Three proposals are open on your desk and all of them say India. One is from a listed firm with tens of thousands of employees and a New York sales office that replies within four hours. One is from a 400-person company in Coimbatore a former colleague vouched for. The third is six people in Pune who built something close to what you need for another client, at less than half the price. You cannot compare them, because the only mental model you have for outsourcing to India came from a magazine article written around 2006.
That model is twenty years stale, and the part that changed most is the part nobody explains to buyers.
India’s technology services sector is not one market. It is at least five, with different economics, different buying processes and different ways of going wrong: IT services, business process management, engineering R&D, software products, and global capability centres, which are not outsourcing at all because the client owns the centre. Deciding what you are buying comes before deciding who you buy it from. Most bad India engagements start with a buyer who picked the wrong segment and then blamed the country.
The size and shape of it, with the caveats attached
First, where these numbers come from. Most headline India IT figures originate with NASSCOM, the industry’s own trade association. They are estimates compiled from member reporting, not audited national accounts, and NASSCOM has restated its own series before, notably the capability centre numbers it rebased in September 2024. Treat them as the industry describing itself. Government services-export data is independent and tells a compatible story.
NASSCOM’s Strategic Review 2026 puts the Indian technology sector at $315 billion in revenue for FY2026, up 6.1%, with exports of $246 billion. The segment split: IT services $149 billion, engineering R&D $63 billion, business process management $59 billion, software products $23 billion, hardware $21 billion. The employee base is around 5.95 million, net additions roughly 135,000, AI-related revenue estimated at $10 billion to $12 billion.
Hold two of those next to each other. Revenue up 6.1%, headcount up 2.3%. The industry that sold bodies for twenty years is decoupling from bodies, which is the most useful fact to carry into a proposal priced per seat.
Independently, the Economic Survey 2025-26 (Ministry of Finance, Chapter 7) records services exports at 10.0% of GDP in the first half of FY2026, with India the world’s seventh-largest services exporter and its global share up from 2% in 2005 to 4.3% in 2024. Computer and software services are over 40% of that, but the fastest-climbing line is not software: professional and management consulting went from 10.5% to 18.3% of services exports across FY23 to FY25. The Reserve Bank put net services receipts at $60.4 billion in the final quarter of FY2026, against $53.3 billion a year earlier.
Five different businesses, all called outsourcing to India
IT services. Application development, maintenance, infrastructure, cloud, testing. The $149 billion core. Sold through an MSA plus statements of work, staffed from a pyramid with a few senior people over a wide junior base. Works when your scope is definable and your own technical management can review the output. Fails when you buy a fixed price for something you cannot specify.
Business process management. The old BPO, now covering finance and accounting, claims and revenue cycle, customer contact, procurement, moderation, data operations. Priced per FTE, per transaction or per outcome, and the pricing basis matters more than the rate: per-transaction makes falling volume the vendor’s problem, per-FTE makes it yours.
Engineering R&D. $63 billion, and the segment Western buyers most consistently underrate. Embedded systems, semiconductor design, automotive, medical devices, aerospace. Longer sales cycles, harder IP questions, engineers who have spent a decade inside one product family. If your work sits closer to hardware than to a web app, the vendor list is entirely different.
Global capability centres. The real 2026 story, and the one that breaks the word outsourcing. A GCC is a wholly owned subsidiary: your company incorporates an Indian entity, hires Indian employees directly, runs them as part of your own organisation. No vendor, no margin, no contract. The Nasscom-Zinnov 2026 report counts 2,117 GCCs operating 3,728 units, employing about 2.36 million people, with revenue near $98.4 billion as of March 2026, and 506 of the Forbes Global 2000 running one. The Economic Survey’s own series lags and defines the category more narrowly: 1,700-plus centres and 1.9 million professionals in FY2024, revenue up from $40.4 billion in FY2019 to $64.6 billion in FY2024. Two credible sources, a wide gap. Nobody has a clean census, and anyone quoting one authoritative GCC number is over-claiming.
The long tail. Thousands of firms between five and five hundred people. No analyst coverage, no press releases, frequently the best value in the market and by far the hardest to verify. This is where the six-person Pune quote came from.
You are almost certainly too small to open a GCC; the threshold where an owned entity beats a vendor usually sits north of 50 to 100 people. It matters to you anyway, because it competes for the same staff. When two thousand owned centres hire directly at the top of the market, with the brand of a global product company and no margin in between, the strongest senior engineers in Bengaluru have a better option than your vendor’s payroll. That is the mechanism behind most “they sent us their B team” complaints. It is rarely a bait and switch. The A team got recruited by a captive. So name individuals in the contract, interview every replacement, and write in a rate floor.
Bengaluru, or a city you have never heard of
Tier-1 means Bengaluru, Hyderabad, Pune, the National Capital Region, Chennai and Mumbai. Tier-2 is Coimbatore, Kochi, Indore, Ahmedabad, Jaipur, Mohali, Bhubaneswar, Visakhapatnam and a couple of dozen more. The Economic Survey notes capability centre expansion into tier-2 and tier-3 cities supporting more dispersed employment, which is the government’s way of saying the map moved.
- Cost fell, but less than the brochure implies. Real estate, facilities and living costs are meaningfully lower. Senior salaries are not, because senior people are nationally mobile and price themselves against Bengaluru. The saving is real at junior and mid bands, and thins fast at the top.
- Attrition is usually lower, for one reason. In a city with fifteen employers hiring your skill, changing jobs means an email. In a city with three, it means moving house. That works in your favour until you need a replacement, when the same thinness works against you.
- Depth is the real question, not availability. Competent engineers exist everywhere in India. Whether forty of them locally have your specific stack, so a resignation is a four-week problem rather than a four-month one, is a city-by-city and skill-by-skill question. Ask how long the vendor’s last three replacements there took, with dates.
- Check where your people actually sit. A vendor with a Coimbatore facility can still staff your account from Bengaluru, at Bengaluru cost and Bengaluru attrition. Put the delivery location in the SOW.
What actually drives the price
Published Indian rate cards are close to worthless. Every one is a marketing artefact, and the real number moves on several axes at once. What to interrogate in the quote you were given:
- Experience band. The steepest variable by far. Zero to three years, three to eight, eight-plus and architect level are four different markets, and the gaps are proportionally wider in India than in the US.
- City. Material at the junior end, close to irrelevant at the top.
- English requirement. Badly understood by Western buyers. India has an enormous English-reading and English-writing workforce, and a much smaller pool comfortable on an unscripted voice call with an irritated American customer. Written technical English costs nothing extra. Customer-facing spoken English is scarce and prices like it.
- Domain certification. A medical coder holding a CPC, a finance lead who has run a US close, a nurse doing clinical review: credentialed roles competing with domestic employers, at a clear premium over a generic graduate hire.
- Shift. Night work for US coverage carries specific Indian overheads: shift allowance, employer-provided night transport and security for women workers under several states’ shops and establishments rules, and higher attrition on permanent nights. A model where India works its own day is materially cheaper than one where it works yours.
- What sits between the rate and the salary. Employer provident fund contributions at 12% of basic wages, gratuity provision, seat and facility cost, supervision, bench capacity, training, margin. All legitimate, all worth asking about.
Which leads to the least asked useful question in an India negotiation: what proportion of this rate reaches the person doing the work? A vendor willing to discuss the split is telling you something real. A rate that buys a below-market salary buys a person who leaves inside a year, and their replacement costs you six weeks of context you paid to build.
Yes, India outsources to the Philippines
This confuses people and it should not. Indian-headquartered providers run delivery centres in Manila and Cebu alongside Eastern Europe, Latin America, South Africa and Egypt. Infosys BPM publishes 42 delivery locations across 14 countries including the Philippines; WNS lists the Philippines among its delivery geographies. IBPAP, the Philippine industry association, puts its IT-BPM sector at roughly $40 billion in revenue and 1.9 million workers, undated on its own site, so treat those as you treat NASSCOM’s.
Four real reasons the work moves. Voice: a deeper pool of workers whose spoken English reads as neutral to a US customer, plus cultural familiarity that reduces friction on unscripted, empathetic calls. Night shift economics: Manila built a labour market, and the transport and social infrastructure around it, on US working hours. Graduate labour cost: at entry level for non-technical process work the two countries are close enough that capability and attrition decide the location, not rate. Continuity: a provider with sites in three countries can move a process after a flood or a grid failure.
For your contract: establish which country and which legal entity performs the work, and whether any part is subcontracted to a third party rather than delivered by a subsidiary. Then require written approval before that changes. Multi-country delivery is fine. Multi-country delivery you discover during an incident is not.
The rules, and how to verify a vendor before you sign
Three frameworks in brief. The Digital Personal Data Protection Act 2023 is India’s general privacy statute, with implementing rules notified in November 2025 and obligations phasing in over roughly 12 to 18 months. One structural point before you lean on it: the Act largely disapplies itself where an Indian entity processes non-Indian residents’ data under a contract with a foreign client, preserving the security safeguards duty and little else. Protection for your customers comes from your contract and your own law. CERT-In’s 2022 directions require Indian entities to report specified cyber incidents within six hours and hold ICT logs for 180 days within Indian jurisdiction, a useful lever when you ask for log access. SEZ or STPI status concerns the vendor’s own export and tax position, not yours.
Then do the checks. India has unusually good free public company data and almost no Western buyer uses it.
- Get the CIN and read it. Every Indian company has a 21-character Corporate Identity Number encoding listing status, industry code, state of registration, year of incorporation, ownership type and registration number. The year is right there in the string. A firm claiming fifteen years of delivery experience whose CIN says 2022 has answered a different question than the one you asked.
- Pull the master data from MCA. The Ministry of Corporate Affairs portal gives free public company and LLP master data: registered office, incorporation date, authorised and paid-up capital, status, date of last annual general meeting and last balance sheet filed, plus directors, signatories and registered charges. Two red flags: filings years out of date, and paid-up capital of a few hundred thousand rupees behind a claim of running a 2,000-seat operation.
- Check the GSTIN. The GST portal lets anyone confirm a taxpayer’s legal name, registration date, status and address. Match it to the entity on the contract; a mismatch between who pitches and who invoices deserves an explanation.
- Sanity-check headcount through EPFO. Almost nobody uses this. Provident fund coverage is mandatory for establishments with 20 or more employees, and EPFO runs a public establishment search where you can find an employer by name or code and view the payment details filed against it. Not a clean headcount, since groups hold multiple codes. Still the difference between “we have 1,200 people” as a claim and as something with a paper trail.
- Ask for attrition on the right basis. Every listed Indian IT firm publishes trailing-twelve-month attrition in its quarterly results, so you have a benchmark. Ask your private vendor for the same metric for the delivery unit that will serve you, not company-wide. One that has never computed it per account has told you how it manages accounts.
- Take references the hard way. Two clients of your size in your time zone, one current and one that has left, then a third you find yourself who was not on the list. The ex-client conversation is worth more than the other two combined.
Related reading
- Will Outsourcing to India Be Restricted? How to Check for Yourself
- Is ₹6 Lakh for 875 Hours a Fair App Development Quote in India?
Six people or sixty thousand
Size is not quality. It is a different set of trades, and the answer depends mostly on how big you are relative to them.
The very large firm sells process and continuity: an escalation path, a compliance function, a bench, cover when someone is ill, a name your board recognises. You also get a pyramid built around utilisation targets, an account manager paid to expand the contract, and a position in their revenue so small that nothing you say reaches anyone with authority. Leverage is purely contractual, so the contract has to be good.
The small specialist sells the opposite. The founder is on the call, the people are senior because there is nobody else, decisions happen in an afternoon. There is also no bench, so one resignation is a crisis, key-person risk sits in two or three heads, and the firm may not survive losing a different client. Often no formal security programme either, which becomes your problem the first time one of your own customers sends you a vendor questionnaire.
The middle, roughly a hundred to a thousand people, is where most buyers spending $200,000 to $2 million a year should look. Big enough for an HR function, a security policy and cover for a sick engineer. Small enough that you are a name rather than an account number.
An opinion I will defend: aim to be between 1% and 10% of your vendor’s revenue. Below 1% you cannot get attention when you need it. Above roughly 15% you become their existential dependency, which sounds like leverage and actually means they agree to things they cannot deliver rather than lose you. Ask what share your contract would be. The pause before the answer is informative.
Two terms worth getting straight before the next call. An offshore development centre is dedicated capacity inside a vendor’s facility: their entity, their employees, your team, your logo on the wall. A global capability centre is your own legal entity employing your own staff. One is a contract, the other is a subsidiary, and pitches use the terms loosely. And NASSCOM membership, which appears on most vendor websites, is a paid trade association subscription rather than an accreditation, audit or certification. It tells you the company is real. It is evidence of no particular control.
If you have a shortlist and want the verification done properly before you sign, or you would rather stand up a small dedicated India team without a three-year master agreement, that is routine work for AB7 Solutions. We run contract staffing, staff augmentation, C2C and RPO alongside BPO and KPO delivery, healthcare support including medical scribing and revenue cycle, and data annotation and AI data operations, so we can give you a straight read on both sides, including when the right answer is a vendor that is not us. Call +1 321 341 7733, email ab@ab7solutions.com or the director at director@ab7solutions.com, or read more at www.ab7solutions.com.
Sources: NASSCOM, Technology Sector in India: Strategic Review 2026 (FY2026 revenue, segment split, exports, headcount; industry-body estimates compiled from member reporting); Nasscom-Zinnov, India GCC report 2026 (2,117 GCCs, 3,728 units, 2.36 million people, $98.4bn, as of March 2026; also industry-body data); Ministry of Finance, Economic Survey 2025-26, Chapter 7: Services; Reserve Bank of India, Developments in India’s Balance of Payments, Q4 2025-26 and RBI Annual Report 2025-26; MeitY, Digital Personal Data Protection Act 2023; CERT-In, Directions under section 70B(6), 28 April 2022; Department of Commerce, SEZ Act 2005; Ministry of Corporate Affairs company master data; EPFO public establishment search; GST taxpayer search; IBPAP (Philippine IT-BPM revenue and headcount, undated on source); company disclosures from Infosys BPM and WNS. Checked September 2026.
Related AB7 service: full list of AB7 services.