In a tweet on June 1st, '46x stock god' Serenity declared AAOI as his most favored stock in the photonics sector, forecasting a 4 to 5 times return over 12-24 months.
In Series One, we used the 9x stock AXTI to elucidate Serenity's methodology and also dissected Serenity's recent favorite, SIVE. This Series Two will accomplish three things: First, tell you what kind of company AAOI is. Second, systematically outline Serenity's logic for AAOI using the methodology we've summarized. Third, present our own viewpoint.
In summary, the logic for AAOI follows the same lineage as SIVE and AXTI previously analyzed in this series: possessing the attributes of both a 'choke point' (high barriers) and a 'bottleneck' (supply-demand imbalance) within a context of highly certain end-market demand.
AAOI: U.S.-Based Photonics Leader
a. One-Sentence Introduction
AAOI (Applied Optoelectronics, Inc., NASDAQ: AAOI) is a company that manufactures high-speed optical communication equipment and optical modules. In simple terms, it is a supplier of 'fiber-optic high-speed transmission components' for AI data centers, cloud computing, and broadband networks.
b. Product Introduction: The Most Promising Segment is the Data Center Photonics Business
AAOI's main product lines can be divided into three major segments:
Data Center Optical Transceivers (Datacom Transceivers): This is the core business segment Serenity is bullish on. It primarily includes 100G, 400G, 800G, and developing 1.6T high-speed optical modules. Major customers are hyperscale cloud service providers (Amazon, Microsoft, etc.), used for high-speed, low-power, low-latency interconnects between GPUs/switches, supporting the physical layer upgrade of AI clusters from copper to optical interconnects. In 2025, this segment generated $196 million in revenue, accounting for approximately 43% of total revenue, a year-on-year increase of 31.8%. This is also the core business for Serenity's recommendation.
CATV and Telecom Products Segments: Main customers are cable TV companies and operators. Although they saw cyclical high growth in 2025, long-term they are seen as mature, stable businesses. CATV revenue in 2025 was $245 million, a 179.5% year-on-year increase, accounting for about 54% of total revenue. Telecom/FTTH is smaller in scale, with 2025 revenue of $13.7 million, constituting 3% of revenue, a 25% year-on-year increase.
Additionally, the company has an important product line: High-Power Semiconductor Lasers & Components. This includes narrow-linewidth pump lasers (400mW class) specifically designed as external light sources (ELS) for Silicon Photonics (SiPh) and Co-Packaged Optics (CPO) architectures. The company is developing ELSFPs (External Laser Source FP modules) to directly enter the CPO supply chain. This segment represents the core 'light-emitting engine' for photoelectric conversion, addressing the physical limitation that silicon material itself cannot efficiently emit light. Revenue from Lasers & Components is not separately disclosed in financial reports but is accounted for within the Data Center, CATV, and Telecom segments based on the final application market. The laser products within the Data Center business are also a crucial part of Serenity's AI narrative.
c. Financial Situation: Rapid Revenue Growth, Significant Reduction in Losses
The company is experiencing rapid revenue growth alongside a substantial reduction in losses. In 2025, revenue was $456 million (an 82.7% year-on-year increase), gross margin was 30.0% (up 5.2 percentage points), and net loss was $38.2 million, a reduction of $149 million compared to the previous year's loss.
d. Rapid Capacity Expansion
The company is rapidly expanding capacity. By the end of Q1 2026, monthly production capacity for 800G transceivers had approached 100,000 units. This implies an annualized revenue run-rate for 800G transceivers of approximately $1.1 billion at current average prices, equivalent to more than twice the company's 2025 revenue. Furthermore, the company is expanding manufacturing capabilities in the U.S. and Taiwan, focusing on supporting laser diodes, 800G, and 1.6T products. The company expects capital expenditures in 2026 to be significantly higher than in 2025 and plans to expand capacity at least through the end of 2027.
Dissecting Serenity's Logic
Serenity's methodology is consistent: starting from highly certain end-market demand, working backwards layer by layer to identify segments that possess both 'choke point' attributes (high barriers, difficult to replace) and 'bottleneck' attributes (severe supply-demand imbalance). This applies to previously analyzed AXTI and SIVE, and naturally applies to AAOI.
a. Optical Modules and CPO Under the Super-Trend of Photoelectric Conversion
i. The Megatrend of Photoelectric Conversion
AI data center interconnect technology is undergoing a clear evolution: from copper cable interconnects, to pluggable optical modules, to the current CPO (Co-Packaged Optics) architecture.
Traditional copper cable interconnects, due to physical limitations, can no longer meet the scaling demands of AI clusters. At extremely high data rates, copper cables suffer from high power consumption, severe heat generation, significant signal attenuation, and limited bandwidth density. As single large model training and inference require tens of thousands to hundreds of thousands of GPUs, the bottlenecks of copper in power consumption, heat dissipation, and signal integrity become fully exposed. Currently, high-speed interconnects within AI data centers have largely completed the switch from copper to optical. Copper is only used sparingly in extremely short-distance, low-speed scenarios.
Within optical interconnects, the industry is further evolving from pluggable optical modules towards CPO architecture. While pluggable modules solved the power and distance issues of copper, the modules themselves occupy panel space, the pluggable interfaces introduce signal loss, and there are still limits on power consumption and bandwidth density, making it difficult to support next-generation, higher-performance AI clusters. CPO packages the optical engine directly near the switch chip or GPU, significantly shortening the photoelectric conversion path, thereby achieving lower power consumption, higher bandwidth density, and superior signal integrity. Currently, CPO is still in the verification and early small-scale deployment phase. Mainstream AI data center construction in 2026-2027 will still primarily rely on 800G/1.6T pluggable optical modules. Widespread replacement by CPO is expected to gradually become apparent around late 2027 to 2028.
Regardless of the form factor, photoelectric conversion (Electrical-to-Optical) is an irreplaceable core component in AI computing infrastructure. As AI cluster scale continues to expand, demand for high-speed, low-power optical interconnects will continue to grow rigidly.
ii. AAOI's Products and Strategy Align with the Major Trend
AAOI's product portfolio aligns highly with current and next-phase industry demands.
Current Phase (Pluggable Module Phase, 2026-2027): Large-scale AI data center construction still uses pluggable optical modules as the mainstream interconnect method. 800G and 1.6T optical transceivers are the highest-volume, fastest-growing categories in this phase. AAOI's 800G/1.6T transceivers handle high-speed optical signal transmission between GPUs and switches and are currently the company's main source of revenue and profit growth. In this phase, AI cluster construction is extremely rapid, while CPO has not yet entered large-scale replacement, maintaining strong demand for high-end pluggable optical modules.
Next Phase (CPO Penetration Period, gradually starting ~2028): As CPO architecture advances, because the optical engine is packaged near heat-generating chips and cannot efficiently emit light itself, it must rely on external light sources (ELS) to provide stable optical signals. This will create new demand for high-power, narrow-linewidth, low thermal-drift lasers. AAOI already possesses core technology for manufacturing high-power narrow-linewidth laser chips (used in CATV, etc.) and is productizing this technology into CPO-specific modules like ELSFPs (External Laser Source FP), currently in development and customer validation stages, to tap into this emerging growth segment.
Therefore, AAOI is currently primarily capitalizing on the volume opportunity of pluggable modules, while simultaneously positioning itself for the CPO era through its laser expertise and vertical integration capabilities, forming a clearer medium-to-long-term advantage in the photoelectric conversion field.
b. Choke Points: Full-Stack, U.S. Domestic, Difficulty for New Entrants
i. Full-Stack Advantage
AAOI possesses full-stack capability, i.e., complete vertical integration from laser wafer manufacturing to module assembly. This reduces supply chain disruption risks and improves supply chain synergy between upstream and downstream. The company performs laser chip epitaxial growth, device processing, packaging & testing, and final module assembly all at its Sugar Land, Texas facility, forming a closed-loop production system. This full-stack model significantly lowers supply chain disruption risks, allowing it to respond more stably to major customer demands in the current environment of generally tight laser and packaging capacity and delicate geopolitics. Simultaneously, because laser and module development and production occur within the same system, process synergy is stronger, facilitating faster technology iteration and product launch for next-gen products like CPO and ELS.
ii. U.S. Domestic Advantage
The ability to conduct the entire production chain in the United States aligns with hyperscalers' strategic preference for supply chain security and 'de-risking' of AI-critical infrastructure. This provides a preferential or exclusive consideration advantage during supplier qualification and initial order allocation.
iii. Supplier Expansion is Difficult and Slow
Major companies do not easily onboard small suppliers, requiring long cycles, further limiting industry capacity expansion. Hyperscalers have stringent reliability requirements for AI training/inference clusters. Any supplier switch or new entrant must pass rigorous aging tests, system-level validation, and multiple rounds of small-scale trial production over 12-24 months. AAOI has established a first-mover advantage through historical delivery records and current first-wave 800G volume shipments. However, the overall industry remains in a supply-constrained state characterized by 'few players + long validation cycles,' with major players unwilling to risk core capacity on unproven new entrants.
Companies without prior experience in manufacturing earlier-generation optical modules find it difficult to quickly enter. Players lacking experience in high-volume optical module production cannot simultaneously solve the four major challenges of yield, cost, supply chain certification, and customer trust in the short term. Optical communication has extremely high requirements for signal integrity, thermal management, and long-term reliability, creating deep barriers of accumulated know-how and customer relationships. Even with capital and technology, new entrants would find it difficult to achieve the leap from zero to volume production within 2-3 years.
c. Bottleneck: AI-Native Demand
i. Industry's Original Capacity Was Essentially Zero; Demand Derived from the AI Megatrend
The industry's demand stems from AI-'native demand,' with no prior surplus capacity. Before the AI boom, 800G+ high-speed AI-specific optical modules were used in niche telecom scenarios, with global effective high-yield capacity close to zero. The explosive new demand from AI data centers has created an incremental market whose magnitude far exceeds the historical installed base. Existing players like AAOI are meeting this wave of demand starting from an almost blank slate. According to McKinsey estimates, there could be a 40%-60% supply gap for 800G in 2027, and a 30%-40% gap for 1.6T potentially lasting into 2029. Notably, this demand is calculated based on downstream needs like data centers, but data centers themselves are supply-constrained and have not yet resolved their own bottleneck issues.
ii. Capacity Build-out Cycles Are Extremely Long
Expanding optical module production lines involves building ultra-clean rooms, procuring precision equipment (MOCVD, electron-beam lithography, etc.), debugging process parameters, training personnel, and yield ramp-up. The entire cycle often takes 12-24 months or longer. The overall industry expansion speed struggles to match the pace of demand explosion.
d. Other: Dilution Risk
Additionally, Serenity often mentions the dilution pressure from a $600 million ATM equity financing as the biggest risk he sees. This mechanical selling pressure is the primary factor currently weighing on the stock price. However, he also states this is a temporary issue. Once the selling pressure from the financing ends, AAOI's fundamental advantages (U.S. domestic full-chain production and revenue ramp potential) will be more easily reflected.
Our Perspective
Our view, within this framework, is: AAOI's bottleneck is very strong, but its choke points are relatively weak. Additionally, the company faces execution challenges. This configuration of a strong bottleneck but weak choke points will significantly increase the company's cyclicality. At the beginning and end of the cycle, pricing power is affected by barriers (choke points), leading to lower profit margins. However, during the peak of the cycle, it will benefit from the bottleneck and soar.
a. Strong Bottleneck, Weak Choke Points
i. Supply-Demand Imbalance is Indeed Pronounced, Bottleneck Characteristics Are Evident
From both supply and demand perspectives, the bottleneck characteristics in AAOI's industry are very apparent.
From the supply side: 1. Industry capacity started from zero, meaning no idle capacity existed previously. 2. Adding new suppliers is difficult. 3. Existing suppliers are expanding capacity slowly.
From the demand side: 1. AI data center construction speed exceeds the optical supply chain's expansion pace. 2. AI data centers themselves are supply-constrained due to multiple capacity shortages. In summary, demand is certain and persistent, while supply faces multiple constraints, making it difficult to reverse the imbalance in the short term.
ii. Choke Points Are Weaker; Fierce Competition Among Incumbents
However, we believe the choke points are less pronounced.
1. While it's hard for external suppliers to enter, competition among existing players is fierce.
Although external suppliers are hindered by technical积累 and certification cycles, competition among the players already in the field is intense. Currently, in the 800G+ high-speed optical module space, the players with scaled mass-production capabilities mainly include AAOI, Innolight (Zhongji Innolight), Accelink (Guangxun Keji), Lumentum, Coherent, and about 5-6 other leading companies, with another group of second-tier companies watching closely. In this competition, each company has its strengths. For example, Innolight has clear advantages in scaled production and cost control, while Accelink, Lumentum, and Coherent have their own strengths in product line completeness, technical积累, and specific niches. For downstream major customers, optical module supply is typically not long-term tied to a single supplier. Instead, they employ a multi-vendor strategy, continuously comparing factors like price, delivery time, yield, reliability, and speed of next-generation product introduction.
2. U.S. domestic advantage only guarantees staying within the system, not guaranteeing long-term volume allocation.
The U.S. domestic advantage does indeed help AAOI get on the hyperscalers' 'approved list' and receive some order preference, but this is merely an 'entry ticket'—an insurance policy downstream customers feel compelled to buy to hedge against geopolitical risk. Actual allocation volume still depends on market factors like cost, capacity, and yield. In other words, the U.S. advantage primarily ensures AAOI's baseline is being on the supplier list, but does not guarantee it a high market share. Furthermore, compared to Chinese companies like Innolight, the U.S. background also has a certain negative impact on overall cost.
iii. Comparison with SIVE's Logic
Let's now compare AAOI with SIVE. We dissected SIVE last time; it is Serenity's most favored stock overall (AAOI is his most favored U.S. stock). By comparison, we find that AAOI's bottleneck is slightly stronger, but its choke points are significantly weaker.
Regarding the bottleneck: AAOI faces an industry starting from scratch, whereas SIVE's industry had traditional uses before, so there was some prior surplus capacity, making its bottleneck slightly weaker.
Regarding choke points: Unlike AAOI, which faces numerous competitors, SIVE's technology is proprietary and it has already locked in upstream capacity, constituting stronger choke points.
b. The Other Side of the Full-Stack Double-Edged Sword: Execution Difficulty
Another choke point mentioned by Serenity, the full-stack model, is not an absolute advantage but a trade-off. From a supply chain stability perspective, it reduces stock-out risk at the cost of increased execution risk. In the current geopolitical context, the importance of stock-out risk has increased, but this doesn't mean execution risk can be ignored. Full-stack vertical integration (from laser wafer manufacturing to module assembly) is AAOI's core competence, allowing for better control over quality, cost, and delivery schedules. However, this also means the company must maintain leadership across multiple high-difficulty stages simultaneously: any bottleneck in epitaxial growth, device processing, packaging, or module testing can hinder the overall ramp-up speed. Especially during critical technology transition windows, a full-stack player needs to simultaneously manage process upgrades and capacity expansion, making execution far more complex than fab-lite competitors or those focused solely on module assembly.
Currently, AAOI is under dual pressure from major factory construction in Texas and the rapid ramp-up of 800G/1.6T production. The execution difficulty brought by the full-stack model is rising significantly. Historically, AAOI has also experienced performance below expectations due to execution issues under its full-stack model. For example, in 2018, during the transition to higher-speed products, there were delays in new product validation and manufacturing yield fluctuations, leading the company to lower guidance multiple times. Similarly, in 2022-2023, due to slower-than-expected new product introduction and capacity ramp, it experienced consecutive losses and gross margin pressure.
c. Valuation is Already High
AAOI's current market capitalization is approximately $13 billion, translating to a PS (Price-to-Sales) multiple of about 28x based on its 2025 revenue of $456 million. Even assuming the company's guidance of exceeding $1 billion in revenue for 2026, the PS would remain at 13x, significantly higher than the historical valuation multiples of traditional optical communication companies.
At this valuation, combined with its industry position of a strong bottleneck but weak choke points, it still possesses multi-bagger potential if the subsequent cycle's strength and pace meet expectations. However, if expectations are not met, the downside potential is also substantial.
Summary: Strong Cyclicality + Uncertainty
AAOI is another typical example of Serenity's methodology—extremely certain end-demand, a real bottleneck exists, with huge upside potential, but the strength of its choke points and valuation safety margin require further verification. We acknowledge its explosive potential during the capacity ramp-up window of 2026-2027. However, full-stack execution difficulty is increasing, and the cycle carries a degree of uncertainty. We recommend making decisions only after verifying valuation, cycle dynamics, and execution.
