AI stocks index builder: build your own AI index fund, set the weights, and backtest it against the AI ETFs.
Pick the chip, cloud, software and power companies you think will actually earn the AI money, decide how much each one gets, and see how the basket would have done next to BOTZ, SMH or the Nasdaq-100. Then buy it at your own broker.
The backtest did not run just now. Tap Backtest it to try again.
Educational only · Never places a trade
In short
There is no single "AI stocks index". Every AI ETF follows its own definition, so two funds with AI in the name can hold almost nothing in common. BOTZ, one of the largest, had Intuitive Surgical, Fanuc, Keyence and ABB as its top four holdings in September 2026, while SMH had 19% in NVIDIA alone, and the pure AI funds charge 0.45% to 0.75% a year. Building your own AI index means choosing the companies and weights yourself, usually 15 to 40 stocks across chips, cloud, software and power, backtesting the basket against those ETFs, and then buying the stocks directly at a broker, where there is no expense ratio.
From each issuer's own fund page
What is the best AI index fund, and what do the AI ETFs actually hold?
Twelve funds people buy for AI exposure, with the fee, the holdings count and the five largest positions as each issuer published them. The first nine carry AI in the name; the last three are the chip and tech funds they are usually compared with.
| Fund | Net expense ratio | Holdings | What it tracks | Largest positions |
|---|---|---|---|---|
| Global X Robotics & Artificial Intelligence (BOTZ) | 0.68% | 81 | Indxx Global Robotics & AI Thematic Index | Intuitive Surgical 8.5%, Fanuc 8.2%, Keyence 8.1%, ABB 7.9%, NVIDIA 7.8% (Sep 25) |
| Global X Artificial Intelligence & Technology (AIQ) | 0.68% | 88 | Indxx Artificial Intelligence & Big Data Index | Palantir 4.1%, Microsoft 3.6%, SpaceX 3.4%, Meta 3.3%, Intel 3.3% (Sep 25) |
| First Trust Nasdaq AI and Robotics (ROBT) | 0.65% | 111 | Nasdaq CTA Artificial Intelligence & Robotics Index | Tempus AI 2.1%, Meta 2.0%, Illumina 2.0%, Cloudflare 1.8%, Recursion 1.7% (Sep 25) |
| iShares Future AI & Tech (ARTY, formerly IRBO) | 0.47% | 49 | Morningstar Global AI Select Index | TSMC 4.7%, NVIDIA 4.7%, Micron 4.5%, AMD 4.4%, Super Micro 4.1% (Aug 26) |
| WisdomTree Artificial Intelligence and Innovation (WTAI) | 0.45% | 57 | WisdomTree AI & Innovation Index | Micron 5.2%, Samsung 4.4%, Sandisk 4.0%, Kioxia 4.0%, NVIDIA 3.8% (Jun 30) |
| Invesco AI and Next Gen Software (IGPT) | 0.56% | Not published in a readable form | STOXX World AC NexGen Software Development Index | See the Invesco fund page |
| Dan IVES Wedbush AI Revolution (IVES) | 0.75% | 30 | Solactive Dan IVES Wedbush AI Revolution Index | See the Wedbush fund page |
| Roundhill Generative AI & Technology (CHAT) | 0.75% | Actively managed | No index, active | NVIDIA, SK Hynix, Alphabet, Knowledge Atlas, AMD (Jun 30, weights not shown) |
| ARK Autonomous Technology & Robotics (ARKQ) | 0.75% | 37 stocks | No index, active | Tesla 10.9%, SpaceX 8.0%, Teradyne 6.5%, Kratos 5.4%, AMD 5.0% (Sep 25) |
| VanEck Semiconductor (SMH) | 0.35% | 25 | MVIS US Listed Semiconductor 25 Index | NVIDIA 19.3%, TSMC 9.3%, AMD 5.8%, Broadcom 5.3%, Intel 5.2% (Sep 24) |
| iShares Semiconductor (SOXX) | 0.34% | 30 | NYSE Semiconductor Index | Micron 8.5%, AMD 8.1%, NVIDIA 6.8%, Intel 6.3%, Broadcom 6.1% (Jun 30) |
| Vanguard Information Technology (VGT) | 0.09% | 321 | MSCI US IMI Information Technology 25/50 | NVIDIA 16.2%, Apple 14.4%, Microsoft 8.3%, Micron 5.1%, Broadcom 3.9% (Jun 30) |
Read the last column before the fee. The funds overlap far less than their names suggest. BOTZ is a robotics fund first: surgical robots, factory automation and machine vision fill its top four, and NVIDIA is fifth. AIQ spreads across 88 names with nothing above about 4%, and includes Intel and SpaceX. ROBT is close to equal weighted, so its largest holding is a 2% position in Tempus AI. ARTY and WTAI lean into memory and semiconductors. SMH puts almost a fifth of your money into one company.
So "which AI ETF is best" is really a question about which definition of AI matches your view. If you think the money lands with chipmakers, SMH or SOXX does that for about 0.35%. If you think it lands in software, the robotics-heavy funds are the wrong tool. And if none of the twelve holds the mix you have in mind, that is the case for building the index yourself. The same list, ranked purely by fee with the dollar cost at each balance, is in our cheapest AI ETF comparison.
Decide what counts as AI before you pick a single stock
Which stocks belong in an AI index?
Nearly every AI basket, fund or homemade, is some mix of four layers. The mix is the decision that matters, because the layers have behaved very differently from one another.
1
Compute and chips
The GPUs, accelerators and custom silicon that train and run models, plus the foundries and equipment makers behind them.
NVDA, AMD, AVGO, TSM, ASML, MU
2
Cloud and hyperscalers
The companies renting that compute out and spending the most on it: the big three clouds and the platforms buying GPUs at scale.
MSFT, AMZN, GOOGL, META, ORCL
3
AI software and applications
Companies selling AI to businesses as a product: data platforms, security, enterprise software with AI priced in.
PLTR, NOW, CRM, SNOW, CRWD, ADBE
4
Power and data center build-out
The physical side: data center REITs, electrical and cooling equipment, networking and the utilities feeding the load.
EQIX, DLR, VRT, ETN, ANET, CEG
The tickers are examples of what each layer contains, not a recommendation. The useful exercise is to write your own thesis in one sentence and let it pick the layers. "Model training keeps scaling" is a compute-heavy basket. "Businesses will pay for AI features" leans on software. "Data centers are the bottleneck" points at power, cooling and networking, which is a basket that shares very few names with a chip ETF.
Then check what you already own. If most of your money sits in an S&P 500 or Nasdaq-100 fund, you already hold Microsoft, NVIDIA, Alphabet, Amazon and Meta at large weights. Adding a cap-weighted AI basket on top mostly doubles those five. An AI index that is built to complement your core, with the mega caps capped or left out, is often the more honest version of the bet. Our guide to how much of your portfolio should be in one stock covers the concentration side of that.
The same 20 stocks can be four different indexes
How should you weight an AI stocks index?
| Weighting | What it does to an AI basket | Use it when |
|---|---|---|
| Market cap | Follows company size. In AI that means three or four mega caps carry most of the basket, so it behaves a lot like the Nasdaq-100 you may already own. | You want the index to look like the market sees AI today. |
| Equal weight | Every name gets the same slice. A 20-stock basket puts 5% in each, so a mid-size software or power name moves the result as much as NVIDIA does. | You want the theme, not the mega caps. |
| Capped | Market cap with a ceiling, say 8% per stock. The big names still lead, but no single chipmaker can become a third of the index. | You want size to matter, within limits. |
| Layer weights | You set the split by layer first (for example 40% compute, 25% cloud, 20% software, 15% power) and weight inside each layer. | You have a view on where the AI money lands next. |
Weighting is where a homemade AI index most often surprises its owner. A cap-weighted basket of 25 AI names can have more than half its value in three stocks, which means the other 22 barely register. Equal weighting fixes that but needs regular resets, because the winners grow past their slice and the basket drifts back toward cap weight on its own. A quarterly reset is a common starting point; the custom index rebalancing tool lets you backtest monthly, quarterly and annual schedules side by side before you commit to one.
In the builder you can set every weight by hand or use equal weight on any plan; market cap and inverse volatility weighting are on the Investor plan. Inverse volatility is worth a look for AI in particular: it gives calmer names such as the cloud platforms more room and the most volatile small software and chip names less, which changes the drawdown profile more than it changes the theme. The market cap vs equal weight comparison goes deeper on the trade-off.
From thesis to a basket you can buy
How to build your own AI index fund in four steps
1
Pick the names
Search the catalog of US stocks and ETFs and add the companies in each layer you believe in. 15 to 40 is the usual range.
2
Set the weights
Equal, market cap, inverse volatility or your own split, and a rebalancing schedule so the weights stay where you put them.
3
Backtest it
Run it against the S&P 500 and the Nasdaq-100. Build a one-line index of BOTZ, SMH or any AI ETF to compare against the fund too.
4
Track it, then buy it
Watch the index daily against its benchmark, export the weights, and place the basket at your own broker when you are ready.
The backtest is the step most people skip and the one that saves the most money. Run your basket through 2022, when AI and chip stocks fell hard, and look at the maximum drawdown next to the Nasdaq-100. If your version fell 20 points further than the index you already own, you are taking a lot of extra risk for the theme, and it is better to learn that on a chart than in your account. Backtested results are hypothetical and past returns do not predict future ones, so use the test to compare designs, not to forecast. Our note on backtesting mistakes to avoid covers the traps, including hindsight in the stock list itself.
Indexes designs and tests the basket. A broker holds it.
How do you invest in your own AI index once it is built?
Buy the basket yourself
Any broker with fractional shares lets you buy 20 stocks at exact weights with a few hundred dollars. Fidelity Basket Portfolios is built for it: up to 50 stocks or ETFs traded in one order, from $1 per position, for a flat brokerage fee of $4.99 a month per Fidelity's page. No expense ratio, and you own every share.
A direct index on an AI benchmark
Public's Direct Index Account, 0.19% a year from $1,000, lists thematic indexes such as the Solactive AI Technology Index among its 100+ benchmarks, and harvests tax losses automatically. You get their AI definition, not yours, with your own exclusions on top. See our Public direct indexing review.
Buy the ETF after all
If the backtest shows an existing fund is close enough to what you wanted, buy the fund. One ticker, no rebalancing chores. The point of building first is that you choose the ETF knowing what it holds, instead of because it has AI in its name.
Owning the stocks directly has one advantage a fund cannot match in a taxable account: each position has its own cost basis. When one AI name falls 30% while the basket as a whole is up, you can sell that one position, realize the loss against gains elsewhere, and buy a similar company in its place. Inside an ETF that loss is invisible to you. The wash sale calculator shows what happens if you buy the same stock back within 30 days.
What Indexes does and does not do: it is index construction, backtesting and tracking software. It does not hold money, place trades or give investment advice. You build and test the AI index here, and hold it wherever you already invest.
Four ways to get AI exposure, compared honestly
Build your own AI index vs an AI ETF vs a basket broker vs a direct index
| AI ETF | Fidelity Basket Portfolios | Public direct index | Indexes, then your broker | |
|---|---|---|---|---|
| Who picks the stocks | The fund's index or manager | You, or a Fidelity model | The index provider, you exclude | You |
| Yearly cost | 0.34% to 0.75% of the balance | $4.99 a month flat | 0.19% of the balance | From $12 a month for the software, plus your broker's cost |
| Backtest before you buy | Fund history only | No | No | Yes, against the S&P 500 and Nasdaq-100 |
| Weighting control | None | Full | Index, cap or equal | Custom, equal, cap, inverse volatility |
| Per-stock tax losses | No | Yes, done by you | Yes, automated | Yes, at your broker |
| Where it falls short | You take the fund's AI definition | No history to test a design | Their theme, not yours | You place and rebalance the trades |
The combination most people land on is the last two columns together: design and backtest here, then hold the basket at a broker that makes multi-stock orders cheap. An ETF wins when you want zero upkeep and a fund already matches your view. For themes beyond AI, the same approach is laid out on our thematic investing page, and the build your own ETF guide covers the broker side in more detail.
AI index fund questions
What people ask before they buy AI exposure
What is the best AI index fund?
There is no single best one, because each AI ETF defines AI differently. For low cost with a pure AI mandate, WisdomTree WTAI charges 0.45% and iShares ARTY 0.47%. For chip exposure, SMH and SOXX charge about 0.35%. Before picking, check the top holdings: BOTZ, one of the biggest, is mostly industrial robotics, not AI software.
Is there an AI index fund?
Yes, as ETFs. BOTZ, AIQ, ROBT, ARTY, WTAI, IGPT and IVES each track an AI or robotics index, and CHAT and ARKQ are actively managed. Vanguard does not offer one. You can also build your own AI index from individual stocks and buy it at a broker with fractional shares, with no expense ratio.
Does Vanguard have an AI ETF?
No. Vanguard does not offer a fund dedicated to artificial intelligence. The closest is the Vanguard Information Technology ETF (VGT) at 0.09%, which holds 321 tech stocks with NVIDIA, Apple and Microsoft at about 39% combined as of June 30, 2026. That is broad tech exposure with AI leaders inside it, not an AI index.
How many stocks should be in an AI index?
Most homemade AI indexes land between 15 and 40 stocks. Fewer than 15 makes the result depend on one or two names. More than 40 starts pulling in companies with little real AI revenue. The AI ETFs range from 25 holdings (SMH) to 111 (ROBT), so there is no standard, only a trade-off between focus and spread.
Is it cheaper to build your own AI index than buy an AI ETF?
Usually, above a few thousand dollars. A 0.68% ETF costs $68 a year on $10,000 and $340 on $50,000. Buying the stocks directly has no expense ratio; a basket service such as Fidelity Basket Portfolios is a flat $4.99 a month. The trade is that you place the trades and rebalance yourself.
What do AI ETFs actually hold?
Less AI than the name suggests, and very different things from each other. As of September 25, 2026, BOTZ's top four were Intuitive Surgical, Fanuc, Keyence and ABB. AIQ's were Palantir, Microsoft, SpaceX and Meta. SMH had 19% in NVIDIA alone. Read the holdings file on the issuer's site before you buy.
Can I backtest an AI ETF against my own basket?
Yes. In Indexes you build your basket, then build a second index holding just the ETF, and compare both against the S&P 500 or the Nasdaq-100 over the same period. That shows return, drawdown and how closely the two moved, which tells you whether the fund is worth its fee for your view.
Sources
Expense ratios, holdings counts, underlying indexes and top holdings from each issuer, read September 27, 2026: globalxetfs.com (BOTZ, AIQ, holdings as of September 25), ftportfolios.com (ROBT), the iShares product pages and fact sheets (ARTY, SOXX), the WisdomTree WTAI fact sheet (June 30), Invesco fund documents (IGPT), roundhillinvestments.com (CHAT), ARK's daily ARKQ holdings file, wedbushfunds.com (IVES), vaneck.com (SMH, September 24) and Vanguard's VGT fact sheet (June 30). Fidelity Basket Portfolios pricing and limits from fidelity.com/direct-indexing/customized-investing. Public's index menu and fee from our Public direct indexing review, sourced from Public's fee schedule and help center. Weighting schemes, benchmarks and plan limits describe the Indexes product as it works today. Expense ratios and holdings change, so confirm them on the fund page before you buy. Indexes is not affiliated with any fund issuer or broker named here. We make index construction and backtesting software: we do not manage money, place trades or give investment advice, and backtested results are hypothetical.
Build the AI index you actually believe in
Choose the layers, the names and the weights, backtest the basket against the Nasdaq-100, and track it every day. Then buy it wherever you invest.
Keep reading