The artificial intelligence (AI) boom that India’s largest information technology giants spent three years championing now comes with a costly caveat: the token bill.
After first flagging the phenomenon of tokenmaxing—or excessive AI token consumption—in January, IT firms are now aggressively tracking these expenses. Swelling AI operational costs threaten to eat up enterprise tech budgets and drag down revenue growth, just as the industry enters what is expected to be its weakest quarter in years.
The fundamental unit of data processed by large language models (LLMs) like OpenAI’s ChatGPT or Anthropic’s Claude is a token—essentially the building blocks used by AI to analyse and generate text, image and software code. For Fortune 500 companies, which form the backbone of India's IT client base, running these models at scale can generate massive token bills, paid either directly to AI vendors or passed through tech partners.
High token consumption risks squeezing traditional IT budgets, as clients divert capital directly to AI infrastructure leaders. Alternatively, if clients shift the financial onus of purchasing AI compute units onto their IT service providers, it could structurally alter contract margins. The squeeze comes at a delicate time for software exporters already seeing traditional revenue streams contract. Automation tools have reduced the billable hours and headcount previously used to drive growth, according to a 24 June report by JPMorgan Chase & Co.
"Incremental data points highlight significant pass-throughs in new projects and token/cloud spending pressuring budgets available for IT services and software," JPMorgan analysts Ankur Rudra, Bhavik Mehta, and Harshit Sharma wrote, adding that the exact shape of an enterprise spending recovery remains obscured by an "AI fog."
Budget strains
The sheer unpredictability of generative AI usage is compounding the budget strain.
"Token costs are an increasing problem as they are growing, substantial, and often unpredictable in when and how many tokens are consumed," said Peter Bendor-Samuel, founder of US-based tech consultancy Everest Group. "They are currently a major factor in IT budget overruns."
Tokenmaxing has also caught the attention of Silicon Valley's top leadership, who are signalling that enterprise clients must move past experimental usage toward strict fiscal discipline.
“The hard truth is that the marginal cost of productivity improvement has to match the marginal cost of the token,” Microsoft Corp. CEO Satya Nadella said in a recent podcast, noting that managing token economics will be a core operational challenge over the next two years.
In response, tech vendors are rapidly rewriting their legacy billing models to protect margins.
Global IT provider EPAM Systems Inc. has begun factoring variable token costs directly into its deal structures, while Tech Mahindra Ltd is testing pricing models linked directly to client token consumption. Cognizant Technology Solutions Corp. has introduced dedicated rate cards for AI compute, while LTM Ltd. (formerly LTIMindtree) is pioneering a quota-based system where clients purchase fixed AI units that deplete as automated tasks are executed.
“Most organizations can speak about token utilization, but the outcomes that will be delivered using those tokens will be a critical measure for how IT services will be delivered. Contractual commitment will be more towards outcomes than token utilization,” said Namratha Dharshan, chief business leader for India Research at Information Services Group (ISG) is a Connecticut-based global tech advisory firm.
Tepid earnings loom
When the June quarter earnings season kicks off this week, five of India’s six largest IT services firms are projected to post their worst first-quarter performance in at least two years, according to research reports by six broking firms said.
Tata Consultancy Services, Infosys, HCL Technologies, Wipro and LTM will register sequential quarterly revenue growth of 1.8% at best, with Infosys expected to lead growth, the reports said.
"We believe that the first quarter of fiscal 2027 will be weak on revenues, impacted by the West Asia crisis and elevated productivity pass-throughs in managed services contracts," Kotak Institutional Equities analysts Kawaljeet Saluja, Sathishkumar S, and Vamshi Krishna wrote in a 1 July note. "Most large companies will struggle to hit the midpoint of their full-year guidance."
Tech Mahindra remains the lone outlier among the top-tier firms. The company is expected to post its strongest first-quarter growth in three years, driven by the execution of large telecom contracts. This includes a landmark five-year, $500 million IT modernization deal with Spanish telecom giant Telefónica SA's O2 unit, which Mint first reported in January.
Mid-size mania
Mid-sized software exporters—typically those generating between $1.5 billion and $2 billion in annual revenue, such as LTM and its immediate peers—are also bracing for a challenging quarter, though recent big deal wins and acquisitions may help.
Coforge Ltd is expected to lead the mid-tier pack, with brokerages forecasting sequential growth of as much as 22%. The Noida-based firm’s performance is heavily front-loaded by its April acquisition of Encora, a US-based digital engineering and data analytics firm. The $2.39 billion transaction stands as the largest outbound acquisition in the history of Indian IT, and is projected to push Coforge’s revenues past the $2.5 billion threshold by fiscal 2027.
Yet, analysts warn that isolated M&A success stories cannot mask a broader, systemic freeze in discretionary corporate tech spending which is expected to spill over into the next quarter.
"We expect demand commentary to stay soft as macro, AI, and geopolitical overhangs continue to weigh on decision-making cycles," Motilal Oswal Financial Services analysts Abhishek Pathak and Keval Bhagat wrote in a client note. "The soft start is likely extending into the July-September period as well."
As enterprise buyers freeze budgets to assess their AI expenses, the near-term outlook for India's tech outsourcing giants remains tied to how efficiently they can manage the transition from billing human hours to optimizing algorithmic tokens.
"Companies might allot a fixed portion of their IT budgets for AI token usage and ask software services firms to generate guaranteed outcomes," said Sushovon Nayak, lead IT analyst at Anand Rathi Institutional Equities. "How much is delegated to these tokens—and whether it represents incremental spend or a cannibalization of existing tech budgets—remains the key monitorable for the industry."