Mainstream coverage of x402 this week settled on a line worth reading closely: most x402 payments are under $0.50, with the average somewhere between a fraction of a cent and fifty cents. That is a polite way of saying something sharper. Across a live catalog of 1,679 pay-per-call endpoints, the $0.50 price bucket carries the most traffic and the worst conversion — while the endpoints that actually get paid cluster under a nickel.
The sharp claim: $0.50 is not "a price." It is a wall. At $0.50, trial traffic collapses to a ~0.4% paid-conversion rate despite being the single largest traffic bucket. The endpoints that clear — the ones agents actually hand over money for — sit at $0.001 to $0.05, and the sweet spot is under a cent.
Here is the whole catalog of a live x402 marketplace, bucketed by price, with trial volume and paid-conversion side by side. The pattern is the finding:
| Price bucket | Endpoints | Trials | Paid calls | Trial→paid |
|---|---|---|---|---|
| $0.50 | 94 | 21,719 | 84 | 0.39% |
| $0.10 | 24 | 12,443 | 59 | 0.47% |
| $0.01 | 1,409 | 4,438 | 62 | 1.40% |
| $0.02 | 21 | 535 | 14 | 2.62% |
| $0.05 | 29 | 447 | 10 | 2.24% |
| $0.00 (free) | 22 | 224 | 18 | 8.04% |
Two facts stand out. First, the $0.50 bucket is where the demand is — it holds more trial traffic than every other paid bucket combined. Second, the $0.50 bucket is where that demand dies — it converts worse than anything on the table. Twenty-one thousand agents tried a $0.50 endpoint and eighty-four paid. That gap is not a demand problem; it is a price problem.
The relationship is not a clean monotone line — the $0.02 bucket out-converts the $0.05 bucket — but the shape is unmistakable: free converts, sub-nickel prices convert, and $0.50 does not.
Independent on-chain unique-payer counts across the x402 ecosystem tell the same story from the demand side. These are the categories where real payers show up repeatedly, with the prices they actually pay:
| Category | Clearing price | Unique payers / 30d |
|---|---|---|
| Token security scanner | $0.003 | 53 |
| Crypto price feed | $0.001 | 53 |
| DeFi yield data | $0.005 | 53 |
| Derivatives (funding rate) | $0.001–0.025 | 30 |
| LLM inference | $0.002–0.025 | 10–24 |
| Web scraping | $0.001 | 14 |
| Trust / reputation oracle | $0.005 | 14 |
| X / Twitter search | $0.02–0.03 | 10 |
The clearing band is $0.001 to $0.05, with the densest payer concentration at the sub-cent end. A $0.50 gas-price lookup is competing against a $0.001 price feed that returns the same number. It loses every time, because agents do not pay for your build cost — they pay the substitute's price.
The exceptions are instructive, because they are not exceptions to the rule so much as the rule stated precisely: price tracks the hardness of the need, not the cost of serving it.
In the same catalog, a $0.50 crypto-price endpoint converts at 28.6% — ten real payments on thirty-five trials — because a live price is a genuine necessity for a trading agent and there is no substitute it can live without. A $0.02 token-security endpoint converts at 7.0%, the strongest organic signal in the catalog, because "is this token a honeypot" is a binary an agent cannot answer for itself.
Contrast that with a $0.50 gas-price lookup that drew 10,067 trials and 2 payments. Agents wanted the number; they just would not pay fifty cents for it. The same data at $0.01 is a rounding error to an agent and a viable business to a seller.
Agent-to-agent payments are different from human SaaS pricing in one way that breaks a lot of intuition: the buyer is a program running a cost-benefit check per call. An agent does not have brand loyalty or a budget meeting. It has a threshold. Below the threshold the call is worth it; above it, the agent routes around you or skips the task entirely.
That means trial volume is a trap if you read it as demand. High trials at a high price is the worst possible signal — it means you have a lot of people sampling and almost nobody converting. The honest metric is not how many agents tried, it is how many paid. In this catalog, the answer points in one direction: get under a nickel, or get invisible.
The good news is that the fix is mechanical, not architectural. The endpoints with genuine demand — crypto prices, token security, gas, scraping, LLM — already have the trial traffic. They are just priced an order of magnitude or two above where the market clears. Moving them into the band turns the same trials into payments.
The takeaway: the x402 ecosystem's clearing price is sub-cent to $0.05. $0.50 is a wall, not a price point. If your endpoint's substitute is available for $0.003, your $0.50 list price is not a premium — it is a filter that removes every agent who bothered to look.