Three companies — TSMC, Samsung Electronics, and SK Hynix — have absorbed approximately 28% of the entire MSCI Emerging Markets index between them, according to reports from The Economic Times. This concentration is more than 2.6 times India's total 10.87% country weight in the same benchmark. In Korea, the top 10 stocks now make up roughly 65% of the KOSPI, with electronics alone at 60.2% of the index. In Taiwan, the top 10 stocks exceed 65% of the TAIEX, with semiconductors at roughly 56%. TSMC alone accounts for 14.2% of the MSCI EM benchmark, Samsung Electronics for 7.8%, and SK Hynix for 6.6%. The question now consuming allocators isn't the gap itself — it's what comes next: whether a single layer of the AI supply chain, advanced fabrication and high-bandwidth memory, has made two of Asia's largest markets dangerously exposed to one input, NVIDIA's order book.
"India doesn't carry that single-point-of-failure risk," said Parvati Rai, Fund Manager at Equentis PMS to ET Markets. "Taiwan and Korea have built phenomenal businesses, but their indices now live and die by a handful of semiconductor names — in both markets, one or two stocks alone account for approximate 40-50% of the index weight, so one export curb or demand wobble and the whole market feels it." Rishabh Nahar, Co-founder at Qode Advisors, framed the risk in sharper terms, stating that "markets like Taiwan and Korea are hostage to a single point in the AI value chain — chip fabrication — which makes their indices extraordinarily sensitive to one variable: NVIDIA's order book and US export policy." India's market cap is spread across banks, consumption, industrials, and increasingly power and infrastructure, which could translate into more resilient and superior risk-adjusted returns over the full cycle, even without a domestic semiconductor champion.
According to Ajay Modi, Director at Piper Serica, India is building capabilities "across semiconductor design, data centres, power infrastructure, cloud, networking, enterprise software, and AI applications," backed by policy support including the ₹76,000 crore Semicon India Programme. The country's data centre capacity stood at approximately 1.5 GW in 2025 and is expected to reach 1.7–2.0 GW by the end of 2026, with ambitions of scaling to 5 GW by 2030. Major hyperscalers are investing significantly, with Microsoft announcing $17.5 billion in investment over four years, Google expected to invest roughly $15 billion through 2030, and AWS committing more than $8 billion. Reliance and Meta are collaborating on one of India's largest AI-ready data-centre projects in Jamnagar.
"What's changed is where the incremental money is going within India — away from expensive consumer discretionary and IT names, and toward hard asset plays that are structurally linked to the AI buildout," said Nahar to ET Markets. "Every data centre conversation eventually becomes a power conversation," he noted, calling power, cooling and grid infrastructure plays "picks-and-shovels" bets with strong domestic policy support. Modi pointed to the scale of capital flowing into India's data-centre buildout, describing what was once a niche real-estate story as having "turned into a multi-decade capex cycle, backed by hyperscaler commitments and strong policy tailwinds." Rather than a tactical rotation into 'hard assets,' what we are witnessing is a structural reallocation toward the physical foundations that will enable AI adoption at scale over the next decade. The market is also betting on the so-called 'reflation trade' — inflation and growth picking up, but not to the point where inflation chokes off the economy.
According to Lighthouse Canton, MSCI India currently trades at a substantial discount to the all-time-high valuation premium it commanded two years ago, while foreign institutional ownership is hovering at a six-year low. This combination has historically acted as "a springboard preceding India's strong period of regional and global outperformance," as reported by The Economic Times. Rai sees allocators moving away from crowded sectors toward capacity-starved sectors like power, transmission and data centres as a way to participate in the AI buildout without paying chip-stock valuations. The entry point matters significantly, with the scale of imbalance showing three chip and memory makers worth more on a single emerging-markets benchmark than an entire national equity market of 1.4 billion people. Global funds that have been underweight India on valuation grounds are finding it difficult to justify that stance when the underlying earnings growth story remains intact.
