2026-11-15 15:26

Lead-in: NVIDIA bets on optical interconnect, yet market remains skeptical about dilution and consolidation. Despite the apparent sincerity of this "powerful alliance," secondary markets are not buying it.

For this seemingly sincere "powerful alliance", secondary markets are unimpressed.

Word count: 3229

Handcrafted Volume | 90% AI Content | 10%

Recently, U.S.-based semiconductor company Marvell (Chinese name: Meiman Technology) announced it has signed a definitive agreement to acquire optical communications device manufacturer Lumentum (Chinese name: Langmeitong) via an all-stock transaction, with an implied enterprise value of approximately $115 billion, equivalent to RMB 76.71 billion.

This is undoubtedly a deal capable of triggering a global tech industry merger wave.

First, in terms of scale, $115 billion surpasses Microsoft’s $69 billion acquisition of Activision Blizzard, making it one of the largest tech M&A deals in history.

Second, both companies are dominant players in their respective high-end DSP and optical chip markets, each holding over 50% market share—this is a textbook “powerful alliance.”

Notably, NVIDIA, currently at the center of controversy, is also involved behind the scenes: early this year, NVIDIA invested $20 billion in each of the two companies. Additionally, BlackRock and Vanguard are major institutional shareholders common to both firms.

Premium of 23.9%, yet secondary markets remain unconvinced

Let’s first examine the two parties involved.

Marvell is a fabless semiconductor design company founded in 1995 and listed on NASDAQ in 2000. Its current business spans electrical interconnects, network switching chips, custom AI acceleration chips, and advanced semiconductor technologies.

According to public data, Marvell holds a 60%–70% market share in the global high-end DSP segment—a clear industry leader. As of the closing price on November 2 (the announcement date), the company’s market capitalization stood at approximately $172.2 billion.

Lumentum, the target of Marvell’s acquisition, is no less formidable. Originally a division of global optoelectronics giant JDSU, Lumentum was spun off independently in 2015 and listed on NASDAQ the same year.

Lumentum’s strongest asset lies in its manufacturing capabilities for indium phosphide (InP) compound semiconductors, along with its portfolio of optical transceivers, high-speed lasers, and optical circuit switches (OCS). In AI data centers, these components convert electrical signals into optical signals and vice versa. Data shows that Lumentum currently commands a 50%–60% global market share in EML (Electro-Absorption Modulated Laser) chips.

Notably, Lumentum’s latest financial performance coincides precisely with the explosion of AI-driven optical interconnect demand. For the first fiscal quarter of FY2027 ending September 26, 2026, revenue reached $1.263 billion, up 136.5% year-over-year; net profit was $360 million, compared to just $4.2 million in the same period last year, with non-GAAP net profit reaching $461 million.

Moreover, the company disclosed that its second large-scale NPO/CPO (Near Package Optics / Co-Packaged Optics) order has already been secured, with buyer commitments exceeding $5 billion. Reports also indicate that Lumentum’s order backlog has extended as far as 2029, with some clients even locking in orders for 2030.

Given that Lumentum is a rising star in the laser industry, Marvell has shown substantial commitment. According to the announcement, each shareholder of Lumentum will receive 4.5 shares of Marvell stock per share held, equivalent to $1,132 per share—23.9% higher than Lumentum’s closing price on October 28, prior to any rumors.

Meanwhile, to secure the deal, both parties have set a substantial breakup fee: if the transaction collapses due to regulatory reasons, Marvell must pay Lumentum a $3.2 billion regulatory termination fee; under specific circumstances, this fee could rise to $5.8 billion.

Yet ironically, despite the apparent sincerity of this “powerful alliance,” secondary markets are not responding favorably.

On the day the transaction was announced, Marvell dropped 12.72%, while Lumentum fell 18.59%. The selling pressure continued in the following days. By the close of November 10, Marvell traded at $154.39, and Lumentum at $585.57—over seven trading days, both companies had cumulatively declined by 20.82% and 31.16%, respectively.

Betting on AI Optical Interconnect: BlackRock and NVIDIA as Shared Shareholders

To understand why Marvell is willing to pay a premium to acquire Lumentum, we must return to the structural bottlenecks facing AI today.

In training and inference processes, massive data must be exchanged between chips, servers, switches, and racks—larger clusters mean more frequent data transfers. When bandwidth is insufficient or latency is too high, systems fall into idle “waiting for data” states. With ten-thousand-card clusters becoming standard and hundred-thousand-card clusters entering the horizon, the physical limits of traditional copper-based electrical interconnects in transmission distance, power consumption control, and bandwidth density are rapidly being exposed.

Thus, high-integration, high-performance optical interconnect technology has emerged as a core necessity for upgrading computing power infrastructure.

From a technical perspective, pluggable optical modules remain the current mainstream, but industry focus has clearly shifted toward more chip-proximate solutions such as NPO, CPO, and OIO.

This is precisely where Marvell finds itself in a delicate position. During the era of pluggable optical modules, DSPs were among the highest-cost and most technically challenging chips in the module stack, and Marvell held a leading share in this domain. However, as technology evolves toward NPO/CPO and eventually OIO, certain architectures may drastically reduce reliance on standalone DSPs. Conversely, Lumentum possesses critical capabilities in continuous-wave (CW) light sources and indium phosphide (InP)-based high-speed lasers—core components in NPO/CPO designs.

Therefore, for Marvell, acquiring Lumentum secures its legacy advantage in the pluggable era while simultaneously securing a strategic entry point into the CPO future.

Additionally, there is another noteworthy detail: both companies share several prominent institutional shareholders.

According to data aggregated by Yahoo Finance from Form 13F filings, as of June 2026, BlackRock held approximately 9.8% of Marvell and 10.1% of Lumentum; Vanguard’s two entities collectively owned about 9.0% of Marvell and 10.1% of Lumentum—both firms rank among the top institutional shareholders in each company.

Furthermore, NVIDIA maintains capital and business ties with both companies: first, on March 2, NVIDIA invested $2 billion in Lumentum, which the latter used to build a new wafer fab in the U.S.; less than a month later, NVIDIA announced a $2 billion investment in Marvell, with plans to collaborate on AI data center infrastructure, integrating Marvell’s products into NVIDIA’s AI ecosystem—including the NVLink Fusion interconnect platform.

While it remains unclear whether these shared shareholders played a behind-the-scenes role in driving this transaction, at minimum, their interests align with the deal’s success.

The wave of optical interconnect M&A has only just begun

In fact, Marvell is far from alone in using acquisitions to expand its optical interconnect footprint.

Globally, the history of optical interconnect M&A spans nearly a decade. For instance, one of the key players in this deal, Lumentum, acquired Oclaro for $1.8 billion in 2018, filling gaps in InP laser and photonic integrated circuit capabilities. In 2019, Cisco attempted to acquire Acacia, a leader in coherent optical DSP, and the deal ultimately closed in 2021 for $4.5 billion. In 2022, II-VI completed its ~$7 billion acquisition of Coherent.

China is no exception. In 2017, Zhongji Equipment acquired Suzhou Xuchuang for 2.8 billion RMB, rebranded as Zhongji Xuchuang, transforming from a motor equipment firm into a global leader in high-speed optical modules.

Another industry leader, Dongshan Precision, invested no more than 5.935 billion RMB in 2025 to acquire 100% equity in SORLIS Photonics and subscribe to convertible bonds. Dongshan Precision, originally focused on consumer electronics and new energy vehicles, committed nearly 6 billion RMB to acquire SORLIS primarily due to its expertise in InP laser chip design and manufacturing, as well as mass production capabilities in 400G/800G optical modules.

Beyond M&A, venture capital investment represents another parallel strategy for market positioning. Fueled by the structural shift toward “optics replacing copper” in AI data centers, numerous on-chain enterprises are securing early positions through industrial capital.

A prime example is Huawei, which through its Hubble Investment has backed companies like Yuanjie Tech, Changguang Huaxin, Zonghui Xinguang, Milight Semiconductor, and Yunnan Xinyao—key players in optical chips and materials. Meanwhile, Alibaba appears on the shareholder lists of Hanguang Xinzheng and Xizhi Tech; Tencent is a shareholder in Xizhi Tech; Xiaomi has invested in Hanguang Xinzheng.

On the other side, A-share listed companies are leveraging Corporate Venture Capital (CVC) to stake claims: Gongjin Shares injected 50 million RMB to increase capital in silicon photonics firm Hongguang Xiangshang, securing a 7.06% stake; Sanan Optoelectronics appeared in the first round of funding for Micro LED optical interconnect firm Xinghuochen Guang; Dahan Technology spent 111 million RMB to acquire a 5% stake in Zhongshi Guangxin from Yue Ling Shares.

When viewed collectively, although the strategies differ, the direction is remarkably consistent: using capital operations to integrate fragmented capabilities across lasers, optical chips, DSPs, optical modules, and packaging into larger platforms. This reflects a broader shift: as AI elevates optical interconnects from “supporting actors” to “main leads,” competition is transitioning from “single-point technology” to “full-stack capability.” Thus, the future winners are unlikely to be a single laser company or a single DSP firm—but rather platform companies capable of seamlessly integrating electrical, optical, packaging, and software elements into a unified system.

Original content by Chuanzhong Network

Source: Chuanzhong Network