This ETF Has Quietly Become One of the Best Ways to Invest in Artificial Intelligence

Goldman Sachs projects total AI spending will reach $1 trillion in 2026, with roughly 60% occurring in the U.S., supporting the view that the AI investment cycle remains in an expansion phase. The article favors the Global X Artificial Intelligence & Technology ETF (AIQ) as a diversified AI allocation: it holds 88 stocks, with 33% of assets in its top 10 positions and AMD representing only 3%. The thesis is that rapidly rotating AI leadership—from hyperscalers and semiconductors to memory and cybersecurity—makes broad thematic exposure preferable to concentrated chip-stock bets.

Analysis

The investable implication is not that a diversified AI ETF is inherently superior, but that dispersion is likely to rise as the market shifts from capex beneficiaries to monetization beneficiaries. NVDA, AMD and MU remain most exposed to hyperscaler capital-intensity assumptions; software and IT-services beneficiaries require proof that enterprise AI converts pilots into recurring revenue. AIQ's broad construction may dilute single-name downside, but it also introduces substantial exposure to low-quality “AI adjacency” holdings whose earnings sensitivity is far weaker than the narrative suggests.

Over the next 1-3 months, hyperscaler earnings and 2027 capex commentary remain the critical catalyst for the semiconductor complex. A capex guide-down would pressure NVDA/MU disproportionately, while beneficiaries of lower inference costs and enterprise deployment—MSFT, ORCL, CRM, NOW and cybersecurity vendors such as PANW—could relatively outperform if management teams demonstrate AI-driven seat expansion or reduced service costs. The key falsifier for a rotation thesis is continued upward revisions to accelerator shipments and HBM pricing alongside weak software AI monetization; that outcome favors staying concentrated in compute rather than broadening exposure.

The contrarian view is that diversification is being marketed as protection after leadership rotation, when it may instead be a late-cycle way to buy lower-beta laggards at elevated thematic multiples. The more attractive expression is selective dispersion: own the infrastructure names with independently visible supply constraints and pair them against software names where AI revenue remains largely embedded in guidance rather than separately disclosed. Over 6-18 months, the durable winners should be firms able to retain AI productivity gains in margins, not merely those spending most aggressively on capacity.

Key Decisions for Investors

• Maintain a 1-3 month relative-value pair: long NVDA / short a basket of unproven AI software exposure via IGV. Use a 10-12% stop on the relative spread; target 15-20% relative return if accelerator demand revisions remain positive while software monetization disclosures disappoint.
• Do not initiate a tactical AIQ position solely on diversification rhetoric. Place it on watch pending holdings-level exposure, expense ratio, liquidity and overlap analysis versus BOTZ, SOXX and IGV; absent that work, the article provides no edge sufficient to justify a trade.
• For MU, remain constructive only while HBM contract pricing and bit-demand forecasts support further earnings revisions. A 2027 capex reduction by major cloud buyers or evidence of easing HBM supply tightness would invalidate the thesis quickly; use defined-risk calls rather than unhedged equity after sharp momentum rallies.
• Build a 6-18 month adoption basket in MSFT, ORCL, NOW and PANW only following measurable evidence of AI-related ARR, backlog conversion, or gross-margin improvement. Pair against AMD if accelerator share gains fail to translate into upward data-center revenue revisions; AMD's 3% AIQ weight does not make the ETF a meaningful hedge.