Is It Really Safe to Invest in AI Stocks and ETFs Right Now? Here's What the Experts Say.

AI stocks remain under pressure over fears of a potential slowdown in AI infrastructure spending, but the article cites continued capex momentum. TSMC raised 2026 capex to $52B–$56B and Citi expects AI data center capex to rise another 40%–50% next year to about $1.5T, supported by AI agent adoption, token growth, and supply constraints. Overall, it argues investors should favor AI leaders (e.g., NVDA, Alphabet, Meta, AMD, Broadcom, TSMC) and maintain long-term exposure via diversified vehicles like QQQ and VOO.

Analysis

The market is still pricing AI as a demand story, but the cleaner edge is on supply discipline. TSM is the best read-through because higher capex only matters if it translates into sustained utilization and pricing power; that is much harder to fake than hyperscaler commentary, and it should keep margins sticky even if end-demand growth normalizes. By contrast, AMD is the most narrative-sensitive name here: if inference/agentic adoption is real but not explosive, its multiple can compress faster than its earnings catch up.

Second-order winners are the infrastructure bottlenecks, not the broad "AI basket." NVDA remains the highest-quality beneficiary if deployment stays constrained, but the real convexity sits with the names that control scarce capacity or custom design wins; AVGO should outperform if hyperscalers keep outsourcing silicon architecture. META and GOOG are less pure capex trades: any externalizing of excess compute is only bullish if it improves asset turns, otherwise it is a sign internal demand is saturating and margins can get diluted by lower-return third-party load.

The key risk is that capex headlines remain strong while bookings, utilization, or enterprise willingness to pay quietly roll over over 1-3 quarters. That would hit high-multiple AI proxies first, then force the market to differentiate between true supply-side winners and simply "AI exposed" names. The contrarian view is that consensus may be underestimating the duration of spending, but overestimating breadth; if spend decelerates, the leaders stay leader-like, while second-tier beneficiaries give back the most.

Key Decisions for Investors

• Long TSM / short AMD for the next 1-3 months: TSM has the clearest pricing-power and utilization leverage, while AMD is more vulnerable to a narrative reset if AI wallet share grows slower than hoped. Stop if AMD starts outperforming TSM by >10% on sustained booking evidence.
• Add to NVDA on pullbacks rather than breakouts: best risk/reward remains the dominant platform exposure, but only if supply remains constrained. Falsify the thesis if gross margin or datacenter growth inflects down for two consecutive quarters.
• Overweight AVGO versus the broader Nasdaq-100 over 3-6 months: custom silicon and design-out activity should be a relative winner if hyperscalers keep spending, and the business model is less exposed to end-market hype than pure GPU beta.
• Use QQQ as a hedge, not a core AI expression: if AI leadership broadens, QQQ participation is enough; if the AI trade is crowded and rolls over, QQQ gives portfolio-level protection without forcing single-name timing risk.