Rates Could Reach 4.5% On AI Spending, And Semiconductors Get Paid First

iShares Semiconductor ETF and VanEck Semiconductor ETF remain rated Buys as AI-driven infrastructure investment fuels unusually strong semiconductor capital spending and demand. The article argues semiconductor demand is outpacing most of the broader market, while the scale of AI-related investment may raise the U.S. neutral interest rate and redirect economy-wide capital flows.

Analysis

The actionable distinction is not broad semiconductor beta but exposure to the AI capex bottlenecks. SMH is more concentrated in NVIDIA, TSMC and Broadcom, while SOXX offers relatively broader exposure; that makes SMH the higher-conviction expression if hyperscaler accelerator procurement remains the dominant spend category, but also the more vulnerable vehicle if custom silicon captures share or a single leading platform disappoints. Second-order beneficiaries extend beyond compute chips to ASML, AMAT and LRCX, where sustained fab investment converts demand visibility into multi-quarter tool backlog and service revenue.

The claimed link between AI investment and a higher neutral rate should not be traded mechanically. Near term, rising real yields can compress semiconductor multiples even while earnings estimates rise; this creates a regime where the group can outperform on fundamentals but underperform on index-relative price action. Over the next 1-3 months, the key catalyst is whether hyperscaler capex guidance broadens from a small set of buyers into enterprise and sovereign demand; over 6-18 months, the larger risk is that AI revenue remains concentrated at the chip layer while utilization, pricing and returns on invested capital disappoint for cloud customers.

Consensus is likely underpricing the transition from GPU-led spending to heterogeneous infrastructure: networking, custom ASICs, HBM memory, advanced packaging and power delivery. That favors AVGO and TSM over a pure one-name accelerator bet, while MU is the highest-beta beneficiary if HBM supply remains tight. The thesis is falsified by simultaneous cuts to hyperscaler capex guidance, weakening foundry utilization, or evidence that accelerator lead times normalize without a corresponding increase in end-user AI monetization.

Key Decisions for Investors

• Use a 3-6 month long SMH / short SOXX relative-value position rather than unhedged sector beta if the objective is continued AI-infrastructure leadership; SMH should outperform if accelerator, foundry and networking concentration remains rewarded. Exit if SMH/SOXX breaks below its 200-day moving average and NVIDIA/TSMC guidance no longer supports premium growth.
• Build a diversified AI supply-chain basket over 1-3 months: long AVGO, TSM and AMAT in equal risk weights. This captures custom compute/networking, leading-edge fabrication and equipment backlog, reducing dependence on a single GPU vendor; reassess after each major hyperscaler earnings cycle.
• Keep MU as a tactical, smaller-sized 3-6 month long only if reported HBM pricing and supply commitments continue improving. Upside comes from operating leverage to memory tightness; risk is rapid conventional-DRAM supply response or HBM qualification delays, which would make the position unsuitable as a core AI holding.
• Do not add broad semiconductor exposure solely on bullish AI narratives until positioning and valuation data are checked: monitor fund flows, forward EPS revisions, semiconductor equipment bookings and hyperscaler capex guidance. If EPS revisions flatten while real yields rise, favor reducing SMH beta or pairing it with a short QQQ rather than adding outright.