Technology and Power I — Platform as Infrastructure

When platforms become infrastructure, the governance question shifts from market competition to public utility.

The Infrastructure Threshold

Platforms cross the infrastructure threshold when the cost of not using them exceeds the cost of the terms they impose. Before that threshold, users have genuine choice: they can decline a platform's terms and operate on alternatives without significant loss. After it, the choice is nominal — the alternatives have been foreclosed by network effects, by the migration of counterparties onto the dominant platform, or by the integration of the platform's services into operational processes that cannot easily be restructured. The platform has become infrastructure in the economic sense that matters: something you cannot practically function without.

The governance implications of this threshold shift are substantial. Markets can govern optional services through competition — users who find one provider's terms unacceptable can switch to another, and this switching threat disciplines provider behaviour. Markets cannot govern infrastructure through competition alone, because the infrastructure threshold is defined by the absence of credible alternatives. Once a platform has crossed into infrastructure, the competitive discipline that would otherwise govern its terms no longer operates.

The Infrastructure Governance Gap

The infrastructure governance gap is the space between what infrastructure requires — reliable, non-discriminatory access on terms that do not extract monopoly rents from the users who depend on it — and what private platform ownership, motivated by shareholder returns rather than public utility, is structured to provide. Physical infrastructure networks — water, electricity, telecommunications — crossed this threshold in prior centuries and were eventually subjected to regulatory frameworks designed to close the governance gap. Digital platforms have crossed the threshold more recently, more rapidly, and in forms that the existing regulatory frameworks were not designed to address.

The Gap in Practice: Three Documented Harms

The first harm is self-preferential treatment: a platform that also operates as a marketplace favouring its own offerings over those of the third parties who depend on it. In 2017, the Commission imposed a record fine of €2.42 billion on Alphabet and Google. The decision found that Google had systematically demoted the results of competing comparison-shopping products on its search results pages, while granting prominent placement to its own comparison-shopping service. The Court of Justice of the European Union upheld the fine in September 2024, ruling that Google had abused its dominant position through self-preferencing in its Google Shopping service. The seven-year gap between the conduct and the final ruling is itself instructive: ordinary antitrust process moves far slower than the harm it is meant to address.

The second harm is data extraction the platform's dependents cannot prevent because the platform's dual role — operator of the marketplace and competitor within it — gives it privileged visibility into their business. The European Commission's investigation found that Amazon's use of non-public marketplace seller data let it avoid the normal risks of retail competition, drawing on granular, real-time information about rival listings, pricing, and stock to shape its own retail decisions. The Commission ultimately made commitments offered by Amazon legally binding under EU antitrust rules, with Amazon agreeing to stop using non-public seller data — covering sales terms, revenues, shipments, and seller performance — for its own competing retail and private-label businesses. The remedy took the form of a negotiated commitment rather than a finding of guilt, which is itself a marker of how uncertain the legal terrain still is.

The third harm is the arbitrary terms change: a unilateral revision of access conditions that dependents cannot reject because they have no functioning alternative. In February 2023, Twitter announced the rapid shutdown of its long-standing free API tiers with only seven days' notice. Tweetbot and Twitterrific, two third-party clients that had served millions of users for over a decade, were shut down after the platform suspended their API access, along with the disaster-response bots, accessibility tools, and research projects that had been built on the assumption of continued access. None of the affected developers had a vote in the decision; none had a comparable alternative to migrate to. That asymmetry — the platform can rewrite the terms of dependency unilaterally, and the dependents can only absorb the change or exit into a vacuum — is the governance gap made concrete.

The platform that has become infrastructure has acquired a form of power that market competition cannot constrain. The API episode shows what governance "by the platform itself" concretely looks like: a unilateral rule change, announced on short notice and enforced through code rather than through any process the affected parties had a voice in. The question is not whether the platform will be governed — it already governs, through terms it writes and revises at will — but whether that governance is made accountable to the users who depend on it, or left to answer only to the shareholders who own it.