Had You Bought This Magnificent Vanguard ETF at the Start of January, You'd Be Crushing the S&P 500 in 2026

Vanguard Information Technology ETF (VGT) outperformed the S&P 500 in 2026 with a 23.3% gain vs the benchmark’s 10.3% YTD, driven by heavy concentration in AI-chip and mega-cap tech (50.6% of assets in NVDA, AAPL, MSFT, AVGO, MU). However, the article flags near-term AI-spend headwinds: UBS survey data suggests 60% of businesses are curbing AI spending by using cheaper, less compute-intensive models, which could pressure semiconductor demand. Net-net, long-term conviction is maintained given sustained data-center/AI infrastructure capex, but timing risk remains for the next leg of semiconductor growth.

Analysis

This is less a clean “buy tech” signal than a concentration warning disguised as sector strength. A handful of AI-capex winners are carrying the entire basket, so the marginal buyer is effectively underwriting continued hyperscaler spending plus sustained pricing power in the semiconductor supply chain. That works until customers start optimizing usage: then the pain shows up first in the highest-duration names and in the second-order beneficiaries of model consumption, not in the broad index.

Near term, the most important catalyst is not macro but earnings commentary over the next 1-3 months: capex guides, backlog conversion, and any evidence that cloud/enterprise customers are trading down to smaller models. If that behavior persists, the market will likely move from “scarcity premium” to “digestion phase,” which typically compresses multiples before it hits revenue growth. NVDA and AVGO remain the cleanest expressions of continued infrastructure spend, while MU and LRCX are more vulnerable to a pause in ordering because their upside depends on the cycle staying tight.

The contrarian point is that AI demand may still be real even if the spend mix changes materially. Cheaper models and efficiency gains can slow unit intensity enough to cap semiconductor upside while benefiting the application layer and large platforms with distribution, particularly MSFT and GOOGL, which can absorb AI into broader bundles. In other words, the trade may be rotating from pure picks-and-shovels to whoever monetizes AI without having to buy every incremental GPU; the consensus is still treating that as a later-year story, but the budget pushback is already visible.

Key Decisions for Investors

• Prefer a hedged expression: long NVDA against short GOOGL for the next 1-3 months. If capex stays hot, NVDA should outperform; if enterprise AI budgets keep tightening, GOOGL is more exposed to customer pushback and pricing friction. Falsify if hyperscaler capex guides reaccelerate materially next quarter.
• Use strength to reduce passive exposure to VGT rather than add to it. The ETF’s performance is being driven by extreme concentration, so the incremental upside from here is less broad sector beta and more a few crowded names. Re-enter only on a 5%-8% pullback or after earnings confirm sustained spend.
• Favor AVGO over MU for a cleaner risk/reward. AVGO has more software-like pricing power and less direct inventory-cycle sensitivity; MU is more levered to any digestion in AI hardware orders. This is a 3-6 month relative-value trade, not a secular call.
• Set a watch item on LRCX and AMD after the next data-center commentary cycle. If bookings or forward demand language softens, rotate out of semicap names quickly; if guidance remains firm, those names offer the highest upside convexity off continued supply tightness.