Ruchir Sharma on Why India Seems to Be on the Wrong Side of the AI Trade

ELI5/TLDR

Ruchir Sharma thinks the AI rally is a real bubble — overvalued, over-owned, over-invested — but says bubbles don’t pop on their own; they need higher interest rates to kill them. Until the US 10-year breaks 5%, the music plays. India is being punished because it doesn’t make chips and its software jobs look exposed to AI; foreigners are pulling money out. His hedge isn’t gold (already too rich) — it’s the boring, high-quality stocks no one wants right now, plus cheap markets like India and the Philippines as optionality if the bubble breaks.

The Full Story

The AI boom is swamping everything

Sharma’s framing: this is the second year in a row where a story that should have hurt markets — last year tariffs, this year oil — has been overwhelmed by AI capex. Average US tariff rates went from 2.5% pre-Trump escalation to ~10% even after the walk-back. Markets shrugged.

“The AI boom is just so much bigger that it’s swamping all other effects.”

The mechanism is a feedback loop: companies keep raising capex, everyone is convinced AI is the future, that conviction drives flows, the flows justify more capex.

This is a bubble — by his own four-fault test

Sharma scores bubbles on four “faults”: overvaluation, overinvestment, overleverage, overownership. AI scores high on three of four. The exception is leverage — hyperscaler capex is still being funded out of cash flow, not debt (yet).

On overownership, the headline stat: 60% of US households now have equity exposure, the highest in the world. America is the only country where households hold more wealth in stocks than in property. In China and most other places, property wealth is roughly 5x equity wealth.

What kills bubbles is rates, not gravity

This is the load-bearing claim of the interview. Three hundred years of bubble history says the same thing: bubbles don’t burst under their own weight. They need a liquidity event — usually higher interest rates that make people start asking “what’s the return on all this capex?”

“Until you have high interest rates or some sort of a liquidity event which tightens money in the marketplace, bubbles don’t just burst under their own weight.”

He’s watching the US 10-year obsessively. It’s range-bound at 4–4.5%. The number to watch is 5%. The 30-year already there is an “orange signal,” not alarm bells.

India is on the wrong side of the trade

Two reasons foreign money is leaving India:

No compute, no chips. The current AI phase is infrastructure build-out and India never invested in it. R&D as a share of GDP: India 0.6%, US ~3%, China ~3%, Korea/Taiwan 4–5%.
Job risk in software/GCCs. 10–15 million Indians work in sectors AI is perceived to disrupt, many of them well-paying. Jobs were already a structural problem.

Result: while Korea is up ~70% and Taiwan ~40% over the past year, India is down ~10% in dollar terms. EM index concentration is now even worse than the S&P 500’s.

His portfolio answer

Decide how much AI you want, then layer on cheap hedges that won’t work today but will if the regime flips:

India and the Philippines — Philippines is at its lowest valuation since the East Asian financial crisis on some metrics.
Quality stocks globally — companies with >15% ROE and decent earnings growth. The “quality” factor has been crushed because everything that isn’t AI got abandoned.
Not gold — gold ran so hard it’s no longer a risk-free hedge; it’s now a position with embedded risk.

On peak US exceptionalism

He called the top in December 2024 — and clarifies he meant relative performance, not absolute. Since then, international markets have outperformed the US, and the dollar has weakened on the margin despite massive inflows. The rupee and peso are exceptions (“crumbled,” the anchor offers; he prefers “weakened”).

Bubbles end with a parabola

A historical detail he flags: from October 1999 to the Nasdaq peak, the index doubled. The last phase is always the most violent. Korean chip stocks moving 10% a day right now — Samsung, SK Hynix at large-cap scale — is parabolic price action. He doesn’t want to be in it, but acknowledges the performance pressure that keeps managers playing.

On geopolitics and the midterms

Markets are bad at pricing geopolitical risk because 99% of the time it doesn’t materialize, and the 1% that does (think WWI) blindsides them. The 1990s felt like a golden era only in retrospect.

On midterms: House likely flips, Senate is 50/50. Trump’s approval is low, but every president’s approval has trended down for 50 years in a more polarized world. The Democrats’ generic ballot numbers are below Trump’s — which is why he keeps getting away with things.

Key Takeaways

Bubbles need a liquidity trigger to pop; gravity alone doesn’t do it. The trigger to watch: US 10-year above 5%.
Sharma’s four-fault bubble test: overvaluation, overinvestment, overleverage, overownership. AI scores high on three. Leverage is the missing fault — capex is still cash-funded.
60% of US households own equities, the highest share globally. The US is the only country where household stock wealth exceeds property wealth.
The average US tariff rate went from 2.5% to ~10% and markets didn’t blink — AI capex absorbed the shock.
India’s R&D spend is 0.6% of GDP vs ~3% in the US/China and 4–5% in Korea/Taiwan. That structural gap is why India is on the wrong side of the AI infrastructure trade.
Korea +70%, Taiwan +40%, India −10% in dollars over the past year. The EM index is now more concentrated than the S&P 500.
Bubble endings are parabolic. From October 1999 to the Nasdaq peak, the index doubled — most of the gain came in the last few months.
“Peak US exceptionalism” was a relative call. Since Dec 2024, ex-US has outperformed and the dollar has weakened despite huge inflows.
Gold has stopped being a clean hedge — the rally embedded risk into the price.
His preferred hedges: high-quality global stocks (>15% ROE), and cheap optionality in India and the Philippines.
Generic-ballot polling matters more than presidential approval. Democrats currently poll below Trump on it.

Claude’s Take

Sharma is doing what he does well: laying out a structured, historically-grounded view without overclaiming a timing edge. The four-fault framework is genuinely useful, and the “bubbles need rates to die” rule is the kind of base-rate thinking most commentators skip. The 60% US-household-equity stat and the property-vs-stock-wealth comparison are the kinds of facts you remember.

The India argument is the sharpest part. He resists the lazy bullish-on-India template and frames it as a liability in this specific phase of the cycle, with optionality if the cycle turns. That’s harder to say on Indian TV. The pivot — that India might benefit later in the adoption phase — is plausible but is the part of the argument doing the most hand-waving.

Where it gets thinner: the “quality stocks as the best hedge” suggestion is a recycled line that’s been wrong for two years and could keep being wrong as long as concentration intensifies. And the midterms detour is fine punditry but not really insight.

Score: 8. Compact, fact-dense, and genuinely useful as a frame for the next 12 months. Loses a point for the standard problem of all bubble commentary — being directionally right and timing-wise useless.