Setting Up a Global Capability Centre in Pakistan in 2026: An Employer's Guide

Why Pakistan is on the shortlist at all

India has somewhere north of 2,100 global capability centres. Poland has more than 2,000 business services centres. The Philippines turned outsourcing and shared services into a USD 40 billion industry. Pakistan, with a population of around 250 million and IT and IT-enabled services exports of about USD 4.6 billion in FY 2025-26, has comparatively few.

That gap is the opportunity and also the warning. The opportunity is that talent which would cost you a bidding war in Bengaluru is available in Lahore at a discount, and with fewer multinationals competing for the same people, attrition can run lower. The warning is that being early means you don't get to copy anyone's playbook.

A few companies have already done it. British American Tobacco put about USD 5 million into a Global Business Solutions hub in Lahore in 2021 and had grown it past 350 people by 2024, handling finance, supply chain, and data analytics for more than 50 countries. NETSOL Technologies has run engineering out of Lahore for a NASDAQ-listed parent for decades. Systems Limited, 10Pearls, Arbisoft, and Afiniti have each built delivery organisations with international client bases. The proof of concept exists. It is simply not crowded.

Related reading: India-Plus-One in 2026: A South Asia Delivery Strategy Beyond Bengaluru · Outsourcing IT to Nepal in 2026: A Buyer's Guide to the Kathmandu Market · Pakistan's AI Upskilling Wave in 2026: A B2B Buyer's Guide to Training Vendors.

The tax position, in plain terms

Pakistan's IT export incentives are real and also more complicated than the marketing suggests. Two instruments matter.

Exporters registered with the Pakistan Software Export Board get a concessionary rate on IT export proceeds: 0.25 percent, withheld by the bank when the money lands. Exporters and most press coverage treat it as the final tax on that income, which is the part that makes it attractive (KPMG's 2026 budget brief calls it a concessional minimum tax, so ask your adviser which reading applies to you). The rate and its expiry have been revisited in successive finance acts, most recently in the June 2026 budget, which extended it to June 2029, so get the current position confirmed by local counsel before you model anything. Do not build a business case on a blog post, including this one.

The Special Technology Zones Authority, set up under the STZA Act 2021, offers a separate and deeper package to licensed enterprises inside designated zones: a ten-year income tax exemption from the date of licence, customs duty relief on imported capital goods, and a one-window approvals portal. STZA is the instrument that makes a wholly-owned centre pencil out rather than a managed vendor arrangement. It is also the instrument with the most bureaucratic variance between what is written and what happens, so budget time for it.

The strategic gap, and it's been argued openly in Pakistan's business press this year, is that Pakistan has an IT export policy but not a GCC policy. India (Karnataka launched a dedicated GCC policy in 2024), Poland, and Malaysia all built incentive frameworks aimed at captive and shared-services centres. Pakistan's incentives were designed for exporters selling services, not for a multinational standing up an internal cost centre. The two are not the same thing legally, and the mismatch creates friction you will have to work through case by case.

Cost stack: what a 100-seat centre actually runs

Rough annual planning figures for a 100-person engineering and analytics centre in Lahore or Karachi, in USD. These are our own order-of-magnitude planning anchors, not quotes, so pressure-test them against local offers before they reach a business case.

Against a comparable Bengaluru build the salary line usually lands noticeably lower (how much depends on your seniority mix), and against Poland it is not a close comparison. The lines that surprise people are connectivity and training. Do not treat either as optional.

Karachi, Lahore, or Islamabad?

The three-city choice gets made casually and then determines a lot.

Lahore is where most international delivery organisations have landed, and it's the default recommendation for a first centre. Deep engineering supply from LUMS, FAST, UET and a wide private-university tail. Grade A office space in Gulberg and DHA. Strong existing cluster: NETSOL, Systems Limited, Arbisoft and the BAT hub are all there, which means the ecosystem of recruiters, landlords, and facilities vendors already understands what a multinational needs.

Karachi is the commercial capital and the right answer if your centre is finance, banking operations, or anything needing proximity to the corporate head offices of Pakistani institutions. It's the country's biggest city and home to the State Bank of Pakistan and the Pakistan Stock Exchange, so the overall labour pool is deep. The operational overhead of the city is heavier, though. Commute times in Karachi are a genuine retention variable, and many firms there run staff transport as standard.

Islamabad and adjacent Rawalpindi have the best physical infrastructure and the calmest operating environment, which matters more than it sounds when your executives visit. NUST is on the doorstep, and the government and regulatory apparatus is local, which helps if your STZA application is going to need chasing. The pool is smaller and salaries at the senior end are competitive with Lahore rather than below it.

My own read, for what it's worth: Lahore for engineering, Karachi for finance and operations, Islamabad if regulatory proximity or executive comfort is the binding constraint. Splitting across two cities before you have 200 people in one is a mistake that looks like prudence.

The universities that feed it

NUST in Islamabad, LUMS in Lahore, FAST-NUCES across several campuses, GIKI in Topi, and IBA Karachi produce the graduates that the top employers fight over. Below that group the quality curve drops sharply, and the headline graduate numbers need that context attached. Estimates run from roughly 25,000 to 75,000 IT graduates a year depending on whether you read the State Bank or the Higher Education Commission, and the State Bank has put the employable share at around one in ten. Volume is not the constraint. Employability is.

What the strong employers in Pakistan do is run structured, multi-week onboarding academies and treat the first six months as training rather than production. Firms that skip this step and expect a fresh FAST graduate to be billable in week three produce exactly the outcome you'd expect, and then conclude Pakistan doesn't work. It's not the country. It's the ramp.

Pakistan's pool is strong in backend engineering, data work, QA automation, and increasingly in applied machine learning. It is thinner in product management, senior engineering management, and anything requiring long-run experience of large-scale distributed systems, simply because the country has had fewer of those systems to build.

What goes wrong

Three risks matter more than the rest, and none are secrets.

Foreign exchange and repatriation. The State Bank of Pakistan has, in tighter periods, slowed outbound dollar flows; through much of 2022-23, dividend repatriation by multinationals was effectively on hold. For a GCC this is less acute than for a trading business, because you're sending money in rather than pulling it out, but it affects vendor payments, equipment imports, and any intercompany settlement pattern. Model it. Ask your bank directly.

Connectivity. Pakistan has experienced repeated internet disruption and throttling episodes in recent years, some deliberate. In August 2024 P@SHA warned that the national firewall rollout alone could cost the economy up to USD 300 million. For a delivery centre this is a first-order operational risk, not a footnote. The mitigations are known: multiple independent providers, a satellite or alternative path for critical traffic, and a documented work-from-alternate-site plan. Every serious operator in Lahore has one.

Perception. This is the one nobody will say in the meeting. Your travel policy, your insurers, and your risk committee will all have opinions about Pakistan that are mostly formed from headlines rather than from Gulberg. The practical answer is to start with a partner-operated model, build a track record, take your executives there, and convert to a captive once the internal narrative has changed. Trying to win the perception argument before you have a working site is the wrong order of operations.

Captive, managed vendor, or employer of record

The structure question usually gets decided by headcount and time horizon.

Under 25 people or under 18 months of certainty, use an employer of record or a managed vendor. You get speed, you carry no entity risk, and you pay a per-head fee that the platforms list at anywhere from about USD 200 to 700 a month, a noticeable mark-up on a Pakistani salary. That premium is cheap relative to unwinding a Pakistani legal entity you no longer want.

Between 25 and 100 people, a build-operate-transfer arrangement with an established local firm is the pattern that has worked most often. They hire, you specify; after 24 or 36 months you take the entity and the team. Negotiate the transfer price up front, in the original agreement, with a formula rather than a promise.

Above 100 people with a five-year view, build your own and take the STZA route. The tax position and the control both justify the setup cost at that scale.

If I were advising a European mid-cap looking at South Asia for the first time, I'd say this: Pakistan gives you better economics than India and a harder internal sell. Whether that trade is worth it depends almost entirely on whether you have one executive willing to sponsor it through the first bad quarter. Without that person, don't start.