Who wins and loses from AI? HSBC maps global impact

HSBC mapped AI exposure across nearly 2,000 business lines and found Taiwan the biggest beneficiary under a moderate disruption scenario at a 53% net AI-enabled revenue share, while India shows the highest net AI-disrupted share at 7%. Under a large disruption scenario, Europe turns negative, led by Austria at 30% below baseline, with Spain and Ireland at 20% below baseline and France at 11% below baseline. Semiconductors and technology hardware are the clearest AI beneficiaries, while commercial/professional services, media, banks and healthcare equipment face the deepest downside.

Analysis

The key market implication is not simply “AI winners vs losers,” but that dispersion is widening across regions with very different factor sensitivity. Taiwan and Korea look like the cleanest second-order beneficiaries because their equity indices have outsized exposure to the AI capex stack: leading-edge semis, advanced packaging, memory, and hardware supply chains that monetize on both unit growth and content uplift. That makes them higher beta proxies for AI spend than US mega-cap software, which may already have more valuation discounting embedded.

The more important negative is that AI disruption appears to hit labor-arbitrage and advice-heavy business models first, then creeps into regulated services later. That suggests the underperformance risk is likely lagged by 6-18 months: initial multiple compression in software/services, followed by margin pressure as pricing power erodes and client churn rises. If the large-disruption path gains credibility, banks and healthcare-adjacent software become vulnerable because AI can compress back-office headcount and advisory fee pools faster than investors expect, especially in markets with slower labor flexibility.

Contrarian take: the market may be overestimating the immediacy of disruption and underestimating the durability of AI beneficiaries outside the obvious US platform names. Many of the “loser” sectors have sticky regulation, embedded workflows, and long contract durations, which can defer revenue leakage for years. Conversely, low-AI-exposure regions and sectors may become stealth hedges because they can benefit from capital rotation away from crowded AI trades without needing direct AI monetization.

The cleanest trade is to own the AI supply chain where earnings revision momentum is still early, while shorting the most valuation-sensitive service exposures that have not priced in margin compression. This is a relative-value theme, not a broad index call: the spread trade should work even if AI enthusiasm persists, because capital spending is already being reallocated toward hardware and infrastructure rather than downstream services.

Key Decisions for Investors

• Long TSM / ASML / KLIC basket vs short a software-services basket (e.g., WDAY, ADP, WPP if accessible) over 6-12 months: best risk/reward is via a pair, not outright longs, because AI capex should stay sticky even if risk appetite fades.
• Overweight KOSPI / Taiwan equity exposure through FX-neutral vehicles for 3-6 months; these are the most direct ways to capture AI hardware beta with less narrative risk than US mega-cap AI names.
• Short European financials and advisory-heavy names on rallies over the next 1-2 quarters, especially where fee income is already slowing; downside should come from margin pressure before credit deterioration shows up.
• Use low-AI-exposure regions/sectors as portfolio hedges: initiate a small long in Latin America or selected Utilities/Materials baskets as a volatility offset against crowded AI longs.
• If buying AI upside, prefer call spreads on semiconductor supply chain names rather than outright equity: the convexity is better if AI spend persists, while limiting multiple compression risk if the market rotates away from the theme.