India-Plus-One in 2026: A South Asia Delivery Strategy Beyond Bengaluru

The concentration risk nobody prices

India's tech industry booked around USD 315 billion in FY2026 revenue and employs close to six million people, on Nasscom's February 2026 numbers. The global capability centre count has passed 2,100, with more than 2.3 million people inside them (that's the latest Nasscom-Zinnov tally). It is, by a distance, the most successful offshore delivery story in history.

It is also, for many companies, a single point of failure that never appears on a risk register. Ask a CIO how much of their engineering and operations capacity sits within a 40-kilometre radius of one Indian city and watch the arithmetic happen in real time. For a lot of mid-caps, I'd bet the answer is over half. For some it's nearly all.

The manufacturing world has spent the last decade or so working through this, and gave it a name. China-plus-one was never about leaving China; it was about not being destroyed by a single policy change, port closure, or currency move. Services have been slower to think the same way, partly because software feels weightless and partly because the India relationship has been so good for so long that questioning it feels ungrateful.

Related reading: Setting Up a Global Capability Centre in Pakistan in 2026: An Employer's Guide · Sri Lanka's Finance and Accounting Outsourcing Market in 2026: A Buyer's Guide · Outsourcing IT to Nepal in 2026: A Buyer's Guide to the Kathmandu Market.

What "plus one" actually buys you

Three things, in descending order of how often they justify the cost.

Wage arbitrage that hasn't been competed away. This is the honest primary driver for most programmes, whatever the risk-committee memo says. In Bengaluru, Hyderabad, and Pune, the combination of GCC expansion and product-company hiring has pushed senior engineering compensation to levels that make the original business case look quaint. A senior engineer in Bengaluru now costs far more than the same role did in 2018, and the counter-offer culture is brutal. Colombo, Lahore, Dhaka, and Kathmandu haven't been through that cycle to anything like the same degree.

Retention. This one surprises people. Attrition in a market with five credible employers behaves completely differently from attrition in a market with five hundred. Teams built in Kathmandu tend to stay built. Colombo used to be the same, until the 2022 economic crisis pushed a lot of mid-career people abroad. Over a three-year product roadmap, the value of institutional memory not walking out the door is larger than most cost models capture.

Actual continuity. If a policy change, a tax ruling, or an infrastructure failure takes an Indian site offline for a fortnight, having 20 percent of capacity somewhere else is the difference between degraded and stopped. This is the reason the board will approve and the reason that, in practice, motivates almost nobody to start.

Four candidate markets, side by side

The table is deliberately blunt, and the cost row is a rough guide rather than a quote. Kearney's comparison puts Sri Lanka around 10 percent below India on IT work; the other three swing a lot by role and seniority. If your requirement is 50 seats of solid engineering at the lowest defensible cost, Pakistan and Bangladesh are the serious answers. If it's 80 qualified accountants doing work you'd otherwise keep at head office, it's Colombo and it isn't close. If you want a small, stable, high-retention product team and you don't need to scale it fast, Kathmandu is underrated (Cotiviti's captive engineering centre there traces back to 2004) and will stay that way for a while yet.

The failure mode: a second site that's just a smaller first site

Here is where these programmes actually die, and it's not cost and it's not talent.

A company opens site two, staffs it with the overflow work nobody at site one wanted, gives it no ownership of anything end-to-end, and reports on it against site one's productivity metrics. Eighteen months later the numbers look bad, someone writes a memo about how the diversification experiment didn't work, and the capacity quietly consolidates back to Bengaluru.

The diagnosis is always the same. Site two was never given a domain. It was given tickets.

There's a related version that kills the idea before it starts, and it's political rather than operational. Site one's leadership is asked to help plan the diversification. They are, in effect, being asked to design the thing that shrinks their own organisation. What comes back is a plan for a small, dependent, low-stakes annexe, delivered in good faith by people who are not consciously sabotaging anything. If the second-site strategy is owned by the person who runs the first site, you already know how it ends.

The programmes that survive do the opposite: they move a complete product area, a complete process tower, or a complete customer segment. Something with a boundary, an owner, and a metric of its own. It costs more to set up because you have to move senior people and real decision rights, and that's precisely why it works.

A European logistics firm I know of moved its entire warehouse-management product line to a 40-person team in Lahore rather than splitting it across Pune and Lahore. Slower to stand up, painful for six months, and two years later that team ships faster than the Pune group it was carved out from. The variable was ownership, not geography.

The cost of the second site, honestly

Every business case for a second location understates the setup and overstates year-one productivity. Here's the correction most programmes need.

Assume 12 to 18 months before the new site reaches the output-per-head of the established one. Not 6. The people are equally capable; the context is missing, and context takes time to transfer regardless of how good your documentation is. Any model showing break-even in year one has been written by someone who wants the approval more than they want the forecast to be right.

Assume a one-off setup cost of several thousand US dollars per seat for a captive, and more once you count everything properly: entity formation, fit-out, equipment, recruitment, and the legal and tax work. A build-operate-transfer arrangement moves that cost into a monthly rate and adds a margin, which is usually the right trade under 100 seats.

Assume double-running. For the first year you'll be paying for the new team and still paying for the capacity at site one, because nobody is going to let you cut the incumbent before the new group has proved itself. That overlap is the single largest line in the real cost of diversification and it's the one most often left out.

Put those three together and a 50-seat second site will cost well into six figures more in year one than doing nothing, and it can pass a million dollars once the double-running is counted honestly. It starts paying back in year two and the return is decent from year three. If your CFO needs the money back inside 12 months, this is the wrong project and you should say so early rather than building a model that pretends otherwise.

Governance, compliance, and the boring stuff that sinks it

Four things to settle before anyone signs a lease.

• Data flows. Map which categories of data will cross which borders, and get a legal view per jurisdiction before you build. India's own DPDP Rules, notified in November 2025, phase in over 18 months, so even the Indian side of that map is still moving. Retrofitting data residency is enormously expensive.
• Intercompany transfer pricing. A second offshore entity is a transfer-pricing exercise as much as an operational one. Involve tax in month one, not month nine.
• Who the site reports to. If site two reports into the site-one leader, it will be resourced like an overflow annexe, because that is what it will be. Give it a line to the same level as site one.
• Travel and executive presence. Budget for senior people to physically go, repeatedly, in year one. Every successful second site I've seen had an executive sponsor with a stamped passport.

How much diversification is enough?

There's no formula, but there is a useful heuristic: a second site should be large enough that losing the first one degrades you rather than stops you, and that threshold is usually somewhere between 15 and 25 percent of total capacity. Below 15 percent, you have a pilot with a nice story attached. Above 40 percent, you're not diversifying, you're relocating, and the change-management burden is a different animal entirely.

Start with one workstream, one location, one owner, and a three-year horizon. Resist the urge to run parallel pilots in Lahore, Dhaka, and Colombo to "compare" them. You'll get three underfunded teams and no signal, and you'll conclude the whole idea doesn't work when what didn't work was doing it three times badly.

The companies that will be glad they started this in 2026 are the ones who treat it as a capability build rather than a cost exercise. The cost case erodes; every offshore cost case always has. What doesn't erode is knowing how to stand up a competent engineering organisation in an unfamiliar country, which is a skill, and a rarer one than the consulting decks suggest.