05-26-2026 
|
Getting your Trinity Audio player ready...
|
AI agents—autonomous systems that negotiate, transact, and coordinate without human intermediaries—need two things to function at scale: addresses they can reach and a ledger they can trust. According to IPv6 Forum President Professor Latif Ladid, IPv6 and a scalable blockchain are the only combination that delivers both.
- The two essential components for AI agents to operate effectively
- Accelerating universal IPv6 adoption mechanisms
- Establishing a robust ‘trust and economy layer’
- Governance considerations surrounding agentic AI
He recently authored a paper for the IPv6 Forum titled “Call for IPv6-Only and Blockchain on Agentic AI,” which referred to AI agents as “the next very large-scale killer app for IPv6” with a prediction that the world may have 900 billion such agents by the end of the decade.
Ladid said this “Internet of Agents”—his term for a global network of autonomous AI systems—will be a network where each node needs a unique, stable identifier and the ability to communicate peer-to-peer.
IPv4, limited to 4.3 billion addresses and dependent on Network Address Translation (NAT) technology, simply cannot accommodate that scale (sound familiar?). IPv6, with its effectively unlimited address space, restores the end-to-end connectivity that NAT broke. It also adds programmable routing (SRv6), application-aware networking (APN6), and real-time telemetry (iFIT) that agent systems can exploit.
But the transition from IPv4 to IPv6 has stalled. Dual-stack, or running both protocols simultaneously, was meant to be temporary. But relying on this has made it seem permanent. According to Ladid, the solution is not to ban IPv4 outright, but to make IPv4 increasingly expensive and operationally painful, while making IPv6-only cheaper and lower-risk.
“Don’t forbid IPv4—just make it the luxury, legacy option while IPv6-only becomes the low-cost, low-risk default,” he told CoinGeek. “Enterprises hate forced migrations but love cost optimization.”
Mechanisms to get universal IPv6 adoption moving faster
Ladid outlined three mechanisms that could actually shift enterprise behavior. First, cloud provider pricing leverage—major clouds already charge for public IPv4 addresses. Hence, the next step is to charge significantly more for IPv4 in internal networking while offering steep discounts for IPv6-only subnets with free NAT64/DNS64 gateways. Second, cyberinsurance premium differentiation—a 5-10% discount for verifiable IPv6-only workloads would get CISO attention fast. Third, procurement mandates with sunset clauses— government and large B2B buyers requiring new contracts to demonstrate IPv6-only operation within 24 months.
The evidence that this is already happening arrived this week. Amazon (NASDAQ: AMZN) announced that Amazon Lightsail CDN distributions now support IPv6-only instances as origins, a signal that major infrastructure providers are treating IPv6-only as production-ready, not experimental.
Late last year, China’s IPv6 Development Report (2025), by the Expert Committee for Promoting Large-Scale IPv6 Deployment and Application, noted that China now has over 865 million active IPv6 users (77% of all internet users in the country). That’s 300 times more than the number in 2017.
But what if agentic AI cannot wait for IPv6 to become universal? Ladid said there’s a real risk of stalling if deployment is made contingent on perfect infrastructure, but no one is actually waiting.
“Systems that wait for perfect underlying infrastructure never ship,” he said. “Agentic AI will ship on IPv4 plus workarounds, then incrementally adopt IPv6 where beneficial.”
He explained that real-world agentic AI will work around IPv4 limitations, just as every other peer-to-peer system has: overlay networks like libp2p and WebRTC, cloud-mediated rendezvous for signaling, and content-based addressing that avoids IP addresses as identifiers entirely. The cost is higher latency, some centralization, and added complexity, but not a stall.Need for a ‘trust and economy layer’
This is where scalable blockchain comes into play. Ladid specifically names BSV Blockchain as the trust and economy layer for agentic AI, citing Teranode‘s 1 million-plus transactions per second and sub-10-millisecond latency. But his argument goes beyond raw speed. BSV provides an immutable audit trail of agent decisions, machine-to-machine micropayments at a scale other blockchains cannot match, and (through the BSV MCP Server) economic identity in the form of wallets that agents control directly.
The parallel between BSV and IPv6 is stronger than it might appear at first. Both aim to restore something that middlemen stripped away—direct addressability in IPv6’s case, direct transactionality in BSV’s. Both have faced adoption headwinds because the incumbent infrastructure is “good enough” for most current use cases. And both risk being dismissed as “solutions looking for a problem” until the problem reaches a scale large enough to break the current leader.
Ladid added that BSV’s real stall factors were economic and governance-related, not technical and the same could be said of IPv6. The lesson for agentic AI builders is that waiting for perfect infrastructure is a trap. Ship on what exists, accumulate technical debt, and migrate when the benefits justify the cost.
Agentic AI’s governance question
Another, more contentious, question is governance. If agentic AI is to achieve its full potential, including the “Zero Employee Company” Ladid describes, at what point does human oversight become an issue, rather than a safeguard?
Ladid said this bottleneck is inevitable. Agents operate in milliseconds; humans take minutes to hours. A fully autonomous enterprise could make millions of decisions per day. Human oversight at that scale is impossible, “you’d need a human per thousand decisions, which defeats the ‘zero employee’ premise.”
The shift, he noted, is from real-time per-action oversight to ex-ante constraints and ex-post auditing. Hard rules coded into agent environments (e.g., spending caps, whitelisted actions, no-delete zones) replace human-in-the-loop approval. Outcome audits and circuit breakers can handle the rest.
“If you insist on human approval before every significant agent action, you cap the company’s scale and speed,” he said. “The bottleneck won’t stop AI; it will just push humans into a slower, higher-level governance role. The question is whether that’s still ‘oversight’ or just post-hoc blame assignment.”
The implications for BSV are direct. If agents are to transact autonomously—paying for compute, data, or other agents’ services—they need a blockchain that can handle machine-to-machine micropayments without human sign-off on every transaction. BSV’s unbounded block size and stable protocol are designed for exactly that: a global ledger where an agent can pay another agent a thousandth of a cent, provably and irreversibly, without asking permission.
Whether the “Internet of Agents” arrives on IPv6 or IPv4-plus-workarounds, on BSV or some other chain, is still an open question. What Ladid makes clear is that the infrastructure decisions being made now by cloud providers, insurers, procurement officers, and protocol developers will determine whether autonomous systems can operate at the scale their proponents promise. The agents are coming. The question is whether the network will be ready.
In order for artificial intelligence (AI) to work right within the law and thrive in the face of growing challenges, it needs to integrate an enterprise blockchain system that ensures data input quality and ownership—allowing it to keep data safe while also guaranteeing the immutability of data. Check out CoinGeek’s coverage on this emerging tech to learn more why Enterprise blockchain will be the backbone of AI.
Watch: Can we trust AI? How blockchain and IPv6 could fix accountability
Recommended for you
