I've been watching the layout of AI-related ETFs, and I find the product design ideas in this sector quite interesting. Basically, there are two main camps: one is holding core funds of AI companies directly, and the other is derivatives around AI infrastructure.

For core AI ETFs, Global X's AIQ tracks the Indxx AI Index, with a fee of 0.68%, managing about $6 billion in assets. If you want a cheaper option, BlackRock's ARTY charges only 0.47%, and both funds have similar holdings, mainly cloud computing, software platforms, and chip design companies. A more aggressive choice is Roundhill's CHAT, which is actively managed, with a fee of 0.75%, focusing on generative AI concept stocks, suitable for those wanting to make concentrated bets.

Expanding outward, chip infrastructure is unavoidable. iShares' SOXX and VanEck's SMH track the semiconductor supply chain, with fees of 0.34% and 0.35%, respectively, both passive index funds. VanEck's fund has over 70% of its top ten holdings, with heavy weights in Nvidia, TSMC, Broadcom, etc. In comparison, iShares' diversification is slightly better.

Data center REITs require attention to interest rate risk. Pacer's SRVR and Global X's DTCR are in this space, with DTCR's fee at 0.50%. These funds may be pressured in rising interest rate environments, but fundamentally, AI computing power demand continues to grow.

Another interesting direction is nuclear energy. VanEck's NLR covers nuclear power operators and uranium mining companies, with a fee of 0.56%; Sprott's URNM mainly invests in uranium mining firms, with a fee of 0.75%. This is a long-term bet on AI power demand.

To build a complete AI ETF portfolio, a simple approach is to pick a core AI fund (either Global X or iShares) and add a few satellite funds: one for chips, one for data centers, and one for cybersecurity (CIBR or HACK are both good, with fees around 0.60%). For a more aggressive stance, you can add some quantum computing ETF QTUM (0.40% fee), which is still speculative but has seen some progress in error correction algorithms recently.

Overall, AI-related ETFs' fees range from 0.34% to 0.75%. Most are passive index funds, with a few active management funds being a bit more expensive. The choice mainly depends on whether you want to focus on the AI platform layer or cover the entire industry chain. My own approach is to first choose a low-fee core AI fund as a ballast, then add specific sector funds based on my view of infrastructure.