Why I Just Bought Even More of These 3 Underappreciated AI Stocks

Silicon Motion Technology reported 46% year-over-year revenue growth in Q4 2025 and guided to about 80% growth in Q1 2026, while IREN highlighted a 5-year, $9.7 billion Microsoft deal that implies $1.94 billion in annual recurring revenue from 200 MW. Broadcom posted 28% revenue growth in Q4 FY25, generated $19.3 billion in Q1 2026 revenue, guided to $22 billion for Q2 FY26, and authorized a new $10 billion buyback. The article is broadly constructive on AI-linked names, but it is primarily an opinion piece rather than a fresh market-moving catalyst.

Analysis

The market is underpricing the “picks-and-shovels” layer of AI infrastructure. SIMO is the cleaner second-order beneficiary because its demand is tied to memory content growth per accelerator, not just GPU unit counts; if AI servers continue to move toward higher-capacity storage and more complex NAND subsystems, controller attach rates can expand faster than headline chip demand. That makes SIMO less exposed to AI capex volatility than NVDA/MU, but more dependent on a sustained build cycle in enterprise storage and SSD refreshes.

IREN is the most convex, but also the most fragile. The equity value here is being driven by a long-duration land grab in power, yet the market still appears to be pricing the optionality without fully discounting the financing stack required to monetize it. The hidden risk is that every incremental MW deployed likely needs fresh capital before it contributes meaningfully to earnings, so the nearer-term catalyst is not ARR math but proof that capital efficiency is improving and that hyperscaler demand can be contracted without punitive dilution.

AVGO is the highest-quality expression of AI capex because it combines growth with capital return, which tends to compress volatility relative to the rest of the AI basket. The second-order effect is that buybacks create a floor for multiple support while the custom silicon franchise deepens customer lock-in; that also puts pressure on smaller ASIC players and reinforces a bifurcation where only scaled suppliers with design-in depth can compound. The contrarian concern is not execution but positioning: AVGO may already be the crowded “safe AI” trade, so upside is more likely to come from continued estimate revisions than multiple expansion.

Key Decisions for Investors

• Long SIMO on pullbacks over the next 2-4 weeks; prefer buying only if it holds relative strength vs. MU/SNDK on heavy volume. Risk/reward is attractive because controller attach-rate upside can lag the memory rally by several quarters, allowing the market to re-rate earnings before consensus catches up.
• Trade IREN as a tactical call option, not a core long: buy long-dated upside via calls or a small common position into weakness, but cap size due to dilution and financing risk. Best setup is 3-6 months, with thesis dependent on additional hyperscaler contracts and evidence of lower cost per MW deployed.
• Stay long AVGO as the highest-quality AI compounder, but consider trimming into strength after post-earnings upside if the stock rerates faster than estimates. The reward remains steady compounding, but near-term upside is more limited than the less-owned names.
• Pair long AVGO / short NVDA for a 3-6 month relative-value trade if AI capex remains healthy but custom silicon keeps taking share from general-purpose accelerators. The trade benefits if investors continue rotating toward cash-generative infrastructure names and away from GPU beta.
• For contrarian exposure, avoid chasing IREN after large green candles; instead wait for a financing-related drawdown to establish exposure. The stock’s biggest risk is not demand but equity issuance, and that risk tends to surface after momentum traders have already piled in.