The idea of an economy where artificial intelligences pay for access to digital services is starting to take shape. Autonomous AI agents capable of purchasing data, APIs or computing resources already exist, driven by blockchain and micropayments. Yet the actual numbers are far from the initial enthusiasm. An analysis relayed by Andreessen Horowitz reveals that activity remains very limited, despite much higher estimates. Between corrected volumes and infrastructure under construction, the emergence of a machine-to-machine economy is still in its infancy.

In brief
- A recent analysis reveals that payments made by artificial intelligence agents are much lower than some initial estimates.
- The corrected data indicates that the actual volume of transactions reached around 1.6 million dollars, far from the 24 million initially mentioned.
- These differences are explained in particular by the difficulty of measuring a sector that is still emerging and by the presence of artificial transactions.
- Despite these modest volumes, the first uses are already appearing, in particular to pay for technical services such as scraping tools, APIs or image generators.
AI agent payouts much lower than expected
The first figures on the economy of artificial intelligence agents attracted attention after the publication of an estimate suggesting $24 million in payments over 30 days. This estimate relayed by Bloomberg was, however, heavily revised after analysis.
Noah Levine, partner at Andreessen Horowitz, explains that the data from Allium Labs shows a very different reality. He emphasizes that “actual AI agent payment volume is closer to $1.6 million”. Research by analysts indicates that gross volumes would reach around $3 million, before correcting for artificial trades.
The adjustments made by the researchers make it possible to better understand the gap between initial estimates and actual market activity :
- The initial estimate relayed: $24 million in monthly payments;
- Data observed on the blockchain: approximately 3 million dollars;
- The corrected volume after removing wash trading: approximately $1.6 million.
These corrections illustrate the difficulty of analyzing a still nascent sector. Transactions carried out by autonomous agents remain rare and the methods to precisely identify them are still under construction. Available data indicates that the AI-automated payments economy already exists, but is still at a very early stage.
The automated payments infrastructure is starting to take shape
Despite these limited volumes, the first concrete uses are already appearing in the technological ecosystem. Payments made by AI agents mainly concern technical services used by developers. According to the data mentioned by Levine, these transactions are used in particular to pay for tools like Firecrawl for web scraping, Browserbase for running browser sessions or even Freepik for generating images. These interactions illustrate an emerging model where automated programs purchase digital assets on demand.
This dynamic also relies on new payment infrastructures adapted to interactions between machines. The x402 standarddeveloped in particular with the support of Coinbase, aims to allow software agents to automatically pay for access to online services. Several major technology companies are already interested in these mechanisms. Companies like Stripe, Cloudflare, Vercel or Google are experimenting with or integrating certain tools allowing autonomous programs to make payments to consume APIs or digital resources.
Still modest today, automated payments between machines provide a glimpse of a market in the making. The infrastructure is being put in place, uses are emerging and the first flows are appearing. If volumes remain limited, the momentum is underway: AI agents are revolutionizing the world of crypto, opening the way to an autonomous economy between software.
Maximize your Tremplin.io experience with our 'Read to Earn' program! For every article you read, earn points and access exclusive rewards. Sign up now and start earning benefits.
