AI infra capex is surging 65% for India’s ‘AI Enablers’: Goldman Sachs

The latest market report highlights that India’s artificial intelligence build-out is moving into a capital-intensive phase, with firms linked to power, data centres and semiconductor infrastructure anticipated to sharply gain spending this year.

Capital expenditure by Goldman Sachs’ basket of 42 Indian “AI Enablers” is anticipated to jump 65 percent in calendar 2026, accelerating from 18 percent expansion in 2025, according to the investment bank. The spending is anticipated to contribute around six percentage points to aggregate Nifty 500 capex expansion, taking the broader index’s capex expansion to 16 percent in 2026 from 7 percent in 2025.

The spending is being led by power generation, power transmission, data-centre operators and semiconductor outsourced assembly and testing (OSAT), Goldman stated.

The bank anticipates the pace of capex expansion among the AI Enablers to moderate to 12 percent in 2027, but that would still be above the broader market.

The numbers offer a different view of India’s AI exposure than the benchmark-level picture. While India has relatively little AI exposure in its major indices, Goldman’s analysis points to a sizeable physical infrastructure supply chain developing underneath the index.

Power is the biggest driver of the capex cycle

Of the six percentage points that AI Enablers are anticipated to contribute to Nifty 500 capex expansion in 2026, power generation accounts for around three percentage points.

Power transmission, data-centre operators and semiconductor OSAT each contribute around one percentage point, according to Goldman’s estimates.

The remaining sub-layers – semiconductor materials, power equipment, data-centre developers, data-centre hardware and semiconductor hardware – have little or no contribution to aggregate Nifty 500 capex expansion in the bank’s calculation.

The gain in spending is additionally accompanied by elevated topline expansion across several parts of the cohort.

Goldman’s consensus estimates show that firms across the AI infrastructure chain are increasing capex alongside topline expansion, rather than simply investing ahead of demand.

The bank anticipates the cohort to stay free-cash-flow positive despite the capex upcycle.

Free cash flow as a share of sales is estimated at around 6 percent for AI Enablers in 2026 and 7 percent in 2027, compared with around 3 percent and 5 percent, respectively, for the Nifty 500.

From 1,800 firms to a $670 billion AI infrastructure basket

Goldman built its AI Enabler universe from around 1,800 Indian listed firms with a combined market capitalisation of around $5 trillion.

It first screened firms on size, liquidity, expansion and investment intensity, narrowing the universe to around 460 firms worth $4 trillion. Goldman then overlaid analysis of earnings-call transcripts and news flow to identify firms showing progress towards AI infrastructure and supply chains.

After further cross-checks, the final basket contained 42 firms with a combined listed market capitalisation of around $670 billion.

The screening framework required a minimum listed market value of $1 billion and average daily trading volume of at least $3 million over six months. Firms were additionally screened for above-median topline expansion, capex expansion, capex intensity or R&D intensity, alongside AI-related news or management discussions.

Goldman cautioned that the methodology can produce both false positives and false negatives, including because its minimum size and liquidity requirements can exclude smaller firms.

The 42 firms are spread across three broad layers – power, data centres and semiconductors – and nine sub-layers.

Data-centre operators account for the largest portion of the basket, at around $398 billion of listed market capitalisation, followed by power equipment at $95 billion, power generation at $47 billion and power transmission at $45 billion.

Data-centre hardware accounts for around $36 billion, semiconductor OSAT for $29 billion, semiconductor materials for $14 billion, data-centre developers for $4 billion and semiconductor hardware for $3 billion.

A small and mid-cap-heavy infrastructure play

The basket is considerably different from the composition of India’s large key market indices.

Of the 42 firms, 12 are large caps, nine are mid caps, 13 are small caps and eight are micro caps, according to Goldman.

Capital goods is the largest sector exposure, accounting for 21 of the 42 firms. Utilities and technology hardware and semiconductors are the other major areas of exposure.

The concentration means the physical AI infrastructure theme is spread across a range of firms that are not necessarily large constituents of the headline indices.

What firms are actually spending on

The power portion of the basket includes Adani Green Energy, Tata Power, NTPC Green Energy, ACME Solar and Clean Max Enviro Energy Solutions.

Power transmission and equipment exposure includes Power Grid, Adani Energy Solutions, ABB India, Cummins India, Siemens, Hitachi Energy India, GE Vernova T&D India, APAR Industries, Kirloskar Oil Engines, Schneider Electric Infrastructure, TD Power Systems, MTAR Technologies, Diamond Power Infrastructure and Waaree Renewable Technologies.

The semiconductor and technology-hardware portion includes Gujarat Fluorochemicals, Navin Fluorine, Himadri Speciality Chemical, Netweb Technologies, CG Power, Waaree Energies, Kaynes Technology, Paras Defence, Sterlite Technologies and Syrma SGS Technology.

The broader basket additionally includes Polycab India, KEI Industries, HFCL, Blue Star, Craftsman Automation, Reliance Industries, Bharti Airtel, Larsen & Toubro, Adani Enterprises, Anant Raj and Brigade Enterprises.

Goldman’s estimates show how differently these firms are positioned within the investment cycle.

For CY27, estimated capex intensity – capex as a share of sales – is 613 percent for NTPC Green Energy, 487 percent for ACME Solar, 242 percent for Clean Max Enviro Energy Solutions and 235 percent for Adani Green Energy.

Among other firms, capex intensity is estimated at 30 percent for Adani Enterprises, 27 percent for Waaree Energies and 26 percent for Kaynes Technology.

These are consensus estimates compiled by Goldman and do not represent individual Goldman earnings forecasts.

AI infrastructure discussions visible in corporate commentary

Goldman stated hard disclosures around AI-related topline and investments stay limited among Indian firms.

That stated, references to AI in earnings calls and investor presentations have increased sharply during the past two years, with the gain spreading beyond the technology sector.

For the infrastructure side of the theme, Goldman found growing references to data centres, power purchase agreements, fibre infrastructure, transformers, switchgear, uptime, substations, hyperscalers, NVIDIA, GPUs and OSAT.

The bank tracked mentions of “AI and Machine Learning”, “Automation” and “Generative AI” in earnings-call transcripts and investor-day presentations.

Discussions specifically related to AI infrastructure across power, data centres and semiconductors have additionally increased in recent years.

Goldman used this textual analysis alongside financial screens and news flow to identify firms with visible exposure to the infrastructure build-out.

The capex cycle comes with a valuation premium

The physical infrastructure opportunity is already reflected to varying degrees in valuations.

Goldman’s 42-stock basket trades at around 36 times forward earnings, an approximately 85 percent premium to the MSCI India index and near the upper end of its own five-year range of 28-39 times.

The bank stated the premium narrows when valuations are adjusted for the cohort’s stronger anticipated earnings expansion.

The AI Enablers’ PEG ratio is around 1.3 times, slightly below MSCI India’s 1.4 times, according to Goldman.

There is additionally a wide valuation gap between individual infrastructure layers.

Semiconductor hardware trades at around 66 times forward earnings, followed by power equipment at 48 times and semiconductor OSAT at 45 times. Semiconductor materials are around 39 times and data-centre hardware around 35 times.

Power generation is around 30 times, while data-centre operators and developers are around 26 times and 20 times, respectively. The physical AI build-out

The spending data points to an AI infrastructure cycle that extends well beyond firms directly selling AI software or computing services.

Power generation and transmission are taking a large share of the incremental investment, while data-centre operators and semiconductor OSAT are additionally contributing meaningfully to the gain in Nifty 500 capex.

Goldman anticipates AI Enabler capex expansion to moderate from 65 percent in 2026 to 12 percent in 2027, but the cohort is anticipated to stay free-cash-flow positive.

For India’s broader market, the more immediate significance is the contribution to corporate investment: AI Enablers alone are anticipated to account for around six percentage points of the Nifty 500’s 16 percent capex expansion in 2026.