AI stocks and AI ETFs in retirement planning.

The AI rally in numbers.

Since the release of ChatGPT in late 2022, AI-related equities have posted double- and triple-digit gains. Nvidia briefly became the most valuable company in the world. Microsoft expanded Azure. Alphabet and Meta deployed record capex into their own models. The Morningstar Global Next Generation AI Index rose about 41 percent in the twelve months to January 2026.

Alongside the rise came corrections. In January 2025 news of the Chinese model DeepSeek wiped roughly 600 billion dollars off Nvidia in a single day. In spring 2025 tariff announcements dragged the sector down again. Investors who sat through both troughs are clearly ahead now. Those who sold in the panic are not.

Thematic investing, what it means.

A broad equity ETF such as MSCI World or FTSE All-World replicates the global market by capitalisation. That means it already holds the biggest AI beneficiaries: Nvidia, Microsoft, Alphabet, Apple, Meta and Amazon together make up roughly a quarter of the MSCI World. Anyone running a savings plan on such an ETF has meaningful AI exposure already, often without knowing it.

A thematic AI ETF does something different. It deliberately narrows the selection to companies that are either AI pure plays or generate a large share of their business from AI. Three products dominate the German market: Xtrackers Artificial Intelligence and Big Data (ticker XAIX), WisdomTree Artificial Intelligence (WTAI), and Amundi MSCI Robotics and AI ESG Screened (GOAI).

The difference is concentration. A broad world ETF spreads risk across 1,500+ companies in 23 countries. An AI ETF typically holds 50 to 100 names, heavily tilted to tech and to the US. In good AI years it outperforms the market. In bad AI years it lags. Both with bigger swings.

Three AI ETFs compared.

Xtrackers AI & Big Data (XAIX) tracks the Nasdaq Global AI and Big Data Index, holds around 90 positions, and in 2026 was by far the largest AI ETF on the German market at roughly 5.5 billion euros. Ongoing costs: 0.35 percent per year, accumulating, physically replicated.

WisdomTree AI (WTAI) follows the Nasdaq CTA Artificial Intelligence Index and skews toward smaller, more specialised AI companies. It is therefore more volatile than XAIX, with wider return dispersion and a much smaller fund size. Costs: 0.40 percent.

Amundi MSCI Robotics & AI (GOAI) broadens the AI focus with robotics and automation, plus ESG screening. That makes it less volatile than pure-AI indices and more diversified. Costs similar at around 0.4 percent. Often the more conservative choice for a first position in the theme.

All three carry a PRIIPs risk indicator of 5 out of 7. That is high, but normal for thematic equity funds. Investors should plan for a horizon of at least ten years and be able to absorb single-year drawdowns of up to minus thirty percent.

How much AI fits in a pension portfolio.

The common rule of thumb for thematic satellite ETFs in a pension portfolio: 5 to 10 percent of the equity allocation, not of total wealth. Someone with 70 percent of their assets in equities and 10 percent of that in AI is therefore running a 7 percent overall theme allocation.

Why not more. Because world ETFs already contain roughly a quarter AI-related names. A 10 percent satellite doubles the weight of the largest AI stocks in the portfolio without destroying diversification. More than 10 percent tips the portfolio toward a tech-concentrated bet, and that is no longer retirement planning but a wager.

Why at all. Because thematic investments in early growth phases can return above-market. Early Dotcom investors typically entered too late, not too early. A small, long-term position in a theme you structurally believe in is a legitimate addition.

Single stocks instead of ETFs, is it worth it.

Buying Nvidia in 2020 and holding is now a multi-bagger. That is the story everyone tells. The stories told less often: C3.ai, mostly underwater since its 2020 IPO. Baidu, five years with no price gain. Intel, once an AI hope, now a turnaround case.

Picking individual AI winners is trivial in hindsight and hard in prospect. In retirement planning the rule is: unless stock analysis is your profession, stick to what an ETF delivers automatically. Single stocks as satellites are fine, as core positions they are not.

The main risks.

Concentration risk: AI ETFs are heavily weighted in a small number of US tech names. A broad benchmark such as MSCI World is already US-heavy despite its global name; AI ETFs more so.

Valuation risk: some AI stocks trade at multiples not seen since the turn of the millennium. Goldman Sachs and Morgan Stanley flagged correction risks of 10 to 20 percent several times in 2026. That is not a crash call, but not an invitation to maximum exposure either.

Definition risk: what ends up in an AI ETF is decided by the index provider. An "AI company" can range from Nvidia (pure chip producer) to a bank that uses AI for fraud detection. The index prospectus is worth reading before a savings plan goes in.

Regulatory risk: the EU AI Act has been in force since 2024 and directly affects many of the companies listed here. More regulation in the US and China is likely. Regulatory risk in 2026 is materially higher than it was in 2022.

AI stocks in the drawdown phase.

Anyone approaching retirement should reduce the satellite share, not increase it. Five years before retirement a high equity allocation becomes problematic in general, and thematic satellites in particular. A market drop in the wrong moment hits retirement planning asymmetrically. The rule is simple: the closer to retirement, the more defensive the portfolio, the smaller or zero the thematic positions.

During drawdown itself, a small AI satellite is still defensible, provided withdrawals do not come from it directly. Withdraw first from defensive buckets. Rebalance or harvest the thematic position only in good years. This is the "bucket strategy" and is well documented in German retirement literature.