IGPT: The Next Phase Of AI Should Favor The Companies Building It

The article rates the Invesco AI & Next Gen Software ETF (IGPT) a BUY, arguing the next phase of the AI cycle should continue to favor semiconductor, memory, and AI infrastructure names. It highlights the fund’s heavy AI infrastructure exposure (notably NVIDIA, AMD, Micron, Alphabet, and Meta) that is expected to benefit from ongoing AI-related capex over the next 6–12 months. The piece flags higher concentration/volatility as added risk, but deems it justified given strong AI investment tailwinds.

Analysis

This is less a broad AI call than a bet that the capex mix keeps shifting toward the picks-and-shovels. The immediate beneficiaries are still the obvious compute owners, but the second-order winners are memory, networking, packaging, and power/cooling suppliers: the bottleneck moves outward from model training to the entire deployment stack. That favors NVDA and AMD on near-term order momentum, but also names outside the basket such as MU, AVGO, TSM, AMAT, and KLAC if hyperscaler spend stays elevated.

The market risk is that investors are already paying for a long runway of AI infrastructure growth, so the next upside leg likely requires either another capex re-acceleration or evidence that returns on invested capital are improving. If hyperscaler budgets flatten for even one quarter, the hardware complex can de-rate quickly because multiples have outrun near-term revenue visibility. The first sign of trouble would be softer cloud capex commentary, longer lead times normalizing, or memory pricing failing to inflect.

Contrarianly, the consensus may be underestimating how narrow this trade has become: if AI spend broadens from model training into inference and enterprise deployment, semis still win, but the biggest relative upside may migrate to components that are not the ETF’s headline names. That argues for owning the infrastructure leaders directly rather than paying an ETF wrapper fee for a concentrated basket that includes some lower-beta names. If NVDA guide, AMD data center ramp, or hyperscaler capex revisions disappoint, the trade likely pauses for 1-3 months even if the 6-18 month secular thesis remains intact.

Key Decisions for Investors

• Long NVDA and AMD into pullbacks, using a 3-6 month horizon; prefer single names over IGPT if the goal is pure AI capex beta, with thesis invalidation on any downward revision to hyperscaler capex or data center gross margin commentary.
• Pair trade: long SMH / short IGV for 1-3 months to express continued spend rotation from software experimentation to infrastructure buildout; exit if enterprise software revenue re-accelerates or AI software monetization becomes visible.
• Add a second-order beneficiary basket on weakness: long MU and/or AMAT as memory and wafer-fab gear lag the headline GPU trade but gain if AI capex remains broad; stop if DRAM/NAND pricing rolls over or order lead times compress.
• Use IGPT only as a tactical vehicle, not a core holding: buy on market-wide tech pullbacks rather than breakouts, since concentration increases drawdown risk and the wrapper may underperform direct semis in strong tape.