With a year-to-date return of 4502.45%, and 25 publicly listed stocks showing gains of 100%–1000%… what are the investment methodologies of the wildly popular stock guru Serenity? How can we learn and reuse them? What are their limitations? The simplest approach: the bottleneck investing method. Serenity's bottleneck investing method, simply put, involves first identifying a definite major trend, then dissecting the industry chain to find the most irreplaceable upstream link, and finally betting before the market fully prices it in. For example, if the market hasn't yet realized that the upgrade of optical interconnects in AI data centers will make a certain upstream material, laser, or testing equipment a scarce asset, then this small link may receive a valuation repricing far exceeding its current revenue. Just like a restaurant's most expensive dish is its main course, but what truly hinders its business might be a niche seasoning; if this seasoning is unavailable, all main courses cannot be made. Bottleneck breakdown: Determined demand × Limited supply × Low attention × Value capture × Catalyst. Essentially, the bottleneck methodology, when broken down, resembles a five-factor model: demand must be sufficiently certain, supply must be sufficiently narrow, market perception must lag, potential value must be sufficiently clear, and there must be verifiable events to catalyze future growth. Only when all five conditions are met simultaneously can small companies potentially achieve excess returns. The first layer: Determined demand. AI data center expansion, cloud vendor ASICs, self-developed chips, inference demand, and bandwidth demand constitute the broader demand context. Serenity repeatedly mentions AMZN Trainium, MSFT Maia, Google TPU, and NVDA's push for 800V DC, indicating that he doesn't view small companies in isolation but rather places them within the context of giants' capital expenditures and architectural migrations. For example, in his AAOI/LITE related tweets, he wrote that the market rewarded the Google TPU supply chain but may have underestimated the optical interconnect demand from AMZN Trainium and MSFT Maia. The second layer: Constrained supply. The so-called bottleneck is not a vague statement like "this thing also benefits," but rather "it's indispensable" and "it's difficult to replicate in the short term." For example, InP substrates, CPO external light sources, CW DFB lasers, SOI wafers, and optical transceiver testing equipment may sound niche, but once AI data centers migrate from electrical connections to optical connections, these aspects will become bottlenecks in production capacity, yield, certification cycles, and customer adoption. Taking InP substrates as an example, InP plays a crucial role in high-speed optical communication lasers, detectors, and some photonic devices, especially in direct bandgap, luminous efficiency, and high-speed modulation scenarios. However, due to limitations such as certification cycles, long equipment lead times, high production process barriers, production capacity expansion failing to keep pace with surging demand, and structural shortages, mass production replication is difficult in the short term. The third layer: low attention.Low attention equals genuine price undervaluation. Many of Serenity's targets are not at the center of the mainstream narrative, but mispricing is more likely to occur in areas where "institutional coverage is low, retail investors don't understand, and the media hasn't fully covered them." Fourth layer: Value capture. Pricing power, profit margin, customer lock-in, and supply share. Converting a true bottleneck into excess returns requires several conditions: can the company secure production capacity, can it set prices, is it facing price pressure from customers, does it need financing for dilution, can profit margins be realized, and has demand already been priced in by the stock price? Fifth layer: Catalysts. Long-term potential is important, but short-term catalysts are also price engines. Short- to medium-term triggers include: financial reports, customer mass production, Jabil fireside chat, CHIPS Act, index inclusion, Nasdaq dual listing, M&A, short-selling crowding, and funds flowing from the local market to US investors. Typical case analysis: 1. $AXTI: The most classic bottleneck case. Serenity's Reddit account was previously suspended for analyzing AXTI. At the time, AXTI had a small market capitalization and a niche business, primarily producing InP substrates, and was considered a "small-cap stock." However, Serenity's understanding was that AI data center optical communication requires underlying materials like InP, and if supply is limited, the entire photonics supply chain will be affected. Subsequently, AXTI's stock price rose nearly tenfold from about $14, further demonstrating its core competence: not focusing on whether the stock price will rise, but first judging whether this link will transform from a "niche material" into a "strategic bottleneck." 2. RPI: Small-cap companies are extremely sensitive to marginal demand. The same change in demand might only cause a 1% revenue disruption for a large company, but for a small company, it could lead to a revaluation of its valuation. For example, increased demand for AI hardware, development boards, and edge devices has a limited impact on giants like Apple, but for a smaller hardware company like RPI, it could directly change its growth curve. Serenity's bullish view on RPI is that if AI agents require a large number of low-cost local nodes or edge orchestration hardware, then this "small computer" could suddenly become an infrastructure for the proliferation of AI applications. 3. $AAOI / $LITE: Expanding from Single-Point Bottlenecks to a Supply Chain Map. Serenity places LITE within the TPU/OCS benefit chain and AAOI within the MSFT Maia and AMZN Trainium ramp-related chains, suggesting that InP might become a bottleneck in 2026, similar to HBM. Bottlenecks aren't just viewed as isolated points, but rather considered within a broader context: after the Google TPU chain is rewarded by the market, the next step might be for companies like AMZN and MSFT with self-developed ASICs related to optical interconnects to be discovered. How can we better utilize Serenity's thinking path? There are six steps: Step 1: Identify the major trends: Has the demand been validated?First, identify the trend, not just stocks. For example, AI computing power expansion, CPO optical interconnects, 800V DC, humanoid robots, stablecoin payments, and RWA tokenization are all trends. Second, map the supply chain: What are the links from the end-user to the upstream? Draw the supply chain diagram. Taking CPO as an example, we can't just know about NVDA; we also need to know about ASICs, switches, optical modules, external light sources, lasers, InP/SOI materials, packaging, testing, fiber arrays, microlenses, etc. Third, identify the bottleneck: Which link is the most difficult to expand production/replace? Determine whether it's a "real bottleneck" or a "pseudo-bottleneck." Real bottlenecks usually have several characteristics: concentrated supply, long certification cycles, high customer switching costs, low yield rates, slow expansion, and reliance on giants' roadmaps. Fourth, gather evidence: Do you have customers, certifications, capacity, and order leads? Use evidence, not emotions, to build confidence. Evidence can include: customer leads in annual reports, management meeting minutes, supplier qualification, CHIPS Act/government funding, index inclusion, patents, recruitment, capacity expansion, cooperation announcements, customer product roadmaps, and peer capex. Step 5: Risk Control. If you're wrong, where did you go wrong? Always create a "contradictory table." Make bold assumptions, but carefully verify them. Buying isn't a one-time solution. If the client doesn't scale up, when will the revenue prove false? If competitors replace them, will the bottleneck disappear? Step 6: Match your position size with your research depth. If you've only read others' summaries, your position should be small; if you can draw up the industry chain yourself, read annual reports, analyze clients, and do scenario-based valuations, your position can be larger. What are the limitations of the bottleneck point investment method? 1. Inferences are prone to overfitting. Serenity excels at connecting regulatory documents, partnership announcements, client websites, and financial statement wording, but this method inherently carries the risk of misjudgment. A client deleting a supplier from their website, a company appearing on Blueprint, or a partner having a relationship with Hyperscaler—these could be strong clues or just noise. 2. When early financials are poor, valuation lacks an anchor. For targets like SIVE, XFAB, and AAOI, Serenity often focuses on future revenue ramp-up, architecture migration, and potential M&A from 2027 to 2029, rather than current profits. This approach offers high odds when the direction is correct, but is prone to misjudgment when the direction is wrong. 3. Liquidity reflexivity risk: Serenity has become a market variable. Once it publicly expresses its bullish view on a small-cap stock, follower funds may directly drive up the price, directly impacting the odds. 4. From a critical perspective, there is also some survivorship bias. The 4500% return, besides having a logic worth learning from, is largely due to catching the massive one-sided bull market in AI computing power.Past experience may not be applicable in the future. Will giants find ways to circumvent current bottlenecks? In short, the bottleneck investment method is effective because the market often prices the overall narrative first, then secondary suppliers, and only then realizes the real shortages in materials, components, testing, and production capacity. This method heavily relies on professional judgment, information piecemeal analysis, tolerance for non-consensus, and position discipline. What we should truly reuse is not Serenity's holdings, but his research sequence: first identify the trend, then the bottleneck, then the evidence, then valuation, then catalysts, and finally, place bets with a manageable position size. [@rayrayweb]