In their February 2026 annual letter, Stripe co-founders Patrick and John Collison addressed the intersection of AI and cryptocurrency, specifically focusing on “agentic commerce”—the ability for AI agents to independently discover, decide, and transact. While optimistic about crypto's long-term role, the report identified critical bottlenecks: current blockchain infrastructure is “not ready” for the demands of autonomous agents.
Engineering Challenges to Level 5 Commerce
The Stripe founders compared the current state of crypto for AI to the mid-1990s internet. They view the limitations as “engineering challenges” that must be solved before the economy reaches fully autonomous agentic commerce. Stripe has made bold moves: acquiring stablecoin platform Bridge and launching the Agentic Commerce Suite to help businesses transition as the underlying blockchain technology matures.
Two primary technical bottlenecks were highlighted: cost predictability and transaction throughput. Stripe estimates blockchains need to support 1 million to 1 billion transactions per second (TPS) to scale. Ben Goertzel, CEO of the Artificial Superintelligence (ASI) Alliance and SingularityNET, called this forecast “entirely plausible.” He noted that standard digital financial transactions already reach millions per second during peak hours, driven by humans through intermediaries. Agentic commerce changes the scale by orders of magnitude: “Instead of a single person initiating an action, we have an entire team of agents operating autonomously… an entire squad generating transactions.”
Overcoming the Blockchain Trilemma
According to Goertzel, achieving Stripe's vision requires overcoming more than raw speed: balancing decentralization, scalability, and security; enabling agents across multiple networks; managing vast data volumes from autonomous squads; and facilitating direct peer-to-peer settlements and “elevated decentralized identity.” He advocates a system of specialized networks—like a highway with dedicated lanes. “By separating traffic, we avoid gridlock. That’s exactly the scalable architecture we need for agentic commerce: a network of shards, where each piece does one thing well and interacts seamlessly with the rest.”
The race for AI supremacy is consolidating power among tech oligarchs investing billions in proprietary infrastructure. Yet a resilient startup ecosystem is mounting a tactical counter-offensive using agility, niche specialization, and open-source collaboration. “We don’t come close to Google or Microsoft in size, but we’re reaching a scale to compete more effectively,” Goertzel said. “Our ‘secret sauce’ is the power of diversity. Being decentralized lets us bring together people, communities, AI algorithms, and datasets from across the globe, unlike monolithic approaches.” He added that many leading researchers now realize building larger LLMs alone won't lead to AGI—a truth SingularityNET embraced from the start with its Hyperon approach.
To demonstrate its cosmopolitan strategy, SingularityNET and the AGI Society have organized the AGI-26 conference to explore diverse interpretations of general intelligence. On minimizing the risk of AI control by a few entities, Goertzel stressed: “We need open, decentralized, democratic methodologies across the entire AI pipeline: deploying AI at scale, fair data supply, teaching AI broad human values, and collective decision-making.” Combining open-source code with decentralized infrastructure and governance, he believes, ensures AI remains transparent, widely accessible, and “highly beneficial for humanity and other sentient beings.”