India’s benchmark stock market may look like one of the global artificial-intelligence trade’s biggest laggards, but Goldman Sachs says a much stronger AI investment theme is developing beneath the index.

A group of 42 listed Indian companies exposed to the physical infrastructure needed for AI has risen about 60% in 2026, compared with a roughly 12% decline in the Nifty, according to a Goldman Sachs Global Investment Research report dated Thursday, 17 September.

The companies, which Goldman calls “AI Enablers,” have a combined market capitalization of about $670 billion and span power generation and transmission, electrical equipment, data centers and semiconductor manufacturing.

The performance makes the group the strongest pocket of the Indian equity market this year by a wide margin, Goldman said. Healthcare, the next-best sector or theme in its comparison, was up about 10%. All three broad AI infrastructure categories tracked by the bank, power, data centers and semiconductors, have risen between 40% and 80% this year.

The findings challenge a widely held view of India as what Goldman described as an “anti-AI” market because the country has relatively little benchmark exposure to the technology companies that have driven equity returns elsewhere.

Only about 16% of MSCI India’s market capitalization is classified by Goldman as AI-exposed, compared with 70% to 80% in South Korea and Taiwan and between 30% and 50% in China and Japan. Indian equities have also been among the weaker major markets since 2025 and have shown relatively little correlation with Asian AI stocks.

That index-level comparison, Goldman argues, misses companies supplying the electricity, equipment, computing infrastructure and semiconductor capacity required to build AI systems.

The distinction is becoming more important as India’s data-center industry expands rapidly. Operational data-center capacity reached about 1.75 gigawatts by the end of the first half of 2026, according to CBRE, which said $38 billion of new investment commitments were announced in the first six months of this year alone.

JLL separately estimates capacity could rise from about 1.6 GW in mid-2026 to 6 GW by 2029 as hyperscalers build facilities capable of handling increasingly power-intensive AI workloads.

India has also built a government-backed AI computing program. The electronics and information technology ministry said in March that 38,231 GPUs had been onboarded through 14 service providers under the IndiaAI compute framework.

From 1,800 stocks to 42

Goldman started with about 1,800 Indian listed companies with a combined market value of about $5 trillion. It first screened them for market size, liquidity, growth and investment intensity and then analyzed earnings-call transcripts and news reports for evidence that companies were generating revenue, committing capital, building order books or forming partnerships connected with AI infrastructure.

The process reduced the universe to 42 companies worth about $670 billion.

Companies needed a market value of at least $1 billion and average daily trading volume of at least $3 million over the previous six months. Goldman then looked for above-median revenue growth, capital expenditure growth, capex intensity or research and development spending before applying its AI-related textual screens.

The result bears little resemblance to a conventional technology-stock basket. Half of the 42 companies are classified as capital-goods businesses. Just 12 are large-cap companies, while the remainder consists of nine mid-caps, 13 small-caps and eight micro-caps. Utilities and technology hardware are also heavily represented.

Goldman divides the companies into nine groups spanning power generation, transmission and equipment; data-center developers, operators and hardware suppliers; and semiconductor hardware, materials and outsourced semiconductor assembly and testing, or OSAT.

Data-center operators account for about $398 billion of the cohort’s total market capitalization, although that figure represents the full value of the companies rather than the value of their AI businesses. Power-equipment companies account for about $95 billion, power generators $47 billion and transmission companies $45 billion.

The companies identified by Goldman include Reliance Industries Ltd, Bharti Airtel Ltd, Larsen and Toubro, Tata Power, Power Grid Corp. of India, Adani Green Energy, Adani Energy Solutions, ABB India, Siemens, Hitachi Energy India, CG Power and Industrial Solutions, Polycab India, Netweb Technologies, Kaynes Technology and Sterlite Technologies.

Its semiconductor-related screen also includes Gujarat Fluorochemicals, Navin Fluorine International, Himadri Speciality Chemical and electronics and engineering companies with exposure to semiconductor hardware or packaging.

Earnings rather than just enthusiasm

One of the report’s more significant findings is that the rally has so far been driven more by earnings than by higher valuation multiples. Since the beginning of 2025, the AI Enablers have delivered a cumulative return of about 53%. Goldman calculates that earnings growth contributed 65 percentage points to that performance, while shrinking valuation multiples subtracted about 12 percentage points.

That differs from an AI rally based mainly on investors paying progressively higher multiples for expected future growth. Consensus forecasts compiled by Goldman call for the group’s earnings to grow 53% in 2026, 39% in 2027 and 29% in 2028. By comparison, MSCI India earnings are expected to grow about 16% in 2027, while the MSCI India small- and mid-cap universe is forecast at about 23%.

Goldman estimates that companies in its AI basket could contribute roughly two percentage points to Nifty 500 profit growth of about 16% in 2027 and 2028. The earnings outlook has also been moving in the opposite direction from the broader market.

Consensus estimates for the AI Enablers’ 2027 earnings per share have been upgraded by 22% so far this year, while estimates for the Nifty 500 have been cut by about 2%, Goldman said.

The shift is particularly pronounced in data-center hardware, where 2027 earnings estimates have been raised by about 95%. Power-equipment forecasts are up 15% and semiconductor-materials estimates by 8%.

The upgrades are not universal. Goldman said estimates for data-center developers have been cut by 15% this year and those for power generators by 14%.

A large capex cycle

The earnings gains are arriving alongside a sharp increase in investment. Goldman expects capital expenditure by its AI Enablers to grow about 65% in 2026 after increasing 18% last year.

The group alone could account for six percentage points of expected Nifty 500 capex growth of 16% this year, with much of the spending coming from power generation, transmission, data-center operators and semiconductor packaging companies.

Capex growth is expected to slow to about 12% in 2027, but Goldman forecasts that the group will remain free-cash-flow positive overall as revenue rises. The scale of India’s wider infrastructure buildout supports the thesis, although industry forecasts vary sharply.

Wood Mackenzie expects Indian data-center capacity to increase from 2.2 GW in 2025 to 12 GW by 2030, with capacity dedicated to AI rising almost 24-fold to about 6.5 GW.

Cushman & Wakefield uses a lower current capacity estimate but also sees a substantial pipeline, putting operational capacity at nearly 1.8 GW and planned or under-construction capacity at about 3.9 GW in the first half of 2026.

The wide range reflects differences in definitions, project pipelines and the treatment of announced versus operational capacity. What is consistent across the estimates is the direction of travel: AI is increasing demand not only for servers and chips but for electricity generation, transmission equipment, cooling systems, fiber and data-center construction.

Valuations remain high

The rally has not left the group cheap on conventional measures. Goldman calculates that its AI Enablers trade at about 36 times forward earnings, an approximately 85% premium to MSCI India and close to the upper end of their five-year valuation range of roughly 28 to 39 times earnings.

The bank argues that much of that premium can be explained by faster expected profit growth. The group’s price-to-earnings-to-growth, or PEG, ratio is about 1.3 times, slightly below MSCI India’s 1.4 times.

There is considerable variation underneath that average. Goldman puts forward multiples at about 66 times for semiconductor hardware, 48 times for power equipment and 45 times for semiconductor OSAT companies, compared with roughly 20 times for MSCI India.

That leaves investors exposed to execution risk if projected earnings or AI infrastructure spending fall short.

Goldman also cautions against treating its screen as a definitive list of AI winners.

Its methodology can produce both false positives because it begins with a top-down screen and false negatives because its liquidity and minimum-market-cap requirements can exclude smaller companies with genuine AI exposure.

The market backdrop remains difficult as well. The Nifty closed at 23,217.6 on Wednesday and was still down roughly 12% for the year, with Indian equities facing pressure from high oil prices, rising global bond yields and foreign investor selling.