By Valentin Ghita, Content and Research, Anonymous Proxies
For most of its life, IPv6 was the protocol everyone planned to adopt next year. That changed in 2026. On 28 March, more than half of the users reaching Google did so over IPv6 for the first time, at 50.10%, as the Internet Society’s Pulse team reported. Websites are further behind: W3Techs finds IPv6 on 32.2% of them. Much of the infrastructure work in 2027 will happen in the gap between those two numbers, and a large share of it will be driven by automated traffic, which already makes up 53% of all web traffic according to the Thales 2026 Bad Bot Report.
1. IPv4 gets cheaper to buy and more expensive to run
IPv4 prices fell hard in 2025. Writing on CircleID, Jack Hazan of IPv4.Global noted that /16 blocks crossed below $20 per address in May 2025 for the first time since 2019 and fell below $13 by the fourth quarter, the lowest prices in about ten years.
Cheaper addresses have not made IPv4 cheaper to operate in the cloud. AWS has charged $0.005 per hour for every public IPv4 address since February 2024, which comes to about $43.80 a year per address. For a crawler fleet, a price monitoring service or an AI agent platform that needs hundreds of outbound addresses, that is a recurring line item.
In 2027, more of these teams will send traffic over IPv6 by default wherever the target supports it, and keep IPv4 for the sites that don’t. AI agents will speed this up, because they are built on new infrastructure that has no reason to inherit IPv4-only habits.
2. Rate limits and reputation move from addresses to prefixes
On IPv4, one address usually means one connection, one household or one server, so blocking an IP is a sensible unit of control. IPv6 breaks that assumption. A single home or server connection typically gets at least a /64, a block of more than 18 quintillion addresses, and software can rotate through them at no cost. Blocking one IPv6 address does almost nothing.
Defenders will respond by applying rate limits, reputation scores and blocks to the /64 or a larger prefix, with the network’s ASN and type (consumer ISP, hosting provider, mobile carrier) layered on top. Security and fraud teams that still log, count and alert on single IPv6 addresses will spend part of 2027 rebuilding those rules.
There is an upside for real users. On IPv4, mobile carriers put thousands of customers behind shared addresses through carrier-grade NAT, so blocking one IPv4 address can lock out a whole neighborhood of legitimate shoppers. On IPv6, each subscriber has a prefix of their own. Prefix-based rules done well will be both harder for bots to dodge and fairer to the people behind them.
3. “Millions of IPv6 IPs” stops being a selling point
Proxy and scraping vendors have started advertising IPv6 pools in the millions. Once targets score traffic by prefix, a million addresses inside a few /64s behave like a handful of IPs. The numbers that will matter in 2027 are how many distinct prefixes a pool spans, which networks announce them, and whether each customer gets a prefix of its own or shares one with strangers whose behavior it cannot control.
Buyers will start asking those questions in the same way they now ask where residential IPs come from, a question that became hard to avoid after Google and the FBI disrupted the NetNut proxy network in July 2026. Providers that can answer clearly will win the larger contracts.
4. Dual-stack becomes the default for data pipelines
None of this ends IPv4. About two-thirds of websites still have no IPv6 address, and many of the sites data teams care about most, from online retailers to public registries, are in that group. ISP-registered IPv4 addresses will keep their value for exactly that reason.
The practical setup for 2027 is dual-stack: tools that try IPv6 first, fall back to IPv4, and record which protocol each request used, so block rates and costs can be compared target by target. The metric worth watching is cost per successful request, split by protocol, because a cheap IPv6 request that gets blocked is more expensive than an IPv4 one that goes through. Teams that understand how IPv4 and IPv6 addressing differ will find this an easy change. Teams that treat IPv6 as a longer IPv4 address will keep getting blocked without knowing why.
5. IPv6 shows up in compliance and audit questions
As more automated traffic moves to IPv6, the questions auditors and security teams already ask about IPv4 will follow it: who owns this prefix, how long was it assigned to us, and can we prove which requests came from it? Logging pipelines that store IPv6 addresses as unparsed text, or truncate them, will make those questions hard to answer. I expect 2027 to be the year many organizations find that out during an incident rather than before one.
Looking ahead
IPv6 crossing 50% at Google was a milestone for human users. In 2027, the same shift will reach the automated half of the internet. Teams that plan for prefixes, dual-stack routing and the real cost of every outbound address will spend far less time fighting blocks they don’t understand.
##
ABOUT THE AUTHOR
Valentin Ghita handles content and research at Anonymous Proxies, a proxy provider operating since 2010. He writes about network infrastructure, web data collection and bot traffic.
