On October 28th, Bloomberg reported that Marvell was in advanced negotiations with optical device manufacturer Lumentum considering acquiring the latter through a all-stock transaction. At the time of the report, Lumentum had a market value of approximately $83 billion, while Marvell was valued at around $221 billion.

This potential transaction raises a question worth considering: Why would a company known for its custom chips and high-speed interconnects venture into the manufacturing of lasers and optical devices on such a large scale?

On October 30th, the semiconductor industry podcast Semi Doped invited two analysts from market research firm Cignal AI, Andrew Schmitt and Scott Wilkinson, to discuss this transaction. Their judgment is that Marvell hopes to combine DSP, lasers, and optical components through vertical integration to form a more complete product system.

If the transaction is completed, Marvell may become one of the few companies that cover multiple aspects including DSP, lasers, and optical modules simultaneously. However, this will also introduce new competitive relationships: on one hand, it will supply chips to optical module manufacturers, and on the other hand, it may enter the markets of these customers.

I. From DSP to lasers and optical modules: Marvell aims to connect the entire chain

The host pointed out that after Marvell acquires Lumentum, it may become a company that owns both DSP and businesses in the fields of lasers and optical modules. Andrew Schmitt believes that the precedent closest to this model is Acacia, but the combination of Marvell and Lumentum may go even further in terms of business scale and coverage.

He first clarified that Marvell is not short of laser supplies. The company is currently mainly purchasing tunable lasers for modules related to ZR and maintains good relationships with multiple suppliers. The core driving force behind the acquisition of Lumentum is vertical integration.

Andrew believes that in a market where technology is changing rapidly, companies need to master both chip and optical technologies in order to better control product development and the supply chain. By having these two types of technology within the same company, collaborative design is possible, leading to the creation of products with improved performance, and also allowing for higher gross margins through the sale of more comprehensive solutions.

Lumentum already has a similar business portfolio. Andrew For example, the company previously acquired Cloud Light to enter the data communication optical module market based on silicon photonics technology. Cloud Light produces optical modules that use its own lasers, while also needing to purchase PAM4 DSP from companies such as Marvell and Broadcom.

In an optical module, the laser and DSP constitute almost the main value components, with the remaining processes mainly being assembly.

This means that there is already an upstream and downstream connection between Marvell and Lumentum. After the acquisition is completed, Marvell will have the opportunity to incorporate key components that previously needed to be purchased externally into its internal system, further integrating DSP, lasers, and optical engines.

Andrew believes that the value of this model is also reflected in profits.

In the past, chip companies could achieve a gross margin of up to 50%, while optical component manufacturers only had a margin of 25% to 30%. Naturally, chip companies were reluctant to acquire companies with even lower gross margins, fearing that it would reduce their own profit margins.

However, he stated that the gross profit margin of the optical components industry has now risen from around 25% to nearly 50%. Lumentum possesses rare technologies in products such as 200G EML, optical circuit switches (OCS), and high-power lasers, and its bargaining power has also significantly increased.

“What if you can earn more money, even though that amount accounts for a smaller proportion of your total income? So what?”

Andrew believes that investors have in the past placed too much emphasis on the gross margin percentage, neglecting the total gross profit that a company can generate. By integrating components into more complete products, even if the combined gross margin ratio changes, it is possible to create more profits.

He also mentioned that six or seven years ago, Marvell might have only needed about $3 billion to acquire Lumentum, but now the potential transaction size could be 30 times that of back then. These are his recollections and estimates, yet they also reflect the change in value of optical components during the AI era.

II. Acting as Both 'Referee' and 'Athlete': Marvell May Compete with Customers

Vertical integration will enable Marvell to have a more complete product line, and it will also change its relationship with customers.

Scott Wilkinson pointed out that the current coherent DSP technology of Marvell is widely adopted by many optical module manufacturers. Taking 400G coherent modules as an example, in addition to Acacia, Ciena, Nokia and some Chinese manufacturers, many companies also purchase the DSP from Marvell.

If Marvell acquires Lumentum, and further enters the optical engine and module market, it may compete directly with these existing customers.

This is precisely the subtle point in transactions: Marvell is not only a chip supplier at the upstream of the industrial chain but may also become a participant in the downstream products. Customers inevitably wonder whether they will continue to rely on a company that competes with them for providing key chips in the future.

Host Austin raised a direct question: If customers switch to other DSP suppliers such as Broadcom, Credo, etc. as a result, is Marvell willing to sacrifice some of its chip market share in exchange for the optical module business?

Andrew believes that this is a business deal that can be calculated accordingly:

"Through vertical integration, I can obtain more value than by selling components separately, even if that means I might sell fewer components."

Scott also reminds us that what really needs to be observed is how Marvell manages the business after the acquisition. Lumentum already sells lasers to external customers and also operates in the optical module business. After the acquisition is completed, Marvell may not keep all of this production capacity for itself.

Andrew believes that whether customers will churn depends on whether Marvell can give the existing DSP customers and Lumentum potential customers enough confidence, ensuring that their existing supply relationships will not be affected.

This is also the cost of vertical integration: companies gain more control over the industrial chain, but it may also force customers to look for new suppliers.

III. The "winner takes all" nature of optical communications: First-mover advantages are not as significant as delivering in the millions

In addition to vertical integration, the two analysts also discussed another distinct competitive pattern in the optical communications industry: the market for new technology standards is ultimately dominated by a few manufacturers.

Andrew refers to this phenomenon as “IEEE lottery.”

Whenever an industry introduces a new standard, usually five or six companies start developing products simultaneously. A few years later, only two companies tend to remain, occupying a combined 80% to 90% of the market share.

Therefore, while it is important to be the first to launch a product, whether one can seize the opportunities in the early stages of market development is what determines whether a company will become the ultimate winner.

Scott further pointed out that in the current AI optical communication market, competition particularly emphasizes production capacity.

"If you enter the market first but can only produce ten thousand units, it doesn't really make much sense. However, if you enter the market second but are able to produce one million units, then the orders will fall into your hands."

He took the Chinese manufacturer Eoptolink (Optics Valley Technologies) as an example. This company entered the 800G Ethernet optical module market a bit later than some of its competitors, yet it has become one of the global major suppliers due to its production capacity advantages. Scott believes that a similar competitive situation is also emerging for 1.6T products.

This is also why Lumentum has received attention. Scott mentioned that the company has been continuously expanding its indium phosphide (InP) manufacturing capacity over the past few years, with production capacity doubling several times, but market demand has grown even faster, and it is still difficult to meet the needs of all customers.

The advantage of Lumentum also lies in having products that are difficult for competitors to quickly replicate. Andrew specifically mentions 200G EML, OCS, and high-power lasers.

He reminded that many companies can claim to be able to develop similar products, but the real test is whether they can scale from producing a few hundred pieces to fifty thousand pieces, or even one million pieces. "Making one hundred pieces is completely different from making one million pieces."

In the AI era, this scale advantage is further amplified. Andrew believes that the value of optical components has also changed: in traditional telecommunications networks, lasers mainly help operators increase transmission capacity; in AI data centers, 1.6T interconnection may increase the training speed of AI clusters by two to three times.

When an optical technology is able to significantly improve the efficiency of an entire computing cluster, the price that customers are willing to pay for it naturally changes.

IV. Will Nvidia join the bidding?

At the end of the interview, the two analysts discussed whether Nvidia could become a potential buyer.

Nvidia has previously invested $2 billion in Lumentum and Coherent respectively, and established strategic partnerships with both companies. Andrew believes that Nvidia also has the motivation for vertical integration: the previous acquisition of Mellanox was aimed at integrating network technology into its own GPU system. Continuing to expand into the fields of optical components and OCS also aligns with this strategy.

Andrew also proposes another possibility: if the negotiation messages between Marvell and Lumentum are deliberately leaked, it might be a signal sent to other potential buyers, ultimately leading to a bidding war.

However, he clearly stated that he does not have any insider information regarding the transaction; this is merely his personal speculation. The previous investment and procurement collaborations of Nvidia also do not prove that it intended to participate in the acquisition.

For Marvell, the core challenge of this transaction ultimately lies in balancing two things: achieving higher overall value by integrating chip and optical technologies, while at the same time preventing existing customers from turning to other suppliers due to competitive factors.

The judgment of Andrew Schmitt is that optical components have become a more valuable part of the infrastructure of AI, and the economic logic of chip companies acquiring optical enterprises is changing. Scott Wilkinson reminds us that in the field of optical communications, where a 'winner takes all' situation is common, production capacity and scale determine the final market share.

Marvell aims to establish a complete industrial chain from chips to optics. Whether they can simultaneously retain customers, technology, and market share will determine how much value this integration can ultimately create.