My conclusion is straightforward: these five stocks are not the same "AI trade"; they represent five different nodes on the AI infrastructure chain. If the market continues to pull back due to inflation, interest rate concerns, or bubble fears, I would place them on a tiered watchlist rather than interpreting "buying the dip" as a single, full-position, aggressive chase. This report discusses MU Micron, MXL MaxLinear, AMD, LITE Lumentum, and VICR Vicor. While they all benefit from AI data center capital expenditures, their risk sources, earnings resilience, and valuation digestion paths differ. [1] [2] [3]
I believe that at this stage of the AI market cycle, the truly important questions are not "Does the AI story still have legs?" but rather three specific ones: First, can capital expenditures translate into real orders? Second, can corporate earnings justify valuations? Third, can investment portfolios withstand high volatility? McKinsey estimates that to meet computing power demands, global data centers may require approximately $6.7 trillion in capital expenditure by 2030, with about $5.2 trillion related to AI workloads. This indicates AI infrastructure is a long investment cycle. However, Fidelity also cautions that earnings growth, valuation, the sustainability of capital expenditure, and the interest rate cycle will determine whether the AI trade transitions from a long-term theme to a short-term bubble. [1] [2]
Conclusion in one sentence: AI infrastructure remains a direction I am willing to research on dips, but entry points must adhere to position-sizing discipline. In a phase where high returns, deep drawdowns, and high volatility coexist, first tier, then act.
I. The Big Picture: AI Infrastructure Is Not a Story Told by One GPU Stock
The easiest mistake the market can make is equating the AI theme simplistically with "buying the GPU leader." In my view, the real structure of AI infrastructure is a capital expenditure chain: front-end needs compute chips, the middle requires high-bandwidth memory, networking, and optical communication, and the back-end needs power supplies, cooling, data centers, and software orchestration. Focusing solely on a single link makes it easy to chase at the wrong pace when valuations are extremely high. Deconstructing the chain reveals whether each pullback is driven by valuation contraction, order deterioration, or is simply a normal shakeout for high-beta assets.
McKinsey's estimates on data center capital expenditure provide an important context for this framework. It doesn't suggest all companies will benefit simultaneously or that all AI-related stocks should rise. Instead, it indicates that if compute demand continues to grow, investment opportunities will diffuse along the "compute—memory—connectivity—optics—power" chain. [1] Morningstar's discussion on the AI stock framework also reminds me that AI stock selection should not rely solely on thematic hype but must simultaneously consider industry positioning, moat, valuation, and uncertainty. [3]
My judgment is that the AI infrastructure opportunity is not a "single line" but a "network." Once the market pulls back, the most worthy research target is not necessarily the one that fell the most, but rather which node's fundamentals remain intact while its valuation gets hammered alongside overall risk appetite.
Public price data from the past year shows these five AI infrastructure stocks have significantly outperformed both the Nasdaq 100 and the SMH semiconductor ETF. LITE, MU, MXL, VICR, and AMD have all seen substantial gains, with LITE and MU performing most notably. However, the same data set also reveals that the maximum drawdown for these five stocks over the past year mostly ranged from about -28% to -32%, significantly higher than the Nasdaq 100's approximate -12.1% maximum drawdown. [9]
This data gives me a clear insight: strong trend does not equal low risk, and high elasticity does not mean it's always a good time to buy. If a stock surges several-fold in a year but can experience a 30% drawdown along the way, the buying thesis cannot just be "bullish on AI long-term"; it must also clearly outline "how to withstand the volatility." In other words, buying the dip is not an emotional slogan but a set of capital management rules.
I will use this table as the starting point for position management. For stocks like MU and AMD with stronger fundamental validation, I am willing to observe in batches during drawdowns. For high-elasticity nodes like MXL, LITE, and VICR, I will first cap the position size before considering price levels. The reason is simple: volatility itself is a cost, and "buying the dip" that ignores this cost can easily turn into passively holding a losing position.
II. The Differences Among the Five Stocks: Don't Buy Based on Who Gained More, But Based on Whose Evidence Chain is More Complete
I do not agree with placing these five companies in the same basket for a crude comparison. MU's core revolves around the memory cycle and AI HBM demand. AMD's core is its data center computing platform. LITE's core is cloud and AI optical communication. VICR's core is high-power server power delivery. MXL leans more towards the AI data center control plane and high-speed connectivity. They all benefit from AI, but their financial elasticity, customer structure, and paths to valuation digestion are not the same.
Based on public company materials: Micron's FY2025 Q4 press release disclosed quarterly revenue of $11.315 billion and FY2025 annual revenue of $37.378 billion, linking the strong performance to AI data center demand; AMD's Q3 2025 press release reported quarterly revenue of $9.246 billion, up 36% year-over-year, with data center revenue at $4.3 billion, up 22%; Lumentum's FY2026 Q3 press release reported revenue of $808.4 million, up 90.1% year-over-year, emphasizing AI, cloud computing, and next-generation communication-related photonics technologies; MaxLinear's public press releases introduce its Coronado and Laguna USB UART solutions for AI data center control plane connectivity; Vicor emphasizes in public materials the demand from AI, HPC, and data center compute growth for its 48V modular power systems. [4] [5] [6] [7] [8]
My sorting is not a simple "ranking by gains." If looking only at past year gains, LITE and MU are the most dazzling. If looking at the fundamental evidence chain, MU and AMD are easier for institutional capital to track consistently. If looking for high-elasticity satellite positions, MXL, LITE, and VICR offer steeper return curves but also demand stricter stop-loss and position caps.
III. Risk-Return Positioning: The Upper Right Corner is Not Paradise, But a Discipline Test
Many investors love to see high-return charts but dislike drawdown charts. My view is precisely the opposite: for high-beta AI stocks, the return rate is merely the outcome; the maximum drawdown is the term you must accept before entering the position. Chart 3 plots the past year's return against the maximum drawdown, showing the five stocks are in the high-return zone, but the drawdowns on the vertical axis are also deep. This indicates
they are not low-volatility growth stocks but high-elasticity assets that need to be digested with position discipline. [9]
I would use a three-tier system to handle such stocks. The first tier is "Core Trackable," referring to stocks with more complete fundamental evidence and more thorough institutional coverage, such as MU and AMD. The second tier is "High-Elasticity Satellite," referring to stocks with clear industry logic but high volatility, such as LITE and VICR. The third tier is "Observational Elasticity," referring to stocks with imaginative product direction but whose financial realization still requires more quarterly verification, such as MXL.
Therefore, my definition of "buying the dip" is not buying whenever the price falls. It is when a price drawdown occurs, fundamentals have not deteriorated, and the capital expenditure chain is still materializing, then absorbing the volatility in batches according to predetermined position rules. Especially for high-volatility stocks like MXL, LITE, and VICR, position size is more important than the entry price.
IV. Industry Chain Scoring: The Five Stocks are Not the Same Trade, But Five Different Nodes
To avoid lumping all AI stocks together as a single concept, I score these five stocks across five dimensions: directness to compute power, sensitivity to AI capital expenditure, cyclical volatility, valuation digestion pressure, and portfolio diversification value. This scoring is not a return forecast or an investment rating, but helps me determine: if I were to create an AI infrastructure watch basket, what specific role each stock plays.
This chart gives me the insight that MU and AMD resemble core evidence assets for the main AI infrastructure theme; LITE and VICR are more like high-elasticity nodes in the chain that are easily amplified by capital flows; MXL leans more towards an "observational" play where valuation could be reassessed post-product adoption. All five stocks have research value, but the buying thesis absolutely cannot be identical for all.
My allocation thinking is: if you only want core AI exposure, prioritize researching MU and AMD with their more complete evidence chains. If willing to bear higher volatility, you can consider LITE and VICR as satellite observations. If allocating to MXL, you must acknowledge its small-cap attributes and revenue realization uncertainties, and cap its position size more conservatively than the others.
V. Operational Framework: The Real Buying Opportunity Arises When Three Things Appear Simultaneously: "Drawdown, Confirmation, Staggered Buys"
I will not treat every AI-related pullback as a buying opportunity simply because the theme is strong. A pullback truly worth acting on must at least satisfy three conditions simultaneously: First, the price has already released short-term sentiment pressure. Second, the company's fundamentals have not deteriorated in sync. Third, there is still cash and risk budget available in the portfolio. Missing any one of these, buying the dip becomes an emotional trade.
Fidelity's framework on AI bubble risk is worth referencing here. It reminds us that while the AI theme might still be a multi-year cycle, investors must track earnings growth, earnings quality, valuation, the sustainability of capital expenditure, and the interest rate cycle. [2] I fully agree with this perspective. AI is not un-investable, but it should not be bought when valuation is most expensive, sentiment is hottest, and positions are fullest, using "long-termism" to mask short-term risks.
In summary, I would place these five stocks into an AI infrastructure watch pool, but I will not treat them all as a buy list with equal weight. For me, the correct sequence is first define the role, then define the position size, and only then define the price.
VI. Conclusion: You Can Buy the Dip, But First Ask Yourself if You Can Withstand the Volatility
The final conclusion returns to the title: Buying the dip on the five leading Nasdaq AI stocks is researchable, but you cannot be lazy about it. If AI data center capital expenditure continues to expand, the storage, computing, optical communication, power, and connectivity segments where MU, AMD, LITE, VICR, and MXL operate have a foundation to continue benefiting. However, if interest rates rise again, cloud capital expenditure slows, AI order realization falls short of expectations, or valuations have already priced in multiple quarters of future growth, these high-beta assets could also pull back rapidly.
My strategy is clear: core positions are prioritized for assets with stronger fundamental evidence chains, satellite positions are allocated to high-elasticity but high-volatility nodes, and watchlist positions are for small/mid-cap opportunities still requiring validation. Buying must be staggered, position sizes must be limited, and risks must be written down in advance. Truly mature AI investing is not getting excited at every dip, but knowing which dip is buyable, how much to buy, and what to do if wrong.
Summary in one sentence: The long-term logic for AI infrastructure remains, but buying the dip is not a charge signal; it's a discipline checklist. First, deconstruct the five stocks into five nodes, then use position sizing and time to digest the volatility.
Risk Notice
This report is for research and discussion purposes only and does not constitute any promise of returns or individual stock trading recommendations. Companies related to AI infrastructure generally possess characteristics of high volatility, high valuation sensitivity, and strong cyclical attributes. Investors need to make independent judgments based on their own risk tolerance. Five key risk categories to monitor closely going forward are: First, if cloud provider capital expenditures fall below expectations, AI hardware chain orders may be repriced. Second, if interest rates rise again, high-valuation growth stocks will face discount rate pressure. Third, segments like memory, optical communication, power supplies, and connectivity carry inventory cycle and customer concentration risks. Fourth, small/mid-cap, high-elasticity targets may experience amplified liquidity and valuation fluctuations. Fifth, if the AI theme sees insufficient earnings realization, the market may shift from "pricing long-term potential" to "pricing current cash flows."
This report is compiled by a guest analyst. The views expressed herein represent the author's personal position and do not represent the views of the BIT platform. This material is for informational purposes only and does not constitute investment advice.
