IPv6

IPv6 (Internet Protocol version 6) is the successor to IPv4, designed to end the address shortage rather than work around it. It uses 128-bit addresses written as eight groups of hexadecimal digits — 2001:0db8:85a3:0000:0000:8a2e:0370:7334 — and has been a finished standard since 1998. Adoption is still incomplete, and that is the single most important fact about it in practice.

Reading an address

The notation looks harder than it is, because two shorthand rules do most of the work.

• Leading zeros drop — 0db8 becomes db8.
• One run of zero groups collapses to :: — so the address above shortens to 2001:db8:85a3::8a2e:370:7334. Only one :: is allowed per address, otherwise it would be ambiguous.

The first 64 bits are the network prefix; the last 64 identify the device on that network. ::1 is loopback, the IPv6 equivalent of 127.0.0.1.

Why "340 undecillion" is a misleading number

The theoretical address space is about 340 undecillion, and quoting that figure hides how allocation really works. Addresses are not handed out individually. A single network receives a /64 — around 18 quintillion addresses — because the protocol assumes it, and a customer typically receives a /56 or /48 containing many of those.

That has a direct consequence for anyone buying IPv6 proxies: sites treat an entire /64 as one entity, because that is what it represents. Being sold "millions of IPv6 addresses" means very little when they all sit inside a prefix that one firewall rule can remove.

Dual-stack and the long transition

IPv6 is not backwards compatible with IPv4. They are separate protocols that cannot speak to each other directly, so most of the internet runs dual-stack — both at once — with clients preferring IPv6 when a DNS AAAA record exists and falling back to IPv4 otherwise. That fallback is why a broken IPv6 path usually presents as a site that is mysteriously slow rather than one that fails outright.

IPv6 proxies: cheap, plentiful, and narrow

Because supply is effectively free, IPv6 proxies cost a fraction of IPv4 and are sold in enormous quantities. The constraints are real, and none of them are about price.

• The target must support IPv6. If a site publishes no AAAA record, an IPv6 proxy cannot reach it at all. A large share of the commercial web still falls into that category.
• Prefix blocking. Anti-bot systems ban by /64 or wider, so a pool's real size is measured in prefixes, not addresses.
• Weaker reputation data. Geolocation and reputation databases are far less complete for IPv6, and some bot detection treats an unfamiliar IPv6 range as suspicious by default.

They work well where the target is IPv6-ready and the job is volume: large IPv6-enabled sites, internal testing, and infrastructure you control. They are a poor choice for defended consumer platforms.

The privacy trade

Without NAT, every device holds its own public address, removing a layer of accidental cover that IPv4 households enjoyed. Early autoconfiguration made this worse by deriving the final 64 bits from the network card's MAC address, producing an identifier that followed a laptop between networks. Privacy extensions now rotate that portion periodically and are enabled by default on major operating systems — though the network prefix still identifies your connection, so this narrows tracking rather than preventing it.

It is also worth confirming your VPN handles IPv6. A tunnel that only carries IPv4 can leak your real address the moment a site offers an AAAA record, which is the IPv6 cousin of a DNS leak and just as quiet.