Focus: Five Nasdaq AI Leading Stocks

Investment Summary

My conclusion is straightforward: these five stocks are not the same “AI trade,” but rather five distinct nodes within the AI infrastructure chain. If the market continues to correct due to concerns about inflation, interest rates, or bubbles, I will place them on a tiered watch list, rather than interpreting “buying the dip” as a one-time, full-position chase. This report discusses MU (Micron), MXL (MaxLinear), AMD (Advanced Micro Devices), LITE (Lumentum), and VICR (Vicor). They collectively benefit from AI data center capital expenditures, but their risk sources, earnings resilience, and valuation digestion methods differ. [1] [2] [3]

I believe that at this stage of the AI rally, what truly matters is not “whether AI still has a story,” but three questions: First, can capital expenditures continue to translate into real orders? Second, can corporate earnings justify the valuations? Third, can an investment portfolio 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 AI workload-related data centers accounting for about $5.2 trillion. This indicates that AI infrastructure is a very 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 shifts from a long-term theme to a short-term bubble. [1] [2]

One-sentence conclusion: AI infrastructure is still a direction I am willing to research on dips, but entry points must obey position discipline. In a phase characterized by high yields, high drawdowns, and high volatility, the strategy is to tier first, then act.

1. The Big Picture: The AI Infrastructure Story Cannot Be Told by a Single GPU Stock

The easiest mistake the market makes is to equate the AI rally simplistically with “buying GPU leaders.” In my view, the true structure of AI infrastructure is a capital expenditure chain: the front end requires computing chips, the middle requires high-bandwidth memory, network connectivity, and optical communications, and the back end requires power, cooling, data centers, and software orchestration. Looking at only one link makes it easy to chase the wrong rhythm when valuations are extremely high. By breaking down the chain, you can understand whether each pullback is merely killing valuation, killing orders, or just a normal washout for high-beta assets.

McKinsey’s projection for data center capital expenditure provides an important backdrop for this framework. It doesn’t mean all companies will benefit simultaneously, nor that all AI-related stocks should rise. Rather, it suggests that if demand for computing power continues to grow, investment opportunities will spread along the “compute – memory – connectivity – optics – power” spectrum. [1] Morningstar’s discussion on the framework for AI stocks also reminds me that AI stock selection shouldn’t just rely on concept hype, but must simultaneously consider industry position, moat, valuation, and uncertainty. [3]

My judgment is that the opportunity in AI infrastructure isn't a single line; it's a network. When the market pulls back, the most worthwhile research subject isn't which ticker fell the most, but which node has an un-falsified fundamental story while its valuation has been unjustly dragged down by risk appetite.

Public price data from the past year shows that these five AI infrastructure stocks have significantly outperformed the Nasdaq 100 and the SMH Semiconductor ETF. The gains for LITE, MU, MXL, VICR, and AMD are substantial, with LITE and MU being the most prominent. However, the same data also reveals that the maximum drawdowns for these five stocks over the past year mostly ranged from approximately -28% to -32%, significantly higher than the Nasdaq 100's maximum drawdown of about -12.1%. [9]

The insight from this data is clear for me: a strong trend does not equal low risk, and high elasticity does not mean you can buy at any time. If a stock has risen several times over a year but can correct by 30% in the process, the investment thesis must go beyond just “bullish on AI long-term” and must clearly address “how to withstand the volatility.” In other words, buying the dip is not an emotional slogan, but a capital management system.

I will use this table as a starting point for position management. For names like MU and AMD, where fundamentals are more validated, I am willing to observe them in batches during pullbacks. For high-elasticity nodes like MXL, LITE, and VICR, I will first set a hard cap on position size before considering price levels. The reason is simple: volatility itself is a cost. “Buying the dip” while ignoring this cost can easily turn into passive holding.

2. Differences Among the Five Stocks: It's Not About Who Rose the Most, But Whose Evidence Chain is More Complete

I do not advocate for a crude comparison of these five companies in the same basket. MU’s core is the memory cycle and AI HBM demand. AMD’s core is the data center computing platform. LITE’s core is cloud and AI optical communications. VICR’s core is high-power server power delivery. MXL is more focused on the AI data center control plane and high-speed connectivity. They all benefit from AI, but their financial flexibility, customer structures, and valuation digestion paths differ.

Based on company public filings, Micron’s FY2025 Q4 press release reported quarterly revenue of $11.315 billion and FY2025 full-year revenue of $37.378 billion, attributing 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 of $4.3 billion, up 22%. Lumentum’s FY2026 Q3 press release reported revenue of $808.4 million, up 90.1% year-over-year, emphasizing photonics technology related to AI, cloud computing, and next-generation communications. MaxLinear’s public press releases introduced its Coronado and Laguna USB UART solutions for AI data center control plane connectivity. Vicor, in its public materials, highlighted the demand for 48V modular power systems driven by AI, HPC, and data center computing growth. [4] [5] [6] [7] [8]

My ranking is not a simple “gain ranking.” If we only look at past year returns, LITE and MU are the most impressive. If we look at the fundamental evidence chain, MU and AMD are more likely to be consistently tracked by institutional capital. If we look at high-elasticity satellite positions, MXL, LITE, and VICR offer steeper return curves but simultaneously demand stricter stop-losses and position limits.

3. Risk-Reward Positioning: The Top Right Corner is Not Paradise, But a Discipline Test

Many investors like to see high-return charts but dislike looking at drawdown charts. My view is the opposite. For high-beta AI stocks, the return rate is merely an outcome; the maximum drawdown is the clause you must accept before entering. Figure 3 plots the past year's return rate and maximum drawdown on the same chart. You can see all five stocks are in the high-return area, but the drawdown on the vertical axis is also deep. This indicates [9]

they are not low-volatility growth stocks, but high-elasticity assets that need to be digested with position discipline.

I use three tiers to handle such stocks. The first tier is “core trackable,” i.e., names with a more complete fundamental evidence chain and better institutional coverage, such as MU and AMD. The second tier is “high-elasticity satellite,” i.e., names with clear industry logic but high volatility, such as LITE and VICR. The third tier is “observational elastic,” i.e., names whose product direction has potential but financial delivery still requires several more quarters of validation, such as MXL.

Therefore, my definition of “buying the dip” is not buying whenever it falls, but when prices pull back, fundamentals haven’t deteriorated, and the capital expenditure chain is still delivering, I will absorb volatility in batches according to a pre-set position rule. Especially for high-volatility names like MXL, LITE, and VICR, position size is more important than entry price.

4. Industry Chain Scoring: Five Stocks Are Not One Trade, But Five Nodes

To avoid lumping all AI stocks into a single concept, I score the five stocks across five dimensions: compute directness, sensitivity to AI capital expenditure, cyclical volatility, valuation delivery pressure, and portfolio diversification value. This scoring is not a return prediction or an investment rating; it helps me determine the specific role each stock plays if constructing an AI infrastructure observation basket.

The insight from this chart is that MU and AMD are more like core evidence assets for the main AI infrastructure narrative. LITE and VICR are more like high-elasticity nodes within the chain that can be easily amplified by capital flows. MXL is more of an observational name where “valuation re-rating might occur after product adoption.” All five have research value, but the investment thesis for each must be completely different.

My allocation thinking is: if you only want core AI exposure, prioritize MU and AMD, which have a more complete evidence chain. If you are willing to take on higher volatility, consider LITE and VICR as satellite observations. If allocating to MXL, one must acknowledge its small-cap nature and income delivery uncertainty, requiring even more restraint on position size compared to the others.

5. Operational Framework: The Real Entry Point Comes When “Pullback, Confirmation, Batches” Occur Simultaneously

I will not treat every pullback as a buying opportunity simply because the AI theme is strong. A truly worthwhile pullback must satisfy at least three conditions simultaneously: First, the price has already released short-term sentiment. Second, the company’s fundamentals haven’t deteriorated concurrently. Third, the portfolio still has cash and risk budget. Missing any one of these turns buying the dip into emotional trading.

Fidelity’s framework on AI bubble risk is worth referencing here. It reminds us that while the AI theme may still be a multi-year cycle, investors must track earnings growth, earnings quality, valuation, capital expenditure sustainability, and the interest rate cycle. [2] I fully agree with this approach. AI is not un-buyable, but you cannot use “long-termism” to cover up short-term risks when valuations are most expensive, sentiment is hottest, and positions are fullest.

In summary, I will place these five stocks in my AI infrastructure watch pool, but I won't treat them all as an equal-weight buy list. For me, the correct order is: first define the role, then define the position, and only then define the price.

6. Conclusion: Dips Can Be Bought, But First Ask Yourself If You Can Withstand the Volatility

The final conclusion returns to the title: “Buying the dip on five major Nasdaq AI leader stocks” can be researched, but you cannot be lazy. If AI data center capital expenditure continues to expand, MU, AMD, LITE, VICR, and MXL – operating in memory, computing, optical communications, power, and connectivity – all have the foundation to continue benefiting. However, if interest rates rise again, cloud capital expenditure slows, AI order fulfillment falls short, or valuations have already pulled forward multiple quarters of growth, these high-beta assets will also correct rapidly.

My strategy is clear: Core positions prioritize assets with a stronger fundamental evidence chain. Satellite positions are for high-elasticity but high-volatility nodes. Observation positions are for small-to-mid-cap opportunities that still require validation. Purchases must be in batches, positions must have limits, and risks must be documented in advance. True mature AI investing isn’t about getting excited over a pullback, but knowing which pullback to buy, how much to buy, and what to do if you’re wrong.

One-sentence summary: The long-term logic for AI infrastructure remains intact, but “buying the dip” is not a charge trumpet; it is a discipline scorecard. First, break down the five stocks into five nodes, then use position sizing and time to digest the volatility.

Risk Disclaimer

This report is for research and discussion purposes only and does not constitute any promise of returns or specific stock buy/sell recommendations. Companies related to AI infrastructure generally possess characteristics of high volatility, high valuation sensitivity, and strong cyclicality. Investors must make independent judgments based on their own risk tolerance. The primary risks to monitor going forward are fivefold: First, if cloud vendor capital expenditure falls short of expectations, orders in the AI hardware chain may be repriced. Second, if interest rates rise again, high-valuation growth stocks will face discount rate pressure. Third, sub-sectors like memory, optical communications, power, and connectivity are subject to inventory cycles and customer concentration risks. Fourth, high-elasticity small-to-mid-cap names may experience amplified liquidity and valuation volatility. Fifth, if AI themes show insufficient earnings delivery, the market could shift from “pricing long-term potential” to “pricing current cash flows.”

This report was prepared by a special analyst. The views expressed in this report represent solely the author's personal stance and do not represent the views of the BIT platform. This material is for reference only and does not constitute investment advice.