NextFin News - A growing debate over restraint in artificial-intelligence capital spending is turning into a tailwind for India's technology-services exporters, lifting the outlook for a sector that has spent much of 2026 under pressure from fears that AI would eat its own lunch. The Nifty IT index, which closed at 31,496.70 on September 1, has recovered about 20% from its July lows even as the same AI narrative that battered the group earlier in the year now works in its favour: clients are shifting from building AI capacity to demanding measurable returns on it. The data in this report are as of September 15, 2026.
The reversal is rooted in a simple arithmetic problem on Wall Street. The five largest hyperscalers - Amazon, Microsoft, Alphabet, Meta and Oracle - are on track to spend between $700 billion and $900 billion on capital expenditure in 2026, a 36% increase over 2025, according to CreditSights estimates, with roughly $450 billion of that tied directly to AI infrastructure. Against that, the AI ecosystem is generating far less in actual sales, a gap that Sequoia's David Cahn has estimated at about $600 billion a year. That mismatch is forcing a rethink: if the returns on AI infrastructure are not arriving as fast as the spending, then the next phase of AI investment will favour deployment, integration and governance over raw compute build-out. That is precisely the work India's system integrators sell.
The Mechanism: From Capex Boom to Deployment Boom
The first-order effect of the AI-restraint debate is visible in how hyperscalers are being treated by investors. Shares of Amazon, Meta and Microsoft have tumbled on days when Alphabet and Meta raised capital-spending forecasts, and more than $1 trillion was wiped from Big Tech market capitalisations in a single week earlier this year on concerns over the scale of AI spending. The second-order effect is what matters for India: a slowdown in the rate of infrastructure build-out does not mean less AI work, it means different AI work.
For two years, the dominant question in enterprise technology was how much compute to buy. The dominant question in 2026 is what return that compute is producing. Brokerage Anand Rathi expects a multi-year growth cycle for Indian IT services beginning in fiscal 2027 precisely because global clients are shifting focus from building AI capacity to generating tangible financial returns. That transition creates demand for system integration, data governance, cybersecurity upgrades and scaled deployment of AI solutions - services where Indian vendors have the labour arbitrage, the domain experience and the client relationships to win.
The mechanism runs through three channels. First, ROI scrutiny pushes enterprises away from experimental pilots toward production deployments, which are integration-heavy and increasingly billed on outcomes rather than pure headcount. Second, as hyperscalers themselves face pressure to justify spending, they are more likely to outsource implementation work rather than build large internal professional-services armies. Third, the deflationary pressure on traditional maintenance and coding work - the bear case for Indian IT - is partly offset by new AI-led transformation contracts that carry higher value per engagement even if they consume fewer hours.
India's largest vendor has already signalled that it is willing to sacrifice short-term billing for this transition. K Krithivasan, chief executive of Tata Consultancy Services, addressed the Nasscom Technology and Leadership Forum in Mumbai and told employees that if they find they can do something faster, better and cheaper with AI, they should tell customers:
"even if it cannibalises revenue." He added, "We are not afraid this technology will take away our livelihood. We believe it is going to open up more."
That is an unusual statement for a services chief executive, whose compensation is typically tied to revenue growth. It is also an honest admission that the old billing model is not coming back.
The Evidence: AI Revenue Is Real, But Still Small
The bullish case rises or falls on whether AI revenue is large enough to matter. The numbers so far are promising but not yet transformative. TCS reported annualised AI services revenue of $2.6 billion in the first quarter of fiscal 2027, up 13.6% sequentially, alongside total revenue of ₹72,275 crore, up 14% year over year, and net profit of ₹13,349 crore, up 4.6%. Its total contract value for the quarter was $9.5 billion, and dollar revenue of $7.624 billion implies an annualised run rate of roughly $30 billion. Infosys has indicated that roughly 8.2% of its total revenue is tied to AI-related services. HCL Technologies and LTIMindtree are also scaling specialised offerings.
But the scale math is unforgiving. TCS's $2.6 billion of annualised AI revenue sits against a total revenue base of roughly $30 billion a year. Even compounding at 13.6% a quarter, AI would need several years to replace the legacy services base that is under deflationary pressure. This is the crux of the debate: the restraint narrative helps the sector's valuation multiple before it helps the sector's earnings. The market is paying today for a future earnings stream that has not yet arrived.
That sequencing explains why the stock move has run ahead of the fundamentals. The Nifty IT index's 20% rebound from July lows was supported by better-than-expected earnings and by global capital rotating out of expensive US technology stocks into relatively cheaper Indian names. Shares of Infosys and TCS have each fallen more than 40% since the start of 2025 as investors assessed AI's long-term impact, leaving valuations that looked attractive to global funds even before the AI-restraint thesis took hold. The sector's rally is, in large part, a re-rating of pessimism rather than a celebration of new earnings.
The depth of that pessimism is worth measuring. In February 2026 alone, Indian IT firms collectively lost about $68.6 billion in market value as the Nifty IT index fell 21%, its steepest monthly decline in nearly 23 years. A move of that magnitude was not pricing a normal earnings cycle; it was pricing an existential question about whether the outsourcing model survives AI. The current rebound is the market beginning to answer that question with something other than panic.
The Counter-Thesis: Deflation Is Still Winning
The strongest case against the bullish read is that AI-led deflation is a structural force, not a cyclical one, and that it will compress pricing faster than new AI revenue can grow. Kotak Institutional Equities has noted that AI adoption could bring deflationary pricing pressure on traditional services, with some brokerage estimates pointing to a 20% to 50% reduction in pricing as AI streamlines processes. CLSA downgraded Tata Consultancy Services, Infosys and Tech Mahindra to Hold and Wipro and Mphasis to Underperform, arguing that larger system integrators are in a tougher spot in the short term. UBS upgraded Tech Mahindra to Buy but downgraded HCLTech and Persistent Systems to Neutral, cautioning that AI-driven revenue may not be sufficient to offset margin erosion in traditional services in the near term.
The cannibalisation risk is not theoretical. The top five Indian IT firms saw their combined workforce shrink by 7,389 in fiscal 2026, and companies have been reluctant to give definitive hiring guidance for fiscal 2027. When a services company's revenue is a function of billable hours and AI reduces the hours required, the burden of proof sits with the bulls to show that new work arrives fast enough to fill the gap. TCS's own workforce stood at 593,798 at the end of the June quarter, down from 613,069 a year earlier, even as the company posted profit growth. TCS's chief human resources officer, Sudeep Kunnumal, attributed the reduction to a combination of voluntary and involuntary attrition rather than AI-driven layoffs alone - but the direction of travel is the same either way.
The answer to the counter-thesis is one of timing and mix, not denial. Deflation in legacy work is real and structural - no amount of integration demand reverses the fact that coding and testing are becoming cheaper. But the same force that compresses legacy pricing also lowers the cost of delivering new AI projects, and the vendors that can migrate talent from maintenance work to AI-centred engagements fastest will see margins stabilise sooner. The restraint debate helps because it slows the pace at which clients rip out legacy contracts, giving vendors time to transition rather than being forced to reprice everything at once.
Cyclical Wave Meets Structural Shift
The correct read is that both forces are at work and they point in opposite directions across time horizons. The cyclical leg is the valuation relief: a rotation out of US technology, a pause in the AI-capex arms race, and a temporary reprieve from the most aggressive AI-disruption fears. That leg is mean-reverting - if hyperscalers resume aggressive spending or if US technology valuations re-rate higher, the rotation trade unwinds and the sector gives back its multiple expansion.
The structural leg is the business-model reset. The era of billing clients for large teams doing routine coding, testing and maintenance is ending. What replaces it is a mix of outcome-based AI projects, platform-led orchestration and managed services around AI governance. That shift is not mean-reverting; it is a permanent change in how software services are priced. The winners will be the firms that can execute the transition without a prolonged margin trough, and the losers will be the ones that defend the old model until it defends nothing.
Mid-cap vendors face the sharpest version of this problem. If hyperscalers prioritise cost-saving measures over new project deployments, mid-tier firms with less diversified revenue and thinner balance sheets may see muted growth in fiscal 2027 even as the large caps benefit from mega-deal renewals. That divergence is already visible in brokerage preferences, with some analysts favouring Coforge and Persistent Systems for agility while downgrading the largest names for exposure to legacy deflation. The sector is not moving as one; it is splitting along the line between those selling transition and those selling continuity.
There is also a currency cushion that the bears often underweight. A weaker rupee lifts reported revenue and margins for exporters, and brokerage forecasts for the top six firms assume a weaker rupee accounts for a meaningful share of the roughly 10% year-on-year revenue and profit growth expected in fiscal 2027. Strip out the currency effect and the underlying volume growth is thinner - which is another way of saying the recovery is real but fragile, and dependent on forces outside the sector's control.
What to Watch
The falsifying signal for the bullish view is straightforward: if annualised AI revenue at the large vendors fails to grow at a double-digit quarterly pace for two consecutive quarters, or if AI-linked revenue as a share of total revenue stalls below 10% through fiscal 2027, the deployment-boom thesis is wrong and deflation is winning. TCS's 13.6% sequential growth in AI revenue is the benchmark; two quarters below 10% would break the narrative. Conversely, if hyperscalers resume raising capital-spending guidance while Indian IT vendors report accelerating AI-led deal wins, the restraint narrative was merely a pause within a continued capex supercycle.
By time horizon: in the short term, sentiment and fund flows dominate, and the sector can rise on valuation relief alone. Over the medium term, fiscal 2027 earnings will separate vendors that are actually converting AI demand into revenue from those merely talking about it. Over the long term, the structural question is whether Indian IT can reinvent its billing model fast enough to survive the deflation of its core product.
The base case is a grinding recovery rather than a V-shaped rebound: large vendors with strong balance sheets and diversified client bases navigate the transition, mid-caps face more pressure, and the index grinds higher on multiple expansion even as earnings lag. The upside case is a faster-than-expected shift to ROI-driven deployments that lifts AI revenue share into the mid-teens by fiscal 2028. The downside case is a resumption of hyperscaler spending discipline that starves vendors of new work while legacy contracts reprice faster than expected.
The market is not pricing a cyclical bounce here; it is pricing the first credible sign that the AI disruption trade may have been overdone. Whether that sign proves durable depends less on what hyperscalers spend on chips and more on what enterprises spend on the people who make those chips useful.
Explore more exclusive insights at nextfin.ai.
