Nobody denied you anything: the fifteen ways to hold power over a technology market¶

A plain-language companion to the working paper "Channels of Power in Production Networks: A Taxonomy for Measuring Technological Dependence" (v0.3, July 2026). The paper carries the definitions, the literature, and the measurement protocols; this text carries the ideas.

The short version. Every study of technological dependence starts from a list, usually unwritten, of the ways one actor can move another. Economists count supply and prices; lawyers count jurisdiction; ownership researchers count shares. Meanwhile the people who actually steer technology markets work channels that appear on nobody's list: the analyst quadrant, the standards committee, the keynote stage, the training curriculum, the maintainer seat, the lobbying meeting. We wrote the complete list: fifteen channels in three families. The third family is the interesting one, because it holds power without ever refusing anybody anything.

The hard family: who can stop or reprice you today¶

Six channels act on whether your technology runs tomorrow morning and what it costs. Someone supplies your components and materials, and can stop. Someone licenses your software and controls its updates, and can terminate or reprice (ask the European firms that lived through the VMware repricing). Someone operates the running services you rent every month. Someone holds your data and your identity systems, which is what makes leaving expensive. Someone owns the app stores, marketplaces, and payment rails between you and your customers, and can delist you. And states reach through all of it with law: disclosure orders on one side, export bans on the other. These six are what dependence studies usually measure, and they are what our companion papers price: who can cut whom, how many independent ways of operating exist, what an exit costs.

The board-changing family: who rewrites the cast list¶

Three channels act on the network itself instead of on today's operations. Ownership can change hands: the acquisition, the buyout, the foreign listing (the capture paper's territory: one structural feature, the governance lock, is associated with staying European in 89 percent of resolved cases against 55). Capital steers who exists and on whose stack they build: venture terms, cloud credits, and the remarkable arrangement in which a chip incumbent takes equity in its own customer. And skills decide which options are real: a stack no local engineer can operate is an option in name only, which is exactly why French qualification rules ask whether a provider can keep running things from its own competences.

The quiet family: who edits your shortlist¶

Now the six channels that never say no to anyone.

An analyst office publishes a quadrant; the quadrant becomes the procurement annex; a supplier absent from it is available and invisible. A standards committee writes the interface everyone must implement, and whoever holds the pen decides what counts as compatible. A keynote and its echo in the trade press set expectations, and in markets with network effects, expectations are self-fulfilling: convincing everyone that one future is inevitable is an investment with returns. A curriculum teaches ten thousand students on one vendor's donated stack; a certification funnel credentials their careers on it; a decade later the market's defaults live inside people. A maintainer seat in an open-source project steers a resource nobody owns. And a lobbying meeting shapes the rules all the other channels run under; this one we have measured: the three big US cloud firms logged 465, 288, and 218 high-level European Commission meetings against 44 for the largest European provider.

Why the quiet family wins: three lists¶

Keep three lists apart. What exists: the feasible suppliers, the list our other papers count. What you consider: what survived the quadrant, the integrator's habits, the procurement template, the defaults. What you buy: where the revenue goes.

Decisions are made on the middle list. But the future of the first list is set by the third, because options live on revenue: a supplier nobody buys from exits, a fork nobody maintains rots, a skill nobody hires for stops being taught. Put those together and the mechanism appears: whoever edits the middle list controls, with a lag, what exists at all. Remove an option from consideration for long enough and you have removed it from the world, without a single refusal, mostly in public, and entirely legally. That is how soft power turns hard, and it is why an audit that counts existing suppliers, even correctly, can certify as diverse a market whose shortlists were written elsewhere years ago.

What to do with the list¶

For Europe's measurement effort, the list is a to-do inventory: the hard family and the lobbying channel are already measured in our datasets; the priority additions are the consideration sets themselves (what European procurement actually shortlists at the layers where we know what exists: the gap between the two lists is a number waiting to be computed), the certification-funnel populations, and the curricula. For policymakers, the lesson is an extension of the one our theory paper proved about certificates: checking suppliers one layer at a time misses the pattern across layers, and checking what exists misses the machinery that decides what gets considered. And for researchers elsewhere: nothing in the list is digital. Pharmaceuticals have formularies where clouds have quadrants; agriculture has seed catalogues; energy has certification bodies. The three families and the three lists travel.

One caveat, as always. The fifteen channels are a synthesis with a completeness claim of the inductive kind: they cover every mechanism documented in our evidence and the literatures we reviewed, and a sixteenth would be a welcome finding. The quiet channels currently carry measurement protocols instead of numbers, and the paper says exactly what to measure before anyone should quote a magnitude.