Focus: Five Leading AI Stocks on Nasdaq

Roger Lee | BIT U.S. Stock Special Analyst

With 21 years of experience in investment banking, asset management, and financial institutions, focusing on AI industry chain, U.S. macro liquidity, and options strategy research.

Investment Summary

My conclusion is straightforward: these five stocks do not represent the same "AI trade," but rather five different nodes in the AI infrastructure chain; if the market continues to pull back due to inflation, interest rates, or bubble concerns, I will place them on a layered observation list rather than interpreting "buying the dip" as a one-time full position chase. This report discusses MU Micron, MXL MaxLinear, AMD, LITE Lumentum, and VICR Vicor. They all benefit from AI data center capital expenditures, but the sources of risk, performance elasticity, and valuation digestion methods differ.[1] [2] [3]

I believe that after entering this stage of the AI market, 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 valuations? Third, can the portfolio withstand high volatility? McKinsey estimates that to meet computing power demands, global data centers may require about $6.7 trillion in capital expenditures by 2030, with approximately $5.2 trillion related to AI workloads; this indicates that AI infrastructure has a long investment cycle, but Fidelity also reminds us that earnings growth, valuation, sustainability of capital expenditures, and interest rate cycles will determine whether AI trades shift from a long-term theme to a short-term bubble.[1] [2]

In summary: AI infrastructure remains a direction I am willing to research on dips, but entry points must adhere to position discipline; in a phase where high returns, high drawdowns, and high volatility coexist, first layer, then act.

I. First, Look at the Big Picture: AI Infrastructure is Not a Story That Can Be Told by One GPU Stock

The most common mistake in the market is to simply equate the AI market with "buying GPU leaders." In my view, the true structure of AI infrastructure is a capital expenditure chain: the front end requires computing power chips, the middle requires high-bandwidth storage, network connections, and optical communications, and the back end requires power supplies, cooling, data centers, and software scheduling. Focusing on a single link can easily lead to chasing the wrong rhythm when valuations are extremely high; by breaking down the chain, one can understand whether each pullback is killing valuations, killing orders, or just a normal washout of high Beta assets.

McKinsey's estimates on data center capital expenditures provide an important context for this framework. It does not suggest that all companies will benefit simultaneously, nor that all AI-related stocks should rise, but rather indicates that if computing power demand continues to grow, investment opportunities will spread along "computing power --- storage --- connectivity --- optics --- power." [1] Morningstar's discussion on the AI stock framework also reminds me that AI stock selection cannot only focus on concept popularity, but must also consider industry position, moat, valuation, and uncertainty.[3]

|----------------|---------------|---------------------|---------------------| | Industry Chain Segment | Representative Target | Main Beneficiary Logic | Key Validation Points | | HBM and Storage | MU Micron | AI servers drive high bandwidth memory and DRAM demand | Price cycle, gross margin, HBM order continuity | | CPU/GPU and Data Center Computing | AMD | Data center CPU and AI GPU platform expansion | Data center revenue growth rate, customer volume, gross margin | | Control Plane and High-Speed Connectivity | MXL MaxLinear | Increased complexity of internal connections in AI data centers | New product introduction, customer certification, revenue conversion | | Optical Communication and Optical Components | LITE Lumentum | Cloud and AI data centers drive demand for optical components | Cloud customer capital expenditures, optical module demand, capacity utilization | | High-Density Power | VICR Vicor | Increased power consumption in AI servers leads to power delivery upgrades | Orders, modular 48V solution penetration, gross margin recovery |

My judgment is that the opportunities in AI infrastructure are not a "line," but a "network." Once the market pulls back, the most worthwhile research is not which target has fallen the most, but which node's fundamentals have not been disproven while its valuation has been dragged down by risk appetite.

Over the past year, public price data shows that these five AI infrastructure targets have significantly outperformed the Nasdaq 100 and SMH Semiconductor ETF. LITE, MU, MXL, VICR, and AMD have all seen substantial increases, with LITE and MU performing the best; however, the same set of data also shows that the maximum drawdown for these five stocks over the past year is mostly between approximately -28% and -32%, significantly higher than the Nasdaq 100's maximum drawdown of about -12.1%.[9]

This set of data clearly inspires me: strong trends do not equal low risk, and high elasticity does not mean it can be bought at any time. If a stock has risen several times in a year but can pull back by 30% during that process, then the buying logic cannot simply state "long-term bullish on AI," but must also clarify "how to withstand volatility." In other words, buying the dip is not an emotional slogan, but a set of capital management rules.

|---------------|----------|--------|--------|--------|------------------------| | Target | Past Year Return | Maximum Drawdown | Current Drawdown | Annual Volatility | My Interpretation | | LITE Lumentum | 1,017.5% | -28.7% | -12.5% | 86.7% | Strongest elasticity in optical communication, but valuation and order expectations are also the most sensitive | | MU Micron | 751.2% | -30.3% | -9.1% | 69.9% | Storage cycle and AI HBM resonance, suitable for watching EPS realization | | MXL MaxLinear | 623.7% | -29.6% | -17.4% | 108.6% | Small-cap high elasticity is more apparent, must be handled with smaller positions | | VICR Vicor | 595.9% | -32.0% | -12.2% | 84.1% | Clear logic for power nodes, but volatility and order realization need to be tracked | | AMD | 340.4% | -27.8% | -5.7% | 66.9% | Relatively more mature, elasticity lower than small caps, but fundamentals are more thoroughly validated | | Nasdaq 100 | 37.0% | -12.1% | -3.3% | 17.3% | Index volatility is far lower than individual stocks, indicating that individual stocks are not substitutes for the index | | SMH Semiconductor ETF | 142.0% | -14.9% | -2.8% | 33.2% | Sector is strong, but individual stock risks still need to be assessed separately |

I will use this table as a starting point for position management. For targets like MU and AMD, which have stronger fundamental validation, I am willing to observe in batches during pullbacks; for high-elasticity nodes like MXL, LITE, and VICR, I will first set a hard limit on position size before considering price levels. The reason is simple: volatility itself is a cost, and ignoring the cost of "buying the dip" can easily lead to passive holding of positions.

III. Differences Among the Five Stocks: It's Not About Buying the One That Rises the Most, But About Whose Evidence Chain is More Complete

I do not agree with putting these five companies in the same basket for a rough comparison. The core of MU is the storage cycle and AI HBM demand, the core of AMD is the data center computing platform, the core of LITE is cloud and AI optical communication, the core of VICR is high-power server power delivery, and MXL leans more towards AI data center control plane and high-speed connectivity. They all benefit from AI, but their financial elasticity, customer structure, and valuation digestion paths differ.

From the companies' public information, Micron disclosed quarterly revenue of $11.315 billion in its FY2025 Q4 press release, with full-year revenue of $37.378 billion, linking strong performance to AI data center demand; AMD's Q3 2025 press release disclosed quarterly revenue of $9.246 billion, a 36% year-over-year increase, with data center revenue of $4.3 billion, a 22% year-over-year increase; Lumentum's FY2026 Q3 press release disclosed revenue of $808.4 million, a 90.1% year-over-year increase, emphasizing AI, cloud computing, and next-generation communication-related photonic technologies; MaxLinear's public press release introduced its Coronado and Laguna USB UART solutions for AI data center control plane connectivity; Vicor emphasized the demand for 48V modular power systems driven by AI, HPC, and data center computing growth in its public information.[4] [5] [6] [7] [8]

|---------------|-------------|---------------------------|------------------------|-------------------| | Target | My Assigned Role | Core Advantage | Maximum Risk | Suitable Buying Method | | MU Micron | Core Beneficiary of AI Storage Cycle | Resonance of HBM and DRAM cycles, clear path for revenue and gross margin recovery | Reversal of storage price cycle, overly rapid capital expenditure expansion | Observe in batches during pullbacks, focus on performance guidance and gross margin | | AMD | Computing Platform Asset | Sufficient verification from data center CPU and AI GPU customers | Intense competition with leaders, AI GPU volume release rhythm affects valuation | Observe with core position thinking, do not chase short-term surges | | LITE Lumentum | High Elasticity Node in Optical Communication | Strong demand for optical components in cloud and AI data centers | Customer concentration, order volatility, valuation sensitive to expectations | Small position, in batches, only add after pullbacks | | VICR Vicor | Power System Upgrade Node | Increased power consumption in AI servers brings structural demand | Uncertainty in order realization and gross margin recovery | Handle with a satellite position, waiting for orders to continue validating | | MXL MaxLinear | Small-Cap Elasticity in Connectivity and Control Plane | Product entry into AI data center connection complexity increases | Small-cap volatility is large, income conversion timing is uncertain | Only suitable for observation positions within high-risk budgets |

My ranking is not a simple "ranking by increase." If only looking at the past year's increase, LITE and MU are the most eye-catching; if looking at the fundamental evidence chain, MU and AMD are more likely to be continuously tracked by institutional funds; if looking at high-elasticity satellite positions, MXL, LITE, and VICR provide a steeper return curve, but also require stricter stop-loss and position limits.

IV. Risk-Return Position: The Upper Right Corner is Not Heaven, But a Discipline Examination

Many investors like to see high return charts but do not like to look at drawdown charts. My view is quite the opposite: for high Beta AI targets, the return is just the result, and the maximum drawdown is the term that must be accepted before entry. Chart 3 places the past year's return and maximum drawdown on the same chart, showing that all five stocks are in the high return area, but the drawdown on the vertical axis is also deep. This indicates that

they are not low-volatility growth stocks, but high-elasticity assets that need to be digested with position discipline.[9]

I will use three levels to handle such stocks. The first level is "core trackable," meaning targets with more complete fundamental evidence and more institutional coverage, such as MU and AMD. The second level is "high-elasticity satellites," meaning targets with clear industrial logic but very high volatility, such as LITE and VICR. The third level is "observational elasticity," meaning targets with imaginative product directions but financial realization still needing more quarters of validation, such as MXL.

|-------|------------|----------------|-----------------------|-----------------| | Portfolio Level | Example Targets | Position Principles | Conditions for Increasing Positions | Conditions for Reducing Positions | | Core Trackable | MU, AMD | Batch allocation, do not fill in one go | EPS upgrades, stable gross margins, drawdown close to historical pressure zones | Downgrade in earnings guidance, or weakening cloud capital expenditures | | High-Elasticity Satellites | LITE, VICR | Position limits significantly lower than core positions | Enhanced order validation, valuation falls to a bearable range | Single-quarter order or customer demand volatility exceeds expectations | | Observational Elasticity | MXL | Only use small positions for observation, do not use heavy positions for speculation | New product conversion to revenue, cash flow improvement | Product introduction delays, income realization below expectations | | Defensive Buffer | Cash, short bonds, or index hedges | Used to wait for secondary pullbacks | Market continues to decline due to macro shocks | Chasing gains leads to cash cushion disappearing |

Therefore, my definition of "buying the dip" is not just buying when it falls, but when the price pulls back, the fundamentals have not deteriorated, and the capital expenditure chain is still being realized, absorbing volatility in batches according to pre-set position rules. Especially for high-volatility targets like MXL, LITE, and VICR, the size of the position is more important than the buying price.

V. Industry Chain Scoring: The Five Stocks Are Not the Same Trade, But Five Nodes

To avoid mixing all AI stocks into one concept, I score the five stocks across five dimensions: directness of computing power, sensitivity to AI capital expenditures, cyclical volatility, valuation realization pressure, and portfolio diversification value. This scoring is not a return forecast or investment rating, but helps me judge: if I were to create an observation basket for AI infrastructure, what role each stock would play.

This chart inspires me to see that MU and AMD are more like core evidence assets in the AI infrastructure main line; LITE and VICR are more like high-elasticity nodes that can be amplified by funds in the chain; MXL leans more towards being an observational target that may see valuation reassessment after product introduction. All five stocks have research value, but the buying logic must not be completely the same.

|-----------|-------------------------|--------------| | Dimension | What I Value Most | Significance to the Portfolio | | Directness of Computing Power | Whether directly benefits from AI servers, data centers, and high-performance computing | Determines whether thematic relevance is strong enough | | Sensitivity to AI Capital Expenditures | Whether cloud vendors and data center capital expenditures can translate to revenue | Determines order realization elasticity | | Cyclical Volatility | Whether industry prices, customer inventory, and capital expenditure cycles are severe | Determines position limits | | Valuation Realization Pressure | Whether current prices have overdrawn future growth over multiple quarters | Determines buying rhythm | | Portfolio Diversification Value | Whether it forms different sources of risk compared to the GPU main line | Determines whether it is worth including in the observation basket |

My configuration thinking is: if I only want core exposure to AI, prioritize researching MU and AMD with more complete evidence chains; if willing to bear higher volatility, consider LITE and VICR as satellite observations; if configuring MXL, I must acknowledge its small-cap attributes and income realization uncertainties, and the position must be more restrained than the other stocks.

VI. Operational Framework: True Buying Points Come from "Pullback, Confirmation, and Batching" Happening Simultaneously

I will not treat any pullback as a buying point just because the AI theme is strong. The pullbacks that are truly worth acting on must meet at least three conditions: First, the price has released short-term emotions; second, the company's fundamentals have not deteriorated simultaneously; third, there is still cash and risk budget in the portfolio. Missing any one of these will turn buying the dip into emotional trading.

Fidelity's framework on AI bubble risks is worth mentioning here. It reminds us that while the AI theme may still be a multi-year cycle, investors must track earnings growth, earnings quality, valuation, sustainability of capital expenditures, and interest rate cycles.[2] I completely agree with this stance. AI is not uninvestable, but one cannot use "long-termism" to cover short-term risks when valuations are at their highest, emotions are hottest, and positions are fullest.

|---------|---------------------------|----------------| | Buying Conditions | Signals to Look For | What I Will Do | | Pullback Release | Individual stocks pull back close to historical pressure zones from highs, and the index has not experienced systemic collapse | First establish an observation position, do not fill in one go | | Fundamental Confirmation | Financial report revenue, orders, gross margins, EPS guidance have not deteriorated | Gradually convert the observation position into a formal position | | Valuation Repair | Price declines are due to risk appetite rather than earnings falsification | Prioritize buying targets with complete evidence chains | | Macroeconomic Environment Bearable | Interest rates have not risen uncontrollably, liquidity has not tightened sharply | Maintain batching, do not leverage | | Portfolio Still Has Cash | Can withstand a secondary decline after buying | Only use planned funds, do not chase orders due to rebounds |

In summary, I will place these five stocks into the AI infrastructure observation pool, but will not treat them all as an equally weighted buying list. For me, the correct order is to first define roles, then define positions, and finally define prices.

VII. Conclusion: Buying the Dip is Possible, But First Ask Yourself If You Can Withstand Volatility

The final conclusion returns to the title: buying the dip in the five major Nasdaq AI leaders can be researched, but one cannot be lazy. If AI data center capital expenditures continue to expand, MU, AMD, LITE, VICR, and MXL are all positioned to continue benefiting in storage, computing, optical communication, power, and connectivity; however, if interest rates rise again, cloud capital expenditures slow, AI order realizations fall short of expectations, or valuations have already overdrawn future growth over multiple quarters, these high Beta assets will also quickly pull back.

My strategy is clear: prioritize core positions for assets with stronger fundamental evidence chains, satellite positions for high elasticity but high volatility nodes, and observation positions for small and mid-cap opportunities that still need validation. Buying must be in batches, positions must be limited, and risks must be written down in advance. True mature AI investment is not about getting excited when seeing a pullback, but knowing which segment of the pullback can be bought, how much to buy, and what to do if wrong.

In summary: The long-term logic of AI infrastructure remains, but buying the dip is not a charge call, but a discipline table; first break the five stocks into five nodes, then use positions and time to digest volatility.

Risk Warning

This report is for research discussion only and does not constitute any profit commitment or individual stock trading advice. Companies related to AI infrastructure generally have high volatility, high valuation sensitivity, and strong cyclical attributes, and investors need to make independent judgments based on their own risk tolerance. The five types of risks that need to be tracked most in the future are: First, if cloud vendors' capital expenditures fall 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, there are inventory cycle and customer concentration risks in segments like storage, optical communication, power, and connectivity; fourth, small and mid-cap high elasticity targets may experience amplified liquidity and valuation volatility; fifth, if the AI theme experiences insufficient earnings realization, the market may shift from "long-term space pricing" to "current cash flow pricing."

|----------|---------------------|-------------------| | Risk Variable | Observation Signals | Response Principles | | AI Capital Expenditure Slowdown | Downgrade in cloud vendors' capital expenditure guidance, order delays | Reduce high-elasticity satellite positions, retain core evidence assets | | Interest Rates Rising Again | 10-year U.S. Treasury yield spikes, growth stock valuations compress | Do not chase high, wait for valuations to be digested again | | Insufficient Earnings Realization | Financial report revenue, gross margin, or EPS guidance below expectations | First reduce positions, then reassess fundamentals | | Industry Cycle Reversal | Weakening storage prices, optical communication orders, or power demand | Avoid misjudging cyclical downturns as short-term pullbacks | | Individual Stock Liquidity Risk | Small-cap targets see amplified trading but distorted prices | Control positions, avoid concentrated bets |

Data Sources and Citation Notes

The market performance, drawdowns, volatility, and risk-return indicators in this report are sourced from Yahoo Finance's public chart data interface, covering the period from June 13, 2025, to June 12, 2026, including MU, MXL, AMD, LITE, VICR, the Nasdaq Composite Index, the Nasdaq 100 Index, and the SMH Semiconductor ETF. The fundamental narratives of the companies are based on each company's investor relations pages, press releases, and public information; AI capital expenditures, AI bubble risks, and AI stock selection frameworks reference public research materials from McKinsey, Fidelity, and Morningstar. All charts are based on publicly available data, and the chart scoring framework is used for research discussion, not representing profit forecasts or investment ratings.

|-------------|----------------------------------|--------------------------| | Chart/Data Item | Usage Criteria | Main Source | | Performance of Five AI Targets and Indices | Past year's daily closing prices, normalized to starting point 100 | Yahoo Finance Public Chart API | | Drawdown Pressure Chart | Past year's maximum drawdown, current drawdown, and annual volatility | Yahoo Finance Public Chart API | | Risk-Return Matrix | Past year's return and maximum drawdown | Yahoo Finance Public Chart API | | Industry Chain Scoring Heatmap | Directness of computing power, sensitivity to capital expenditures, cyclical volatility, valuation realization pressure, and portfolio diversification value | Company public information, financial report press releases, public market data | | AI Capital Expenditure Background | Global data center capital expenditures and AI workload demand estimates | McKinsey Public Research | | AI Bubble Risk Framework | Earnings, valuation, sustainability of capital expenditures, and interest rate cycles | Fidelity Public Research | | AI Stock Selection Framework | AI stock pool, valuation, moat, and uncertainty | Morningstar Public Research |

References

1. McKinsey & Company, The cost of compute: A $7 trillion race to scale data centers, April 28, 2025.https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-cost-of-compute-a-7-trillion-dollar-race-to-scale-data-centers

2. Fidelity, 5 signs of an AI bubble to watch for, February 10, 2026.https://www.fidelity.com/learning-center/trading-investing/ai-bubble

3. Morningstar, The Best AI Stocks to Buy Now, June 9, 2026.https://www.morningstar.com/stocks/best-ai-stocks-buy-now

4. Micron Technology, Micron Technology, Inc. Reports Results for the Fourth Quarter and Full Year of Fiscal 2025.https://investors.micron.com/news-releases/news-release-details/micron-technology-inc-reports-results-fourth-quarter-and-full-8

5. Advanced Micro Devices, AMD Reports Third Quarter 2025 Financial Results.https://ir.amd.com/news-events/press-releases/detail/1234/amd-reports-third-quarter-2025-financial-results

6. Lumentum, Lumentum Reports Fiscal Third Quarter 2026 Results.https://investor.lumentum.com/news-releases/news-release-details/lumentum-reports-fiscal-third-quarter-2026-results

7. MaxLinear, MaxLinear Enhances Control Plane Connectivity for AI Data Centers.https://investors.maxlinear.com/press-releases/detail/612/maxlinear-enhances-control-plane-connectivity-for-ai-data

8. Vicor, AI, HPC and Data Center Power Delivery Solutions.https://www.vicorpower.com/industries-and-innovations/artificial-intelligence

9. Yahoo Finance Chart API, daily prices for MU, MXL, AMD, LITE, VICR, \^IXIC, \^NDX and SMH, retrieved June 15, 2026. https://finance.yahoo.com/

This report is compiled by a special analyst. The views expressed in this report represent 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.