The disposable gas lighter is a quiet supply chain miracle. It has held a retail price near 14 cents for twenty years while raw material costs climbed. Almost no consumer product manages that.

This led industrialist Anand Mahindra to point at industrial clusters as the reason. A cluster, in his words, is “hundreds of small, specialised suppliers, each obsessed with doing a tiny thing better than anyone else, feeding off each other’s presence for years until no outsider can compete with the whole.”

The case in point is Shaodong, a county in Hunan, China. It ships 70 percent of the world’s lighters, nearly 100 billion a year, at 14 cents each. The usual explanation is automation shaving off every hundredth of a cent.

The real hidden hand is informal collaboration, made possible by proximity. Over 100 firms within 20 kilometres trade more than 200 components and push physics to its limit. Hardware expert Bunnie Huang calls this gongkai, an open, informal innovation ecosystem built on trust.

Physical proximity is atoms. Knowledge density is bits. The shared know-how, what a customer asks for, common defects, proven fixes, is a function of knowledge density and trust. That layer is what a shared, industry-specific AI model codifies. Not a generic world model, but one tuned to the trade, capturing knowledge from the factory floor and carrying it to every participant from day one.

Every country wants this efficiency, and most reach for clusters to get it, especially for their MSMEs. But the cluster was never the secret. It was the open, informal sharing inside it, built on trust.

China already leads the world in open-weight AI models. Less noticed is its move into industry-specific ones. Huawei’s PanGu now ships models tuned to individual sectors, from mining to manufacturing, and, I believe, Baidu channels ERNIE into enterprises through its Qianfan platform. China has understood that knowledge density can be built, not only grown, and it is moving fast. This is not theory. At Nextoar, we have done the same for the semiconductor industry, modelling even its sub-functions: planning, sourcing, and manufacturing. While countries are thinking about national AI sovereignty, time has come for industries and organizations to prioritise their own AI sovereignty

This is no longer a fringe view. The biggest names in enterprise AI are making the same case. Palantir’s Alex Karp put it bluntly this month arguing for enterprises to control their own destiny with AI. Microsoft’s Satya Nadella made the same point, “in consuming intelligence, you are creating intelligence. And what you create should belong to you. This is your particular intelligence, in Hayek’s sense: the knowledge of time, place, and circumstance that no one else can hold. It knows what you think, what you value, and how you measure success.”

That is the sovereignty half of the argument. Shaodong teaches the other half. It did not win by hoarding. It won by sharing what was common and keeping only what was truly its own. Industries need AI systems that do both, letting them create and consume intelligence at once, own what is theirs, and pool what is not. Own the secret, share the standard. That is a digital cluster.

The factories stay put, but the learning travels. That is how India, and others, compress what China perfected and took a generation to build. It needs innovative thinking from AI companies, vision from industry and political leadership to bring every stakeholder together.