proposes Adding Requirements for Cryptoplatform Developers

A new amendment to the CLARITY bill proposes creating a category called "non-decentralized financial trading protocols." This includes individuals or groups with direct or indirect control over a protocol's functions, operations, or consensus rules. These entities would be required to register with the Commodity Futures Trading Commission (CFTC). The bill also mandates the CFTC and Treasury Department to develop specific rules and standards for these platforms. The stated goal is to distinguish truly decentralized projects from those that are effectively controlled, for example, through a majority of governance tokens or a project treasury. The amendment aims to resolve previous debates by clearly assigning compliance responsibilities, including for customer funds, to developers and governing communities even if they don't hold the assets directly. Additionally, the updated bill proposes legal protections for stablecoin issuers and exchanges that freeze assets suspected of being involved in illicit activity, shielding them from civil lawsuits demanding unfreezing. All crypto platforms would also be subject to bank-like regulations. The senator sponsoring the amendments hopes they will garner broader support for the CLARITY bill in an upcoming procedural Senate vote. The bill retains an "ethics provision" prohibiting government officials and their spouses from issuing, promoting, or engaging in crypto business, though a separate proposal could allow certain individuals, like former President Trump, a tax deferral on crypto asset sales.

cryptonews.ru17m ago

Bill CLARITY Proposes Adding Requirements for Cryptoplatform Developers

cryptonews.ru17m ago

Novo Holdings points to lack of investment in quantum applications

Novo Holdings reports that from 2014-2025, about 70% of the $13.9 billion in private investment in the quantum sector went to hardware and components. Investment in business applications remains insufficient relative to their potential value. By early September 2026, disclosed private funding for quantum companies had already surpassed the total for all of 2025. The report argues that while initial hardware investment was necessary, the next phase requires funding for specialized algorithms and services alongside computer development. These solutions should help customers with specific tasks like selecting molecules for research. Developers who create hard-to-replicate services—combining algorithms, proprietary data, and client integration—will gain an advantage. The life sciences, including pharmaceuticals, are identified as a highly promising field for early quantum applications. However, even early fault-tolerant computers are not expected to handle entire drug development workflows. Instead, they will tackle critical, small-scale tasks where classical approximations fail, such as calculating chemical bond behavior. More routine calculations will still be performed by classical computers. For a quantum service to be commercially valuable, its results must tangibly influence a client's next decision, like which experiment to run. Useful applications in materials science, energy, and chemistry may emerge earlier than in pharma. The report outlines two scenarios: early fault-tolerant systems in the late 2020s/early 2030s and more scalable systems from the mid-2030s. Key risks include progress in classical computing and AI potentially reducing quantum advantages, and the possibility of in-house development by pharmaceutical firms or cloud providers.

cryptonews.ru19m ago

Novo Holdings points to lack of investment in quantum applications

cryptonews.ru19m ago

U.S. Treasury's $5 Billion+ Buyback of 10-Year Bonds Failed to Stop Their Sell-Off

On September 10th, the U.S. Treasury, led by Scott Bessent, conducted a significantly larger-than-usual buyback of long-term government bonds, purchasing $5.2 billion worth. However, this intervention failed to halt a continued sell-off, with the yield on the benchmark 10-year Treasury note rising above 4.85% for the first time since 2023 and hitting nearly 4.98% the next day. Investors viewed the scale of the operation as insufficient for the massive $32 trillion U.S. debt market. They cited persistent pressures including rising oil prices, inflation risks, a large budget deficit, and expectations of further Federal Reserve rate hikes as reasons the buyback couldn't reverse the trend. The Treasury had announced it was prepared to buy up to $6 billion in long-term debt, triple its standard $2 billion operation. Yet, it only accepted $5.2 billion of the over $10 billion in investor offers. Market participants, including strategists from Bank of America and Amundi, criticized the move as too timid and potentially counterproductive, signaling government "nervousness" without addressing core issues. Key factors driving yields higher include the U.S. national debt exceeding $40 trillion, soaring debt servicing costs, strong new borrowing needs, and inflationary pressures from high energy prices. Additional concerns were raised by proposals for large new government spending. Analysts concluded the market is effectively "fighting the U.S. Treasury," with the intervention unable to overcome these fundamental headwinds.

cryptonews.ru20m ago

U.S. Treasury's $5 Billion+ Buyback of 10-Year Bonds Failed to Stop Their Sell-Off

cryptonews.ru20m ago

Fly Brain Successfully Made to Engage in Crypto Trading

Researchers have conducted an experiment using a digital connectome (brain map) of a fruit fly, the Drosophila MaleCNS v1.0, containing 166,700 neurons and 25.6 million synaptic connections. In this project, real-time Bitcoin ($BTC) to US Dollar Coin ($USDC) price data from Coinbase was converted into an RGB visual signal. This signal was fed into the fly brain model, with "buy" actions encoded as "sweet" and "sell" as "bitter." A fixed neural reader then translated the brain's activity into trading commands—buy, sell, or hold—which were routed through an intermediary service to place spot orders on Coinbase Advanced. The experiment ran in a safe, simulated environment with virtual $100, not real money. The programmer noted that real-money trading is technically possible by creating a limited portfolio (max $100 USDC) with restricted API keys. However, the fly's model would be capped at trades of $10 each, with a maximum of 24 trades per day, no leverage, no shorting, and trading halted after a $20 portfolio loss. The system incorporates a learning mechanism: profitable outcomes stimulate 15 dopamine-like neurons (PAM11), while losses activate two aversive neurons (PPL101), encouraging the model to avoid unprofitable trades. Despite this, the developer acknowledges that the fly brain has not yet demonstrated an ability to trade profitably; synaptic changes do not equate to a successful trading strategy. This remains a research experiment. In related news, a Solana-based bot recently turned $0.227 USDC into roughly $696,000 in seconds by capitalizing on a sharp price drop of the memecoin ANB.

cryptonews.ru21m ago

Fly Brain Successfully Made to Engage in Crypto Trading

cryptonews.ru21m ago

Trading

Spot
活动图片