The SPDR S&P Semiconductor ETF (NYSEARCA:XSD) and the VanEck Semiconductor ETF (NASDAQ:SMH) both offer exposure to semiconductors, yet they are two very different bets. XSD spreads chips money almost evenly across roughly 40 U.S. semiconductor names. SMH funnels most of its dollars into the 25 largest, tilting hard toward the AI mega-caps. Over the last five years that structural choice produced a 375.62% return for SMH versus 197.24% for XSD. In 2026 so far, the ranking has flipped.
What Each Fund Is Actually Betting On
SMH is a concentrated wager that the mega-cap winners keep winning. Its top five holdings are AMD at 10.33%, Broadcom at 9.57%, Micron at 9.39%, Taiwan Semiconductor at 8.75%, and NVIDIA at 8.40%. Add ASML and Intel and roughly 55% of the portfolio sits in seven names tied to AI training, advanced packaging, and leading-edge lithography. If hyperscaler capex stays elevated and the biggest fabs keep pricing power, SMH captures nearly all of that upside.
XSD is the opposite thesis: breadth wins when the AI trade broadens. Its modified equal-weight index caps positions near parity, so the top names, Marvell Technology at 3.06%, Power Integrations at 3.05%, and Cirrus Logic at 2.98%, sit next to analog, power, RFID, and FPGA suppliers like Silicon Labs, Lattice, and Impinj. That construction bets on small and mid-cap semis catching up as edge AI, automotive silicon, and industrial chip cycles recover.
Where the Difference Shows Up
The divergence is visible in this year’s performance. XSD is up 70.75% year to date, beating SMH’s 64.66%, because small and mid-cap chip names finally caught a bid after two years of lagging NVIDIA and Broadcom. Over one year the funds are essentially tied at 108.61% for XSD and 109.83% for SMH.
Zoom out and the mega-cap bet dominates. SMH has returned 2,146.44% over ten years against 1,179.03% for XSD. That gap is the AI decade in a single number: owning NVIDIA and TSMC at heavy weights was worth roughly double an equal slice of the industry. The last week shows the risk pattern in miniature, with XSD down 7.95% versus 4.43% for SMH as smaller chip names sold off harder.
The Practical Comparison
| Factor | XSD | SMH |
|---|---|---|
| Weighting | Modified equal-weight | Market-cap weighted |
| Top holding weight | 3.06% (MRVL) | 10.33% (AMD) |
| Expense ratio | 0.35% | 0.35% |
| 5-year return | 197.24% | 375.62% |
| Foreign exposure | U.S. only | Includes TSM, ASML |
Fees are a wash. The real cost difference is behavioral: SMH rebalances less aggressively, letting winners run, which is efficient in a trending AI market but leaves you exposed if one or two mega-caps roll over. XSD trims winners and adds to laggards at each rebalance, damping single-stock risk but cutting the biggest compounders down to size. SMH also gives you Taiwan Semiconductor and ASML, foreign names XSD excludes, which matters if you want direct exposure to the foundry and lithography bottleneck. For a broader look at the AI infrastructure names driving these funds, see our Next Nvidia Playbook.
The Verdict
SMH is the right choice for investors who believe the AI cycle runs through a handful of dominant chip and equipment names, and who want the TSMC and ASML exposure that pure U.S. funds lack. XSD fits investors who think the rally needs to broaden, want less single-stock risk, and are willing to give up some AI beta to own the analog, power, and specialty names that lead when the cycle turns. If NVIDIA and Broadcom keep setting the pace, SMH wins again. If 2026’s early broadening holds and mid-cap semis lead, XSD keeps closing the gap.
Contact [email protected] for any questions or corrections.
Top Gaining Stocks
Top Losing Stocks