The biggest winner at yesterday's Taipei Computex event was MRVL.

Jensen Huang has poured over $6.5 billion into a bunch of optical companies in just three months, including Marvell (MRVL). At the event, he stated that Marvell will be the 'next trillion-dollar company'. MRVL skyrocketed by 32.52% in a single day, with trading volume exceeding 100 million shares, marking the largest single-day gain in the company's history.

Alongside this, the entire optical module sector has been set ablaze: Coherent surged 17.3%, Lumentum jumped 13.3%, Corning rose nearly 12%, Ciena climbed over 8%, and Nokia also joined the rally.

Regarding the logic of optical modules, I've actually done a deep dive analysis on this. In an article from two weeks ago (AI Era, the 'Great Famine' Moment for Fiber and Copper), I broke down the underlying logic of the physical limits of copper cables, the fiber shortage, and the interconnection demand for AI data centers very clearly. The amount of fiber needed for AI clusters is 36 times that of traditional CPU racks, and the production cycle for optical rods is 18 to 24 months, with rigid supply constraints that have no short-term solutions.

MRVL surged 32% in a single day to close at $290, with a market cap exceeding $250 billion; Nokia's AI optical network logic is also being increasingly recognized, starting to become a frequently mentioned mainline target. These two names seem to have already become the representatives of optical modules.

However, the editor wants to delve deeper and break down the optical module industry chain by segments, identifying those targets that have not been fully priced by the market but also hold significant power in their respective choke points. Many investors only focus on the 800G and 1.6T optical modules themselves, but the entire industry chain can be divided into at least six sub-segments and levels, from which we will select some targets in each sub-segment.

Thus helping readers find which targets in the U.S. stock market's optical modules, besides MRVL and NOK, are truly worth paying attention to?

The methodology behind this stock selection is inspired by 'Wall Street's new stock god' Serenity, who is currently the most prominent analyst in the optical communication and AI supply chain 'bottleneck theory,' known for his 'chokepoint' logic, specifically digging up the invisible bottlenecks upstream of AI data center optical interconnects, rather than directly chasing large-cap stocks like Nvidia or Broadcom.

Optical fiber materials: GLW (Corning)

This layer is the foundation of the entire industry chain and the hardest to replace. The core raw material for optical fibers is called the preform, with an expansion cycle lasting 18 to 24 months, and the process is extremely complex. The amount of optical fiber required for an AI data center is 36 times that of a traditional CPU rack, with Meta's Hyperion data center alone requiring 8 million miles of optical fiber.

GLW

Corning is the global leader in the optical fiber market, an absolute oligopoly. GLW currently has a market cap of about $150 billion, with a stock price of $197.

Founded in 1851, the inventor of low-loss optical fiber, no other company can replace its production capacity in the near term.

The core of the moat: the manufacturing of preforms is a highly integrated process of 'materials science + precision manufacturing,' with yield accumulations counted in decades. Corning has accumulated over 40 years in this process. The net profit from the optical communication business is expected to grow by 28% year-on-year in 2024, by 71% in 2025, and a staggering 93% year-on-year growth in Q1 2026, with the growth rate still accelerating.

On May 6, 2026, Nvidia announced a multi-year strategic cooperation with Corning, with a maximum total investment in warrants of up to $3.2 billion, requiring Corning to increase its optical connectivity manufacturing capacity in the U.S. by tenfold and expand optical fiber production by over 50%, along with the construction of three new factories. This is Nvidia's largest single bet in the entire optical supply chain.

Over the past 12 months, the stock price has risen by over 315%. Serenity has included Corning in his 'material oil' supply chain bottleneck map, positioning it as a stable core holding.

Indium phosphide substrate: AXT (AXT)

This is the most undervalued part of the entire industry chain and also the most famous call from Serenity.

Why is indium phosphide (InP) so critical? Because silicon cannot emit light. All lasers in optical modules require InP as the substrate material. Nvidia's optical connectivity chips, all high-speed transceivers, and the optical engines in CPO are fundamentally dependent on InP wafers. Even more critically: the growth process of InP single crystals is an order of magnitude more complex than that of silicon, with an expansion cycle often exceeding two years, making it impossible to solve quickly with capital alone.

In 2025, global InP substrate shipments are expected to be around 600,000 to 700,000 pieces, while actual market demand is 1.5 to 2 million pieces. The supply gap exceeds 70%, and it cannot be eliminated in the short term.

The dynamics at this layer are extremely concentrated, with three companies controlling 80-90% of the global market share: Japan's Sumitomo Electric, Japan's JX Metals, and America's AXT.

AXTI

AXT controls 60-70% of the global InP substrate output, making it the largest single supplier in this oligopolistic market. AXT currently has a market cap of about $4-5 billion, with a stock price of $141.

This is Serenity's breakout work and one of the most frequently discussed targets (with over 70 public tweets). His core assertion about AXTI is succinct: 'The entire AI industry's construction relies on this company, including all players such as Google, Nvidia, and Microsoft; they all depend on its indium phosphide substrate. If there is no indium phosphide substrate, the entire story of AI optical interconnect will end in 2026.'

AXT is the largest single supplier in the global InP substrate market, with a client list including Google, Nvidia, Microsoft, and all major optical module manufacturers. In the Q1 2026 report, revenue reached $26.9 million, a year-on-year increase of 39%, with InP business revenue reaching $13.6 million. The company has just completed $632.5 million in financing, specifically to expand InP capacity, aiming to double it in both 2026 and 2027. Order demand is expected to grow about 2 times in 2026 and double again in 2027.

Serenity initially built his position around $12, targeting a price of $150. As of writing, it has approached $141, with floating profits exceeding 1000%. This trade most clearly validates his methodology: small market cap, a client list full of trillion-dollar giants, and its products being the only choice upstream; once such a company is 'discovered' by the market, its elasticity is several times that of large-cap companies.

However, the risk to note is that AXT's production lines are mainly in China, creating exposure to geopolitical supply chain risks, which is a core risk variable.

Lasers and optical chips: LITE, COHR, MTSI

This is the manufacturing segment in the optical module industry chain with the highest technical barriers. What is truly scarce in the AI era is not the optical module as a whole but the chips that generate optical signals themselves, namely EML (electro-absorption modulated laser) and CW Laser (continuous wave laser). A 1.6T optical module requires 200G/lane EML, and globally, only one company is mass-producing such chips.

LITE

Lumentum is the only company globally producing 200G/lane EML at scale; no second company is at this technological node. LITE currently has a market cap of about $64 billion, with a stock price of $1,025.

This is the single point monopoly with the highest technical barrier in the current optical module sector. Nvidia's large-scale pre-purchase of this capacity has pushed delivery times to after 2027, triggering supply tightness across the industry. Serenity lists LITE as the core benchmark of the optical wave narrative, discussing its dual role in pluggable modules and CPO multiple times.

In March 2026, Nvidia invested $2 billion with several billion dollars in long-term procurement commitments. OCS (Optical Circuit Switch) backlog orders exceed $400 million, with additional CPO orders in the hundreds of millions, slated for delivery in the first half of 2027. FY2026 Q2 revenue reached $665.5 million, a year-on-year increase of 66%, with the company guiding for over 85% year-on-year growth in the next quarter. The Greensboro wafer factory will be fully operational by early 2028, ultimately supporting an annualized capacity of $5 billion. On Computex day, shares surged 13.3% to about $1025. Rothschild's target price is $1270, Jefferies $1200, JPMorgan $1130.

COHR

Coherent is the company with the highest degree of vertical integration in the optical device industry chain, owning the entire production line from InP raw materials to lasers, packaging, and modules. COHR currently has a market cap of about $66-73 billion, with a stock price of $427.

Originally founded in 1971 as II-VI Incorporated, it was renamed after acquiring Coherent in 2022. Coherent's moat comes from doing what no other company can: owning the entire production line from InP raw materials to lasers, packaging, and modules. It is one of the largest InP system manufacturers globally, with its own InP wafer fabs, laser production lines, and packaging and module capacity. No other company has all four segments in-house.

The data center business's book-to-bill ratio has exceeded 4 times for several consecutive quarters, indicating that demand far exceeds capacity. Nvidia has also invested $2 billion, along with years of procurement commitments. FY2026 Q3 revenue reached $1.806 billion, a year-on-year increase of 21%. The 6-inch InP production expansion is ahead of schedule. The stock has risen about 455% over the past 52 weeks. It was included in the S&P 500 at the beginning of 2026, continuously attracting passive funds. Serenity has included COHR in its core map of DCO and laser suppliers.

MTSI

MAACOM is the invisible leader in the field of high-frequency analog chips for optical modules, with many of the hardest-to-replace components like TIAs (transimpedance amplifiers) and laser drivers coming from it. MTSI currently has a market cap of about $26 billion, with a stock price of $365.

Many people do not know that the most expensive and hardest-to-replace components in optical modules are not just DSPs but also high-frequency analog chips like TIAs (transimpedance amplifiers) and laser drivers, many of which come from MACOM. The competitive landscape is quite concentrated because designing high-frequency analog circuits and process optimization requires years of accumulation; it's not something that can be quickly replicated just by throwing money at it.

MACOM's product roadmap has covered the 1.6T and 3.2T ecosystems, with FY2026 Q2 revenue at $28.89 million and a Q3 midpoint guidance of about $33.5 million, with growth still accelerating. Gross margin is 55.9%, making it one of the few mid-cap companies in the optical module industry chain with robust high margins. Serenity has included it in the analog/mixed-signal IC and optical module chain, positioning it as an invisible shovel seller. EBC analysts call it 'the cleanest mid-cap choice for investors who want optical exposure without taking on pure broadcast vulnerabilities.'

DSP chips: AVGO, MRVL, CRDO

In an 800G optical module, DSP accounts for 30-40% of the cost and has far higher margins than the module itself. Often, the real profits come from DSP companies rather than the optical module companies. This layer is the most 'semiconductor-like' segment of the optical module industry chain, with the competitive landscape determined by SerDes technology accumulation and customer certification barriers.

AVGO

Broadcom is the absolute dominator in optical module DSP chips and also the largest player by shipment volume in CPO switches. AVGO currently has a market cap of about $1.5 trillion, with a stock price of $487.

Almost all high-end 800G and 1.6T solutions cannot avoid Broadcom's PAM4 DSP and SerDes, with a clear market dominance. They are also the most active promoters of CPO, having shipped over 50,000 Tomahawk 5 Bailly CPO switches throughout 2025, with the third-generation Tomahawk 6 Davisson having a switching capacity of 102.4 Tbps, making them the largest player in the CPO field by shipment volume.

Serenity lists AVGO as a downstream reference in the supply chain, alongside NVDA and MRVL, but notes it is not a focus for him in 'small-cap bottlenecks.' Its size limits its elasticity, serving as a ballast in the industry chain.

MRVL

Marvell is the target mentioned earlier by Jensen Huang at Computex, which saw a single-day surge of 32.52%, hitting a historic high. MRVL currently has a market cap of about $250 billion, with a stock price of $290.79.

After acquiring Inphi in 2021, Marvell's capability in optical module DSP has significantly increased, with many 800G modules' 'brains' being Marvell DSP. By February 2026, they completed the acquisition of Celestial AI and XConn, forming a complete tech stack covering silicon photonics, CPO optical engines, and CXL switching, being the only company currently covering all four dimensions of 'custom ASIC + 1.6T optical DSP + silicon photonics + CXL switching.'

Jensen Huang personally called it 'the next trillion-dollar company' at Computex, with a single-day surge of 32.52%, pushing its market cap above $250 billion. Serenity actively holds positions and has posted multiple affirmations of MRVL's potential while warning of valuation and belief risks. FY2026 full-year revenue is projected at $8.2 billion, with data center business accounting for over 75%. Custom ASIC business is expected to exceed $10 billion in FY2029, with FY2028 revenue target set at $16.5 billion. In the short term, it has significantly exceeded the sell-side consensus target price (mean $222 vs post-surge $290), so caution is advised for overheating risks.

CRDO

Credo is the leader in the AEC (Active Electrical Cable) sector, using copper cables to cut into part of the optical module market. This is one of the most complex narratives in the industry chain. CRDO currently has a market cap of about $60-80 billion, with a stock price of $226.

The most standout product is the purple AEC (Active Electrical Cable), which provides low-latency short-distance connectivity with copper cables, directly cutting into the market for pluggable optical modules. The core logic is: not all short-distance connections need optics; AEC offers a dual win in cost and latency in specific scenarios. FY2026 Q2 revenue reached $268 million, a staggering 272% year-on-year increase, with a year-to-date rise of over 120%. Serenity has included it in the interconnect/switching chain, positioning it as highly elastic yet narratively complex. The potential for disruption and being disrupted coexist, making this the most complex story in the industry chain.

Optical module assembly and foundries: AAOI, FN

This layer is the most familiar track to the market, and also the most competitive segment, with the greatest pressure from price wars. Companies that can succeed here either have vertical integration of upstream devices (Coherent, Lumentum) or extreme manufacturing efficiency.

AAOI

AAOI is the fastest-growing pure-play target for 800G and 1.6T optical modules, and also one of the most frequently called targets by Serenity, with over 123 public tweets. AAOI currently has a market cap of about $12-15 billion, with a stock price of $197.

This is also one of the few pure optical module companies in the U.S., with the core logic being the purest and fastest shipment of 800G. By 2026, it has secured over $200 million in large orders (including from super-scale cloud vendors like AWS and Microsoft), with a full-year revenue target exceeding $1 billion. The year-to-date rise is about 441%, making it the highest-gaining target in the entire optical module sector.

The cost of high elasticity is high concentration: revenue sources are concentrated among a few super-large clients, and any changes in order rhythm will be sharply reflected in the stock price. Suitable for short-term plays, not ideal for long-term holdings.

FN

Fabrinet is the only player in the optical foundry field with scale; regardless of who wins in the industry, it benefits. FN currently has a market cap of about $24 billion, with a stock price of $654.

The cleanest expression of optical foundry logic. A large proportion of the products from optical device manufacturers like LITE and COHR are manufactured by Fabrinet in Bangkok, Thailand. It does not bet on a single technology route and acts as a beta amplifier for the entire industry’s expansion. Pluggable modules, CPO, and OCS all require precision optical packaging, and there are only a few scale foundries in the precision optical packaging field, with Fabrinet being one of them.

FY2026 Q3 revenue and EPS set historical records, with Q4 revenue guidance between $1.25 billion to $1.29 billion. The annual increase is about 150%, with low volatility, making it suitable for core positions.

Optical network equipment and next-generation materials——NOK, LWLG

NOK

Nokia is the second leader in this Computex rally, easily overlooked. After acquiring Infinera, it simultaneously covers both DCI and AI backbone networks. NOK currently has a market cap of about $26 billion, with a stock price of $16.25.

Serenity has mentioned NOK multiple times, discussing its connections with policies and the supply chain, as well as its product layout after the Infinera acquisition. After acquiring Infinera, Nokia became one of the few suppliers in the optical network equipment market to simultaneously cover both 'DCI data center interconnect' and 'AI backbone networks.'

In Q1 2026, Nokia's optical network business grew by 20% year-on-year, with AI & Cloud-related sales growing by 49%, and Q1 profit growing by 54%, with the stock price hitting a 16-year high. New product lines include 1.6T coherent pluggable modules and compact amplifiers for multi-fiber applications, actively securing DCI orders from super-scale cloud vendors. With a market cap of about $26 billion, it is relatively undervalued as the second leader in this sector, making it an easily overlooked target in the Computex rally.

LWLG

Lightwave Logic is the smallest market cap and highest-risk target on this list, betting on electro-optic polymers becoming the standard material for next-generation silicon photonic modulators. LWLG currently has a market cap of about $1.7 billion, with a stock price of $12.67.

LWLG is not an optical module company; it focuses on next-generation modulator materials in silicon photonics——electro-optic polymers (EO Polymer).

Traditional silicon-based modulators have a bandwidth of 40-50GHz, while LWLG's polymer modulator targets specifications above 110GHz, supporting 400G/lane, representing a 2-3 generation leap in existing technology. In March 2026, LWLG signed an agreement with Tower Semiconductor to incorporate its polymer modulator solution into Tower's silicon photonics PDK, meaning any customer using Tower's production line can directly access LWLG's material solutions. Currently, four Fortune 500 companies are in Stage 3 prototype development, including collaboration with a second global Fortune 500 client to develop a 400Gb/s CPO solution, with commercialization supply agreements expected to materialize in 2027. PhotonCap (suspected of being Serenity's optical engineering research account) has written an in-depth analysis report on LWLG's collaboration with Tower.

Logic: If EO polymers become the standard modulator material in silicon photonics, LWLG's IP licensing model will be a high-leverage winner. But currently, there is no substantial revenue, making it suitable for extremely high-risk positions and not for core holdings.