How Will Paywalled Content Get Paid For?
For thirty years the answer to the above question was a subscription. A person paid an annual fee, logged in, and consumed what they could during working hours; the fee reflected roughly what one professional could absorb, because a human analyst reading filings or pulling market data is bounded by time and the length of a working day. Consumption had a natural ceiling, and the pricing team never had to think about it, the working day did the metering. Agents are the first customers to arrive without a ceiling at all.
An agent does not tire and does not stop at the first sufficient answer, it will keep querying a source as long as the expected value of the next query exceeds its cost. The subscription model assumes a human constraint that does not apply to agents, so the pricing model has to change.
Two ways to sell access
The first approach is borrowed from consumer software: a flat monthly fee for unlimited use, (known as the Spotify model). It holds together when the marginal cost of one more play is near zero and total consumption is capped by human time, since no subscriber listens to more than twenty-four hours of music in a day.
Premium data breaks both conditions. Serving a query against a large, fresh, licensed dataset carries real marginal cost, and an agent's consumption answers to no human limit. Flat-fee unlimited access, priced for people and consumed by machines, applies pressure on revenue that eventually forces a repricing.
The second model is similar to your electricity bill: You pay for what you draw, metered to the kilowatt-hour, and the bill scales with how much you use. This is the model the current buildout of AI data centers is making visible to everyone, as the same infrastructure that runs the models turns power itself into the headline cost of the industry. It fits agentic consumption almost exactly, because an agent's usage is already discrete, measurable, and machine-reported. The catch is infrastructure. Utilities took a century to build a meter on every line, a known account behind every meter, and a billing system that settles on a cycle. Data providers have to build the same three things for an arguably harder problem. The consumer is an agent they have no prior relationship with and must authenticate on the spot. The consumption is a single query against specific content, not undifferentiated power. Plus, the settlement has to clear per-query at machine speed rather than once a month. The billing model is right, but the plumbing to implement it is not there yet.
Whichever way it settles, the direction is the same: pricing is moving from seat to query. Flat fees won’t disappear entirely, but most premium access will end up metered because providers can’t keep absorbing agent volume without the revenue to match it.
Three postures, and none of them prices the query
Few companies have built for this, and the responses so far fall into unrecognizable postures. Some providers have sold their archives for training in flat, one-time licensing deals, and the large news settlements are the clearest examples: a lump sum, negotiated after the material had already been used, covering the right to train on a corpus. These deals turn leverage into a check and price a single transfer rather than ongoing consumption, so they say nothing about what happens when an agent queries the content live, every day, at scale.
Others have exposed data to agents through entitlements, extending an existing enterprise license to cover an agent acting for a licensed customer. This is the most forward-looking posture in the market today, but it leaves per-consumption pricing unsolved, because access still rides on a seat or a contract shaped for a human relationship. Others again have chosen not to participate, holding their data behind the walls that have always protected it. That posture is coherent, but it is getting harder to hold as agents route around the walls rather than through them.
What almost no one has built is a way to charge an agent for exactly what it consumes, verify that the agent is authorized to consume it, and settle without a human in the loop. This is a banking problem to be solved, not a data problem. That authentication, metering, and settlement layer does not exist yet, although agents are already consuming premium content anyway.
The consumption you cannot see
When an agent retrieves premium content, the content does not simply get read and forgotten. It can be summarized, recombined, and in some cases absorbed into the behavior of another model. The clearest example is on the model side, where capable open-weight systems have reportedly been trained in part by distilling the outputs of leading closed models, learning to reproduce their behavior without ever touching their weights. The same dynamic reaches data. A provider that exposes premium content without controlling how it propagates gives up more than per-query revenue, because it risks the reproduction of the very thing that makes the data valuable, at a speed and scale human licensing never had to consider.
An anonymous query can’t be held to any terms, but known, authenticated ones can. Getting paid for access is the near-term goal, retaining control over what that access produces is the one that protects the asset, and it is a governance problem as much as a pricing one.
A conversation worth having
The consumption model for premium data is going to be rewritten over the next few years, because the thing consuming the data has changed. The infrastructure behind it has to solve pricing, authentication, and governance altogether, not one at a time. Anchorage Digital took a comparable bet in 2021 when we became home to the first federally chartered bank in the U.S., before the infrastructure for regulated institutions to hold and move digital assets existed, and we built for a market that had not yet arrived.
Our teams are now onboarding Merchants and data providers onto the Anchorage Digital Agentic Banking Marketplace, giving their content a governed, metered, per-query path to the agents that increasingly want to consume it. If you own premium data and you are starting to think about what agentic consumption means for your business, please get in touch.