India's AI-refuge rally is an oil trade

One election result, two opposite verdicts. The Democratic sweep that drove Korea's Kospi down 5.3% — on fears of "government-mandated pauses in data center construction" — and confirmed the Nasdaq's 10% correction lifted India the same week ^[1][2]. The Sensex gained about 1,600 points and the Nifty 1.8% on November 5, a fourth straight session of foreign buying that put roughly Rs 14,863 crore — about $1.8 billion — into the market and left the Nifty the only gainer among Asia's five largest benchmarks this quarter ^[3]. The pause fear has a concrete face: Texas, which holds roughly 27% of planned US data-center capacity, just elected a governor who ran on ending data-center development ^[1]. That is the American leg, and it has been picked over all week. The explanation now being claimed for India's leg was ready before any votes were counted. Before Diwali, the billionaire banker Uday Kotak argued that the leverage behind the global AI buildout had produced serious overvaluation, and he made a specific prediction ^[4].

The kind of leverage which global companies are taking to fund their AI aspiration is also leading to significant overvaluation in many cases. — Uday Kotak

The flow data is starting to read like vindication. Fidelity's multi-asset manager is adding Indian equities ^[5], and Davidson Kempner is moving into Indian first-lien lending even as hedge fund managers gathered in Hong Kong warn that private credit is the AI trade's weakest point ^[6]. But look at what is actually moving the money. Brent crude fell nearly 10% to about $94 after the vote, on hopes of a faster resolution of the Iran war — relief for an oil-importing economy ^[3]. Indian bonds rose for a fourth straight session, the benchmark 6.94% 2036 yield down six basis points to 7.1258% as the US 10-year slipped to 4.84% ^[7]. That is the same event moving a second instrument through oil and rates, not through chips. The counter-note belongs right here. The same Fidelity manager is also adding Nasdaq names and memory chips, on the view that the selloff opened a valuation gap he considers disconnected from strong hyperscaler earnings ^[5]. A prediction can come true for the wrong reason, and so far this one has. The flow has a history, and it whipsaws. In July, a record Rs 40,031 crore came into Indian assets ^[8] as investors fled $1.6 trillion of US tech value and bought beaten-down Indian IT stocks, pushing the Nifty IT index up 16% in a rally SBI Securities tagged the "anti-AI trade" ^[9]. By September, foreign investors were sellers again — $3.2 billion out on oil-and-yield pressure ^[10]. And in March, the war shock pushed Brent above $116 and a record $12.3 billion fled ^[11]. The two big outflows came on macro spikes; the big inflow came on an AI unwind; and the money flipped back to selling within weeks. What precedes a reversal is visible in the tape, even if no one can promise the next one. That makes the third movement the interesting one: what India is doing with the room. On US election day itself, AdaniConneX — the Adani Group's data-center venture with EdgeConneX — was in talks for $1.5 billion of offshore bank debt on a five-to-seven-year tenor, meaning it matures in five to seven years ^[12]. The structure, not the timing, is the point: contract-backed debt financing for data centers, the exact model that just broke CoreWeave. CoreWeave's October defense of that model:

We structure the majority of our financing so that cash flows from customer contracts repay the associated debt and generate additional free cash flow. — CoreWeave

Reliance has committed ₹10 lakh crore — ten trillion rupees, about $120 billion — over seven years to sovereign data centers, and its chairman insists this outlay is different ^[13].

This is not a speculative investment. — Mukesh Ambani

The same insistence, one version already stress-tested. Union Bank of India forecasts that the country's AI financing will shift toward debt funding over the next three to five years, and recommends structured GPU-cluster financing, while flagging that GPUs go obsolete on 18-to-24-month cycles ^[14]. The central government has proposed a tax holiday to 2047 for foreign cloud companies serving overseas clients from Indian data centers, and state governments are bidding with subsidized land, power and water ^[15]. Nearly $60 billion of foreign AI data-center investment is planned, set against $1.16 billion of private AI investment in India today versus $109.1 billion in the US ^[15]. The counterweights are real and belong in the line. Crisil, also on election day, projected capacity growing five-fold by 2030 and noted that nearly half the new capacity is already committed under 15-to-20-year contracts that "limit speculative development" ^[16]. Ed Yardeni argues spending stays resilient because other states will court what Texas refuses — developers relocate ^[17]. Played straight, that does not refute the shift; it describes it. The buildout is not being vetoed. It is being re-zoned, and its leverage is moving offshore. Gartner has drawn the line where all of this meets. Its October warning put more than $3.5 trillion of AI infrastructure at risk of impairment by 2030, on the arithmetic that data-center owners need $8.5 trillion of cumulative revenue to justify the capex and a 12% return, while only $3.6 trillion is forecast ^[18]. The debt being negotiated on election day would come due five to seven years out, landing almost exactly where Gartner expects the bill to be presented. The refuge thesis and its contradiction will be tested in the same stretch of years.