Forget Betting Everything on NVIDIA. This Chip Fund Rode the Same AI Boom and Is Up 79% in 2026

NVIDIA is up 11.35% year-to-date through June 16, 2026, but the article argues the broader semiconductor stack has outperformed, with VanEck Semiconductor ETF (SMH) up 79.69% YTD. NVIDIA still posted strong underlying growth, including 92% data center revenue growth to $75.246B and 199% networking growth, but zero H20 shipments to China and the exclusion of China data center compute from Q2 outlook highlight export-license risk. The piece recommends a partial rotation from concentrated NVDA exposure into SMH to preserve AI exposure while diversifying across foundry, memory, networking, and custom silicon beneficiaries.

Analysis

The key shift is that AI capex is broadening from a single-vendor bet into a multi-node industrial cycle. That matters because the next leg of returns likely accrues to the bottlenecks, not the brand leader: foundry capacity, advanced packaging, HBM memory, networking silicon, and custom accelerators should capture more of the dollar-per-watt spend as hyperscalers diversify supply chains. In other words, the trade is moving from "who sells the best GPU" to "who owns the scarcest inputs".

NVIDIA’s China exposure is the cleanest source of asymmetry: export restrictions create a binary overhang that can reprice the stock in days, while upside from the core AI thesis is now more likely to be earned in months and can be diluted by multiple compression. The second-order risk is that investors anchoring on NVDA as the AI proxy are effectively taking a hidden regulatory macro bet; as that risk becomes more visible, capital should migrate toward beneficiaries with less policy sensitivity and more direct leverage to capex throughput.

The contrarian point is that the broad chip basket may be the better expression of the same thesis, but it is not a free lunch. If AI spend pauses even briefly, the high-beta beneficiaries in the stack can de-rate faster than NVDA because their margins and valuation support are more cyclical. So the right framing is not abandoning NVIDIA, but reducing concentration and letting the rest of the ecosystem finance the same theme with better diversification and less headline risk.

Key Decisions for Investors

• Trim 25-40% of oversized NVDA exposure and rotate proceeds into SMH over the next 1-3 trading sessions; the goal is to preserve AI upside while cutting single-name/export-control risk. Risk/reward improves if NVDA suffers a 5-10% regulatory drawdown while the basket is cushioned by broader capex participation.
• Prefer a barbell within semis: long SMH or a basket of AMD/AVGO/INTC over the next 3-6 months, with emphasis on names levered to custom silicon, networking, and foundry demand. This captures the second-order beneficiaries of AI spending that NVDA does not monetize directly.
• For tactical hedging, buy 1-2 month NVDA put spreads around known policy/event windows if China license headlines remain active. The setup is asymmetric because downside can gap on headlines while upside is typically more gradual.
• In taxable accounts, compare the after-tax cost of a partial NVDA swap versus using new cash to build SMH. If embedded gains are large, incremental buying of SMH may be the cleaner expression unless the tax drag is less than roughly one-third of expected diversification benefit.
• Avoid chasing the strongest semiconductor names indiscriminately; if AI capex sentiment cools, the highest-multiple beneficiaries can correct 15-20% faster than NVDA. Size any basket exposure with a 3-6 month horizon and predefine a drawdown stop.