0x1361dec71361…1361deca
AI Agents Now Drive 76% of OpenRouter Token Use, Solana Foundation Says
AI agents drove 76% of OpenRouter token use in October, and x402 payments hit $100 million in volume — Solana is building the rail, says its foundation AI lead.

Outputs
Agents accounted for 76% of OpenRouter token processing in October 2026, up from 8% a year earlier
The x402 standard has processed over 400 million payments and $100 million in volume, from zero a year ago
More than 120 teams are building AI products on Solana, per foundation AI lead Rishin Sharma
Token prices fell 270-fold from GPT-4 ($37.50 per million) to GPT-5 nano ($0.14), while monthly token use rose about 330-fold
Solana's BlockRun, the largest x402 server, handled 60+ million transactions in two months
AI agents now account for 76% of all tokens processed through the model marketplace OpenRouter, up from just 8% a year earlier, according to Rishin Sharma, who leads AI at the Solana Foundation. Sharma presented the figures at the Breakpoint conference in London, where he argued that autonomous software, not humans, is becoming the primary user class of the internet — and that blockchains are positioned to become its payment rail.
"The biggest opportunity in crypto is already underway," Sharma said on stage.
What has Solana actually shipped?
The foundation launched Solana.ai last week, a homepage for the network's AI ecosystem. More than 120 teams are building AI products on Solana across payments, trading and infrastructure, Sharma said. Alongside the hub, the foundation has released:
- Payment channels rated at 1 million payments per second
- A private AI service offering attested private inference on Solana
- Pay.sh, a server integrating more than 100 APIs that agents can pay to use
The numbers behind the shift are stark. Sharma's slides showed the price of a million tokens falling from $37.50 for GPT-4 to $0.14 for GPT-5 nano — a 270-fold drop — while monthly token consumption has risen roughly 330-fold. Bots accounted for more than 53% of global web traffic in 2025, up from 51% the prior year, according to Imperva's Bad Bot Report, which Sharma cited.
"Which model is smarter matters a little bit less, and what matters more is what the model can actually do," Sharma said. "What are the tools it's using? What are the data it's able to access? And what money can it functionally use?"
How large is the agent payments market?
Payments are where the change is most visible. The x402 standard, which lets AI agents pay for online services, has processed more than 400 million payments and $100 million in volume, up from zero a year ago, Sharma said. BlockRun, a Solana-based service he described as the largest x402 server, has handled more than 60 million transactions in the past two months alone.
The standard's institutional backing is notable. The x402 Foundation is backed by Coinbase and the Linux Foundation, and the Solana Foundation partnered with Google Cloud on a stablecoin payments service for AI agents in May. Sharma added that the foundation is working with Google, Alibaba and ByteDance to let those companies charge for their AI models through agent payments.
Are agents real economic actors yet?
Macro investor and Real Vision founder Raoul Pal, speaking earlier at the conference, was more skeptical. Agents today are not meaningful economic actors, he argued. "They're like kids," Pal said. "These are young kids who spend nothing on a few sweets and stuff." As stocks and other assets move onchain, however, he expects agents to begin trading and settling on blockchains, with value accruing to "the underlying layer one."
For developers, what does this change in practice? The answer is positioning: if agentic traffic already dominates token consumption and automated software is a majority of web traffic, the differentiating layer shifts from model quality to tooling, data access and money rails.
The foundation's next steps are concrete. Sharma said it plans to publish benchmarks measuring how well AI models perform financial tasks, and to launch an on-ramp letting users convert cash into stablecoins and hand them to their AI assistants. Franklin Templeton, for its part, has described agentic AI as a killer use case for crypto.
via google.com (Original)
More from Nathan Brooks
Show full bio
Market editor covering money, funding and market shifts around crypto and web3 industry at Mempool Brief.
1,512 articles