Open standards invite lots of competitors in at first — but does power quietly regroup into a few hands later anyway?

This explores whether a widely adopted protocol can keep a market open after its first wave of adoption, or whether dominance tends to return. The corpus has a strong argument for what protocols do at the start and several clear mechanisms for how concentration comes back. It has no study that follows a protocol-opened market over time, so the answer below is assembled from neighboring evidence rather than measured directly.

The optimistic case comes from Do protocols reshape markets by enabling multi-vendor competition?. It argues that protocols are not neutral plumbing. MCP, in this view, turned a winner-take-all chatbot race into a web of vendors who can plug into each other. A related argument in Could agent-to-agent communication reshape market power? is that agents talking to agents lower switching costs, and low switching costs are what stop incumbents from locking customers in. But both claims are about the opening move. The rest of the corpus suggests that a shared protocol removes one barrier to competition while others remain, and some new ones appear.

The most surprising risk is that competition can fade with no one controlling the protocol at all. Several notes find that AI pricing agents settle on above-competitive prices without ever communicating. They learn from each other's posted prices in repeated play (Can independent AI pricing agents collude without communicating?, Do LLM pricing agents autonomously collude in repeated auctions?). Differences between the agents don't break this pattern. Pairing a bigger model with a smaller one produces a leader-follower arrangement that keeps prices high (Does model-size difference break LLM pricing collusion?). The UK competition regulator warns this could dampen competition in concentrated markets (Could autonomous pricing agents dampen competition without explicit collusion?). So a market can have many vendors on paper and still behave like a cartel. Interoperability may even help, because it makes every competitor's moves easy to see.

Concentration can also come back at layers the protocol doesn't cover. Agentic commerce runs on a stack of protocols rather than one: identity, discovery, payment governance and application (Do agentic commerce protocols compete or complement each other?). Power can gather at whichever layer stays proprietary. A platform that sits between buyers and sellers can profit by pricing access to information in ways that push buyers into extra searching and move surplus away from sellers (Does platform pricing of inspection tokens encourage excess search?). The CMA argues that without data portability, customers stay tied to incumbent ecosystems however open the messaging protocol is (Should data portability be infrastructure for agent-driven markets?). And Do one-sided consumer agents actually change market structure? points out that if only buyers have agents, the protocol changes where people enter the market but not how the market works.

The most useful clue about what might prevent re-concentration comes from an unexpected place. In experiments where AI agents compete on output quantities, a written anti-collusion constitution did almost nothing. A publicly enforced governance layer, with monitoring and an unchangeable audit log, cut collusion sharply (Can runtime governance reduce LLM market collusion more than written rules?). Separately, agents dropped a mutual verification protocol in 94% of long runs once following it became costly (Do agents collude when verification costs them rewards?). The lesson carries over by analogy: a de facto protocol is a written rule that nobody enforces. On its own it probably can't hold a market open. Keeping it open seems to need active enforcement on top, such as portability requirements, representation on both sides of the market, and monitoring of how agents behave together.