My conclusion is straightforward: these five stocks are not part of a single “AI trade,” but rather five distinct nodes along the AI infrastructure chain. If the market continues to pull back due to concerns over inflation, interest rates, or bubbles, I would place them on a tiered monitoring list rather than interpreting “buy the dip” as an opportunity to go fully long all at once. This report examines MU (Micron), MXL (MaxLinear), AMD (AMD), LITE (Lumentum), and VICR (Vicor). While they all benefit from AI data center capital expenditures, their sources of risk, earnings elasticity, and methods of valuation absorption differ significantly. [1] [2] [3]

I believe that now that AI markets have entered this phase, what truly matters is not whether AI still has a story, but three key questions: First, can capital expenditures continue to translate into real orders? Second, can corporate profitability justify valuations? Third, can investment portfolios withstand high volatility? McKinsey estimates that to meet computing demands, global data centers may require approximately $6.7 trillion in capital expenditures by 2030, with about $5.2 trillion related to AI workloads; this indicates that AI infrastructure involves a long investment horizon. However, Fidelity also cautions that profitability growth, valuations, the sustainability of capital expenditures, and interest rate cycles will determine whether AI investing evolves from a long-term theme into a short-term bubble. [1] [2]

Final takeaway: AI infrastructure remains a direction I’m willing to research on dips, but entry points must adhere to position sizing discipline; in this phase of high yield, high drawdown, and high volatility, prioritize layering before acting.

First, look at the big picture: AI infrastructure is not a story that can be told with just one GPU stock.

The most common mistake in the market is equating AI trends simply with “buying the leading GPU stocks.” In my view, the true structure of AI infrastructure is a capital expenditure chain: the front end requires computing chips, the middle layer needs high-bandwidth memory, network connectivity, and optical communications, and the back end demands power supply, cooling, data centers, and software orchestration. Focusing on just one segment makes it easy to misjudge timing when valuations are already extremely high. Only by breaking down the entire chain can you understand whether each pullback is driven by valuation compression, order cancellations, or simply normal consolidation in high-beta assets.

McKinsey’s analysis of data center capital expenditures provides important context for this framework. It does not suggest that all companies will benefit simultaneously or that all AI-related stocks should rise; rather, it indicates that if computing power demand continues to grow, investment opportunities will spread along the chain of “computing power—storage—connectivity—optics—power supply.”[1] Morningstar’s discussion of the AI stock framework also reminds me that selecting AI stocks cannot rely solely on conceptual hype, but must also consider industry positioning, moat, valuation, and uncertainty.[3]

My assessment is that the opportunity in AI infrastructure is not a "line," but a "network." When the market retraces, what’s most worth examining isn’t which asset has fallen the most, but which node’s fundamentals remain intact while its valuation has been dragged down along with risk appetite.

Over the past year, publicly available price data shows that all five AI infrastructure assets significantly outperformed the Nasdaq 100 and the SMH Semiconductor ETF. LITE, MU, MXL, VICR, and AMD all posted strong gains, with LITE and MU showing the most outstanding performance; however, the same data also reveals that the maximum drawdowns for these five stocks over the past year generally ranged between -28% and -32%, significantly higher than the Nasdaq 100’s maximum drawdown of approximately -12.1%. [9]

This data set clearly inspired me: a strong trend does not equal low risk, and high volatility does not mean you can buy at any time. If an asset has risen several-fold over a year but experienced a 30% drawdown along the way, your rationale for buying cannot simply be “long-term bullish on AI”—you must also clearly articulate “how to withstand the volatility.” In other words, buying on dips is not an emotional slogan; it’s a capital management strategy.

I will use this table as a starting point for position management. For assets like MU and AMD with stronger fundamental validation, I’m willing to observe and average down during drawdowns; for high-elasticity names like MXL, LITE, and VICR, I’ll first set a fixed position cap before considering price levels. The reason is simple: volatility itself is a cost—ignoring this cost and “buying the dip” can easily lead to being stuck with losing positions.

Three: Differences Among the Five Stocks—It’s Not About Buying the One That Rises the Most, But the One With the Most Complete Evidence Chain

I don’t agree with crudely comparing these five companies as if they belong in the same category. MU’s core lies in memory cycles and AI HBM demand; AMD’s core is data center computing platforms; LITE’s core is cloud and AI optical communications; VICR’s core is high-power server power delivery; and MXL is more focused on AI data center control planes and high-speed connectivity. While all benefit from AI, their financial resilience, customer structures, and valuation digestion paths differ significantly.

According to public company disclosures, Micron reported quarterly revenue of $11.315 billion and annual revenue of $37.378 billion for FY2025 in its Q4 press release, attributing its strong performance to demand from AI data centers; AMD reported Q3 2025 revenue of $9.246 billion, a 36% year-over-year increase, with data center revenue reaching $4.3 billion, up 22% year-over-year; Lumentum reported Q3 FY2026 revenue of $808.4 million, a 90.1% year-over-year increase, highlighting photonics technologies related to AI, cloud computing, and next-generation communications; MaxLinear’s public press releases detail its Coronado and Laguna USB UART solutions for AI data center control plane connectivity; Vicor, in its public materials, emphasizes the growing demand for 48V modular power systems driven by increased computing power in AI, HPC, and data centers. [4] [5] [6] [7] [8]

My ranking is not simply based on "price increase." If you look only at the past year's gains, LITE and MU stand out; if you consider fundamental evidence, MU and AMD are more likely to be consistently tracked by institutional capital; if you're seeking high-elasticity satellite positions, MXL, LITE, and VICR offer steeper return curves, but also demand stricter stop-losses and position limits.

IV. Risk-Reward Position: The Top-Right Corner Is Not Heaven, But a Test of Discipline

Many investors enjoy seeing high-return charts but dislike looking at drawdown charts. My view is the opposite: for high-Beta AI assets, returns are merely the outcome, while maximum drawdown is a term you must accept before entering. Figure 3 plots both the one-year returns and maximum drawdown on the same chart, revealing that all five stocks lie in the high-return zone, yet their drawdowns on the vertical axis are also substantial. This indicates

They are not low-volatility growth stocks, but high-elasticity assets that require position discipline to absorb. [9]

I categorize these stocks into three tiers. The first tier is “Core Trackable,” comprising stocks with more comprehensive fundamental data and stronger institutional coverage, such as MU and AMD. The second tier is “High-Elasticity Satellites,” featuring stocks with clear industry narratives but high volatility, such as LITE and VICR. The third tier is “Observational Elasticity,” including stocks with promising product directions but requiring additional quarters of financial validation, such as MXL.

Therefore, my definition of "buy the dip" is not simply buying whenever the price falls, but rather gradually accumulating positions according to pre-defined allocation rules when the price retraces, fundamentals remain intact, and capital expenditure pipelines continue to deliver. For highly volatile assets like MXL, LITE, and VICR, position size is more important than entry price.

Five. Industry Chain Score: Five shares are not a single transaction, but five nodes.

To avoid treating all AI stocks as a single concept, I scored five stocks across five dimensions: direct compute exposure, sensitivity to AI capital expenditures, cyclical volatility, valuation realization pressure, and portfolio diversification value. This scoring is not a return forecast or an investment rating, but rather a tool to help me understand what role each stock plays in an AI infrastructure observation basket.

This chart suggests that MU and AMD are more like core evidence assets in the AI infrastructure theme; LITE and VICR resemble high-elasticity nodes in the chain that are easily amplified by capital; MXL, on the other hand, is more of an observational target tied to potential valuation re-rating after product adoption. All five stocks hold research value, but their investment rationales must differ significantly.

My allocation strategy is: if you seek only AI core exposure, prioritize researching MU and AMD, which have more comprehensive evidence trails; if you're willing to accept higher volatility, consider LITE and VICR as satellite positions; if you choose to allocate to MXL, you must acknowledge its small-cap status and uncertain revenue realization, and maintain a more restrained position size compared to the others.

Six: Operation Framework – The true buying opportunity arises when retracement, confirmation, and dollar-cost averaging occur simultaneously.

I won’t treat every pullback as a buying opportunity just because the AI theme is strong. A pullback worth acting on must satisfy at least three conditions simultaneously: First, the price has already released short-term sentiment; second, the company’s fundamentals have not deteriorated in tandem; third, the portfolio still has cash and risk budget available. Missing any one of these turns buying the dip into an emotional trade.

Fidelity’s framework on the risks of an AI bubble is worth highlighting here. It reminds us that while the AI theme may still unfold over a multi-year cycle, investors must monitor earnings growth, earnings quality, valuation, capital expenditure sustainability, and the interest rate cycle. [2] I fully agree with this perspective. AI is not unbuyable—it’s just that you shouldn’t use “long-termism” to mask short-term risks when valuations are at their peak, sentiment is overheated, and positions are fully loaded.

In summary, I’ll add these five stocks to my AI infrastructure watchlist, but I won’t treat them as equally weighted buy candidates. For me, the correct order is to define the role first, then the position size, and finally the price.

Seven: Conclusion—Buy on dips, but first ask yourself if you can withstand the volatility.

The bottom line returns to the title: Buy the dip in the five leading Nasdaq AI stocks—do your research, but don’t get lazy. If capital spending on AI data centers continues to expand, MU, AMD, LITE, VICR, and MXL—covering storage, computing, optical communications, power, and connectivity—are well-positioned to benefit further. However, if interest rates rise again, cloud capital spending slows, AI orders fall short of expectations, or valuations have already priced in multiple quarters of future growth, these high-beta assets could quickly correct.

My strategy is clear: allocate the core position to assets with stronger fundamental evidence chains, satellite positions to high-flexibility but high-volatility nodes, and observation positions to smaller-cap opportunities still requiring validation. Purchases must be made in batches, positions must be capped, and risks must be written down in advance. True mature AI investing isn’t about getting excited when prices pull back—it’s about knowing which pullback to buy, how much to buy, and what to do if you’re wrong.

The long-term logic behind AI infrastructure remains intact, but buying on dips is not a call to charge—it’s a discipline; first, break down the five stocks into five distinct nodes, then manage volatility through position sizing and timing.

This report is for research and discussion purposes only and does not constitute any promise of returns or recommendations to buy or sell individual stocks. Companies related to AI infrastructure generally exhibit high volatility, high valuation sensitivity, and strong cyclicality; investors should make independent judgments based on their own risk tolerance. The five key risks to monitor going forward are: first, if cloud providers' capital expenditures fall below expectations, orders along the AI hardware supply chain may be repriced; second, if interest rates rise again, highly valued growth stocks may face pressure from higher discount rates; third, niche segments such as storage, optical communications, power supply, and connectivity face risks related to inventory cycles and customer concentration; fourth, small- and mid-cap stocks with high elasticity may experience amplified liquidity and valuation fluctuations; fifth, if AI-themed companies fail to deliver sufficient earnings realization, the market may shift from pricing based on long-term potential to pricing based on current cash flows.

The market performance, drawdown, volatility, and risk-return metrics in this report are compiled and sourced from publicly available data via the Yahoo Finance charting interface, covering the period from June 13, 2025, to June 12, 2026, and including MU, MXL, AMD, LITE, VICR, the Nasdaq Composite Index, the Nasdaq 100 Index, and the SMH Semiconductor ETF. Company fundamental narratives are based on each company’s investor relations pages, press releases, and public disclosures. AI capital expenditure, AI bubble risks, and AI stock selection frameworks are informed by publicly available research from McKinsey, Fidelity, and Morningstar. All charts are compiled from public data; the chart scoring framework is intended for research and discussion purposes only and does not constitute a yield forecast or investment rating.