Technology8 min readApril 28, 2026

The new geography of global technology

For thirty years, engineering capability behaved like a geological deposit: concentrated, hard to move, and worth paying a premium to sit near. That is no longer true. The interesting question is not whether talent has dispersed — the numbers settle that — but what a company owes a site on the day it stops being an outpost and starts carrying real product responsibility.

71%: Of engineering headcount outside the founding HQ country
2.6×: Attrition in delivery centres versus capability centres
11 mo: Median time before a new site owns its first product surface

Why the hub formed, and why it is loosening

The hub model was rational. Tacit knowledge — the kind that travels in corridors rather than in documents — transferred faster inside dense clusters. Capital sat near the founders. Senior engineers who had shipped at genuine scale lived within an hour of one another, and hiring the first one made hiring the next one easier. None of this was sentiment. It was a real reduction in the cost of coordination and in the risk of a bad hire. The premium attached to a Bay Area or Stockholm postcode was, for a long time, defensible.

Three forces have eroded it. Distributed work moved the tacit-knowledge argument out of geography and into tooling — imperfectly, but sufficiently. Local capital matured: a founder in Lisbon, Bengaluru or Nairobi no longer has to relocate to raise a serious round. And diaspora engineers who spent a decade inside the hubs have gone home carrying the standards they learned there. The result is not a flat world. It is a world with many more places where a genuinely good engineering organisation can be assembled.

The delivery centre and the capability centre

Most companies get this wrong on the first attempt, and the mistake has a recognisable signature. A site is opened because the cost per engineer is forty percent lower. Work arrives as specification. Nobody on the ground decides what gets built; they decide how quickly the thing decided elsewhere is finished. For two years the arithmetic looks excellent. Then attrition climbs, the strongest engineers leave for firms that will give them ownership, and the cost advantage is consumed by rework and replacement.

A capability centre is a different object. It owns a product surface end to end: the problem, the roadmap, the on-call rota, and the customer conversations that feed all three. Its engineering director argues with headquarters and sometimes wins. Cost there is a consequence, not a purpose. The test is simple and uncomfortable — name one decision the site made last quarter that head office disliked and lived with anyway. If there is none, it is a delivery centre with better branding.

A named product surface the site owns end to end
A hiring bar set locally, not inherited as a quota
At least one decision head office lost this quarter

A site that has never won an argument with headquarters is not a capability centre. It is a delivery centre with a better address.

NORD/ONE, Technology practice

Share of engineering headcount located outside the founding headquarters country, across six years of a typical portfolio company.

Proximity as research, not as a cost line

The cost framing obscures the more valuable asset. An engineer in São Paulo or Jakarta, living inside the payment habits, connectivity constraints and regulatory texture of that market, knows things no survey delivers. Intermittent bandwidth is not a hypothesis to them; it is an ordinary Tuesday. Products built by people who share their users' conditions tend to arrive with the right defaults already chosen, which is worth more than the twelve weeks of research otherwise required to find them.

The pattern repeats: features designed for a supposedly constrained market travel upward into the rest. Offline-first sync, low-data modes, aggressive tolerance for unstable networks — these begin as accommodations and end as advantages everywhere, because the assumption of a perfect connection was never as safe as headquarters believed. To see a distant market only as somewhere to serve, rather than somewhere to learn from, is to forfeit the compounding half of the investment.

The organisational bill

None of this is free, and the cost is not rent. It is decision latency. When a company spans nine time zones, the default is that every meaningful choice waits for a window in which headquarters is awake — and sites learn quickly that waiting to be told is faster than proposing. Decision rights have to be written down and pushed to where the context lives, or the organisation quietly recentralises within eighteen months regardless of what the operating model claims.

Ask who owns the roadmap. Not who contributes to it — who can change it on a Tuesday without booking a call. In most companies the answer is a single postcode, and every subsequent argument about structure is downstream of that fact. The firms handling this well have accepted something genuinely difficult: a distributed engineering organisation requires the founding country to give up work it enjoys and is good at. Dispersal is not a staffing exercise. It is a transfer of authority, and it should be budgeted like one.

NORD/ONE is a fictional company. Figures in these articles are illustrative.

Thinking about a move like this?

We work with leadership teams on exactly these questions.

Let's talk

Continue reading

Markets