IPv4's 32-bit address space (~4.3 billion addresses) ran out, and IPv6's 128 bits make addresses effectively unlimited again.

IPv4 is over 40 years old and was designed for a handful of research computers — not billions of smartphones, IoT devices, and servers.

The core driver — address exhaustion:

• IPv4 = 32 bits = 4,294,967,296 addresses (already fewer than humans on Earth)
• The IANA central pool of free IPv4 blocks was exhausted in 2011; the regional registries followed soon after
• Stop-gaps like NAT (sharing one public IP among many private hosts) kept IPv4 alive but broke the "every device is directly reachable" end-to-end model

IPv6's answer — 128 bits:

• 2¹²⁸ ≈ 340 undecillion (340 × 10³⁶) addresses
• Enough to give every grain of sand on Earth its own address many times over — so NAT becomes unnecessary

Tip: Remember the jump as "32 → 128 bits" — that's 4× the bits, but because it's exponential, it's an astronomically larger space, not 4× larger.

Go deeper:

• IPv4 address exhaustion (Wikipedia) — the 2011 IANA pool depletion, per-RIR dates, and NAT as the stop-gap that drove IPv6.
• IPv6 (Wikipedia) — why the IETF built IPv6 and how 128-bit addressing answers exhaustion.

Bigger address space (no NAT), stateless autoconfiguration, a simpler fixed header, no in-transit fragmentation, built-in IPsec, and QoS support via flow labels.

The six headline improvements:

Tip: In practice the killer feature is simply the address space — most of the others (IPsec, QoS) also exist as bolt-ons in IPv4, but you can't bolt on more addresses.

Go deeper:

• IPv6 (Wikipedia) — the canonical list of IPv6 improvements (simpler header, SLAAC, multicast, no router fragmentation).
• RFC 8200 — IPv6 Specification — the authoritative spec the improvements derive from.