India is the one large market not in the AI trade. That’s now an asset, not a shortcoming
DSP Mutual Fund August 2026 Netra Webinar The DSP Netra August 2026 briefing opens with a sharp reframe: the world is paying record prices for AI capacity it cannot yet monetise. India is the one large market not in that trade — and that is now an asset, not a shortcoming. India’s software-heavy, semiconductor-light composition is a diversification asset right now, not just a hedge against a future reversal. DSP Netra August 2026: the AI concentration risk case This month’s DSP Netra August 2026 briefing comes from Sahil Kapoor, Head of Products & Market Strategist at DSP Mutual Fund. Read alongside the Kotak August outlook — which framed India as the “anti-AI destination if the trade reverses” — and the UTI July non-consensus call. Where Kotak positions India as a hedge, DSP goes further: India’s market composition is a diversification asset right now, and the bargain universe available today didn’t exist in 2024. AI HARDWARE TRADE Fence-Sit INDIA COMPOSITION Diversification Asset BARGAIN UNIVERSE 450–500 Names LT RETURNS Reset to 10–12% Summary The DSP Netra August 2026 call turns on the AI capex concentration risk. Technology now sits at 41% of MSCI ACWI market cap — larger than at the dot-com peak, and the entire move has come from hardware and semiconductors, not software. AI capex is now 2.4–2.7% of US GDP — roughly $1 trillion in 2026 alone — with five hyperscalers having spent $1.3 trillion over five years, more than the world’s largest oil and gas companies since COVID. HBM and memory prices have multiplied 20 to 30 times in a year, which means nominal spend is overstating real capacity creation. The return math doesn’t work yet: a $2–3 trillion cumulative capex programme needs $3–4 trillion of annual revenue to justify itself, while global IT spend after 40–50 years is $6 trillion. The gap has to come out of the $45 trillion global wage bill — and that isn’t a short exercise. India’s setup is the mirror image. IT services are now just 7.4% of the Nifty — below the global financial crisis low of 8.8%. Nifty price-to-book is below its long-term average with ROEs materially higher than at the December 2020 lows. Out of the BSE 1400-plus universe, 450 to 500 companies now screen as bargains with sub-20-times multiples available across financials, IT, select healthcare, auto and insurance — a set that simply did not exist in 2024. Two flat years have delivered a time correction, not a price correction. DSP’s calls: fence-sit on the AI hardware trade; own India as a diversification asset now; reset long-term return expectations to 10–12% before costs; and hunt the bargain universe. The detail The AI concentration risk — 41% of MSCI ACWI, larger than the dot-com peak Technology, counting IT and communication services together, is now 41% of MSCI ACWI market cap — larger than at the dot-com peak. And within it, the entire move has come from hardware and semiconductors, not software. A large part of the reported capex is price, not capacity. Some HBM and memory prices have multiplied 20 to 30 times in a year. GPU rental pricing and the semiconductor producer price index have both moved sharply higher, which means nominal spend is overstating real capacity creation. Funding is increasingly moving from cash flow to debt — having been almost entirely cash-flow-funded until recently. And China is undercutting on price with open-weight models while adding grid capacity in a single year equal to all of Germany’s — which puts the incumbents’ revenue visibility, and therefore their funding, at risk. The return math — the cascade that doesn’t compute yet Every $1 of infrastructure capex becomes $1.5 at the compute layer, $2.7 at the model layer and $4 at the end user. So a $2 to $3 trillion cumulative capex programme needs $3 to $4 trillion of annual revenue to justify itself. Global IT spend, after forty to fifty years, is $6 trillion. The gap has to come out of the $45 trillion global wage bill. That is the unanswered question — and it doesn’t get answered in a quarter. Business investment in computer and peripheral equipment in the US is growing 75% year on year; globally the number is close to $600 billion and approaching the dot-com peak as a share of GDP. That is roughly five to six years of normal demand bought in two years. Why India sits outside — IT at 7.4% of the Nifty, below the GFC low India sits outside the AI hardware trade almost entirely. IT services are now just 7.4% of the Nifty — below the global financial crisis low of 8.8%. That composition, until recently framed as a lack of exposure, is now the setup. All of MSCI EM’s return this year has come from technology, and within technology from semis and hardware. India’s more diversified market composition — software-heavy, semiconductor-light — could become a diversification advantage if the semiconductor and hardware trade normalises. Indian valuations have quietly reset. Nifty price-to-book is below its long-term average on both trailing and forward, at levels last seen in December 2020 but with materially higher ROEs. The same price, better quality. The bargain universe — 450 to 500 names that didn’t exist in 2024 Out of the BSE 1400-plus universe, 450 to 500 companies now screen as bargains, with sub-20-times multiples available across financials, IT, select healthcare, auto and insurance. That is a large enough set to build a genuinely better-quality portfolio — a set that simply did not exist in 2024. Two flat years have delivered a time correction rather than a price correction. The Nifty has spent over 103 days below its 200-day average with modest drawdowns — which is exactly the starting point that improves forward returns. Reset long-term returns to 10–12% before costs — the honest math Reset long-term return expectations to 10 to 12% before costs. Domestic sales growth tracks nominal GDP at 10 to 12%. Export growth lands at 7 to 11% including currency. Margins














