Solana Foundation outlines AI agent roadmap at Breakpoint 2026
Rishin Sharma laid out plans for AI model benchmarks, inference access, and a card-to-stablecoin onramp built for agents
The Solana Foundation wants its next wave of users to be software. At Breakpoint 2026 in London, Rishin Sharma, the foundation’s head of AI initiatives, framed AI agents as central users of onchain infrastructure.
His roadmap covers financial benchmarks for AI models, model inference, and a card-to-stablecoin onramp so agents can get funded.
What Sharma put on the table
Sharma’s talk, titled “The Next 12 Months of AI on Solana,” took place on November 17, 2026. It came one week after the launch of Solana.ai, a hub that already lists more than 120 teams building AI-based products on the network.
Among those teams, payments have become the leading use case.
Sharma pointed to heavy agent-driven activity on x402, a payment standard built for machine-to-machine transactions. According to the research findings, x402 has processed over 60 million transactions in the past two months alone.
Across its lifetime, the standard has handled more than 400 million payments and $100 million in volume.
On the infrastructure side, Sharma said Solana’s payment channels can handle 1 million payments per second. He also highlighted Pay.sh, which integrates more than 100 APIs that agents can pay to access.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Sharma also cited partnerships with Google, Alibaba, and ByteDance focused on pay-per-request access for AI models. Instead of a monthly subscription, an agent pays for exactly what it uses, one call at a time.
The roadmap: benchmarks and an onramp
Two forward-looking items stood out in the plan. The first is a set of financial benchmarks for AI models.
The second is a card-to-stablecoin onramp aimed at funding AI agents.
Sharma also referenced model inference as part of the plan. Inference is the process of running a trained AI model to produce answers, and it’s the part that costs money every time someone asks a question.
Why agents, and why now
The pitch rests on a striking usage shift. According to the research, agents accounted for 76% of OpenRouter token volume by October 2026. A year earlier, that figure was just 8%.
OpenRouter routes requests across many AI models, so its traffic offers a rough read on who is actually using them.
Sharma’s presentation slotted into a broader Breakpoint theme the conference called a “token supercycle.” The idea ties AI functionality together with stablecoins, tokenized assets, and programmable capital on Solana.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.

