In our recent WisdomTree Survey, we asked 4000 savers and investors¹ in Europe where they expect to find growth over the next five years, and the answer isn't just Artificial Intelligence (AI) anymore. The AI investment theme has grown so large that it is now splitting into more specialised sub-themes. For example, AI software (33%) and AI infrastructure (29%) were the two investment themes investors believed will matter most over that horizon.

Yet most of those same investors feel they have already "done AI." They own a global equity index like the MSCI World, S&P 500, or perhaps the Nasdaq-100, and assume the exposure is baked in. After all, some of the largest companies on the planet make AI chips, AI software and AI infrastructure. Surely that's enough?

What a broad index actually buys you

A global equity index is built to represent the whole market, weighted by market capitalisation, i.e. company size. That means its AI exposure is really an exposure to AI companies that have already become giants. Today, that's a short list of very well-known companies: Nvidia, Apple, Microsoft, Meta and so on. But those indices are very concentrated, and the seven largest US technology names, often called the ‘Magnificent Seven’, now account for a quarter of it².

That's a lot of concentration in a handful of AI names, but concentration isn't the same as coverage. A market-cap-weighted index is, by design, a snapshot of yesterday's winners, not a map of where a theme is heading next. It holds Nvidia because Nvidia has already won. It won't hold significant weight in the next key company until that company has also grown large enough to force its way in, usually long after most of the growth has already happened.

The index's exact definition also matters. Some stocks that performed very strongly in the AI theme in 2026, like SK Hynix, Samsung or TSMC, are not part of the MSCI World, as Korea or Taiwan are still considered emerging countries by MSCI. To access those companies, an investor would need to use an MSCI Emerging Markets or MSCI All Country World index, for example.

AI is a diversified theme with many emerging players

At WisdomTree, we map the AI theme across three distinct layers:

• The companies building AI's physical foundations, i.e. the AI infrastructure.
• The businesses designing the software and systems that make it usable.
• The machines bringing AI into the physical world, such as drones, robotaxis or humanoids, also known as Physical AI.

Our own AI strategy (which tracks the Nasdaq CTA Artificial Intelligence Index) is built around all three. Focusing on companies developing AI's building blocks, such as chips and data infrastructure, and companies designing AI products. Crucially, though, inclusion and weighting in this strategy depend not solely on company size but on their sensitivity to the growth of the AI ecosystem.

Owning MSCI World can feel like owning AI. Yet a dedicated AI strategy overlaps with it by just 22%³ (calculated as the share of one index's weight made up of holdings it also shares with the other), far less than you’d expect.

Looking at the top 10 of our AI strategy, the differences between an AI-focused strategy and the MSCI World are stark:

• Tempus AI (applies AI to genetic and clinical data to match cancer patients with the right treatment faster) and Nice (uses AI agents to handle customer service calls and chats once run entirely by people) are our 4th and 6th biggest holdings and are not even in the MSCI World Index.
• Marvell Technology (designs custom AI processors and networking chips for cloud giants' data centres, an alternative to off-the-shelf GPUs) is our 5th biggest holding and only accounts for 0.2% in the MSCI World Index.
• Synopsys (makes chip-design software, now using AI to automate much of that design work) is our 10th biggest holding and only accounts for 0.1% in the MSCI World Index.

Figure 1: Top 10 stocks in the WisdomTree AI strategy versus their weights in the MSCI World

Source: MSCI, Nasdaq, WisdomTree. As of 30/09/2026.

The difference also shows in performance, with the strategy gaining 48.08% over the last 12 months versus 15.24% for MSCI World⁴. It is worth noting that even with a pure tech benchmark like the Nasdaq 100, the overlap remains very contained at 36.6%² and the performance differential is also very wide, as the Nasdaq only returned 23.8% over the last 12 months³.

Figure 2: AI strategy performance versus global equity indices

Source: WisdomTree, Bloomberg. As of 11 July 2026. Historical performance is not an indication of future performance and any investments may go down in value.

The infrastructure layer

Spending by the large cloud computing groups, or ‘hyperscalers’, on AI data centres has climbed from around $20 billion a quarter in 2020 to close to $150 billion in early 2026⁵. Over roughly the same period, prices for the memory chips used in AI servers have surged by as much as eight-fold, as demand outstripped supply. While exposure to that layer in the MSCI World Index would mostly consist of semiconductors with Nvidia and the hyperscalers themselves, there is a complete infrastructure layer: compute, memory, networking, power and cooling, that keeps AI running. It is built around names like Vertiv, Delta Electronics and SK Hynix, companies that have very small weights, if any, in global equity indices. The WisdomTree SemiAnalysis Artificial General Intelligence Infrastructure UCITS Index invests across that full value chain and exhibits only 18%² overlap with the MSCI World. TSMC, SK Hynix, LAM Research or Amphenol are companies in the top 10 in that index that have a small to zero weight in the MSCI World.

The part that isn't on your screen

Then there is a layer most equity indices barely touch: AI leaving the screen entirely. The installed base of drones is forecast to grow more than sevenfold by 2030. Robotaxi services are scaling across dozens of US cities this year. Humanoid robots are moving from prototype to mass delivery; UBTech alone has taken orders worth more than 800 million yuan from the likes of Foxconn and BYD⁶. None of this shows up meaningfully in MSCI World, the S&P 500, or even in a broad AI strategy. The companies involved, from Rainbow Robotics to Red Cat Holdings, are mostly too small, too new, or too specialised to register. The WisdomTree Physical AI UCITS Index shows a minuscule 7.7% overlap² with the MSCI World and only two stocks in the top 10 even belong to the MSCI World.

Why this matters now

AI has not stood still, even over the last twelve months, from the surge in memory prices, to the arrival of ‘agentic’ AI workflows, to humanoid robots reaching commercial scale. A strategy that simply holds the winners of two years ago risks looking as dated as the winners of a previous tech cycle. AI strategies are built to evolve with the theme and aim to capture opportunities arising from AI-related growth.

Owning a broad index isn't wrong. For most portfolios, it's a sound core holding. But if the goal is genuine exposure to AI growth, its infrastructure build-out, and its move into the physical world, a global index gives you only a narrow slice of a much bigger opportunity. AI-focused strategies can be used as a satellite to that core holding to tilt the portfolio toward AI-powered growth stocks.

Investors should consider that thematic strategies may carry greater concentration risk and volatility than broad equity indices.

¹ 4,000 adults across Europe with at least £5,000/€5,000 in savings, including 1,000 UK investors.
² Source: MSCI, Bloomberg Finance L.P. As of 5 October 2026.
³ Source: MSCI, Nasdaq, WisdomTree. As of 30 September 2026.
⁴ Source: MSCI, WisdomTree. From 30/09/25 to 30/09/26.
⁵ Source: WisdomTree, Bloomberg. As of 11 July 2026. Historical performance is not an indication of future performance and any investments may go down in value.
⁶ Source: UBTECH, as of 17 November 2025.