This is a question I have considered for decades now, and it impacted my decision to become an IPv4 broker 15 years ago. Now we have completed more than 600 brokered IPv4 transfers, and I want to address the issue with a marketplace perspective.
I began brokering IPv4 sales in the early 2000’s, before the RIRs ran out and before there were any transfer policies in place. In those days we transferred only legacy IPv4 blocks, and normally through the acquisition of shell companies who owned those blocks.
In late 2010, I was approached to bid for IPv4 addresses being auctioned by Nortel out of bankruptcy. Since there had been no public sales, and thus no public pricing of IPv4 addresses, the price range was really an unknown. To choose an appropriate price, we had to consider the implication of the question above. What would be the time horizon for an IPv4 market? How long will it last before IPv4 values approach zero as the IPv6 transition completes?
If the Internet could continue to function in a runout condition, then the pressure to move to IPv6 would be abated, and in fact, could lead to an aborted transition. This could mean IPv4 for decades more, and in that event, wouldn’t the addresses be worth more than they would be if the transition to IPv6 was imminent?
Here we are more than 7 years past the Miami ceremony celebrating IANA exhaust in 2011, mentioned in other answers. I was at that ceremony, mulling over this question. In the intervening years, I have grown to believe that IPv4 will in fact continue to function effectively despite being in an exhaust condition.
Three big factors: NAT, Security, Market.
NAT was far more efficient and effective than it was perceived by those who denigrated it. We have still not effectively plumbed the depths of oversubscription ratios that can be attained by multiple levels of NAT, but the Internet functions in Nigeria, despite the country having huge user counts but tiny IPv4 allocations, with an effective oversubscription ratio of more than 256:1. The TCP and UDP port needs per user have become predictable, and the static direction of port ranges to specific users allows for the information that Law Enforcement requires to be provided without onerous logging of who used what port when.
Security considerations have driven the Internet away from transparent end-to-end connectivity and towards a client/server architecture, where the server can be responsible for most security, and the client can be relatively dumb. This type of architecture dovetails with NAT, in that it allows for a limited number of statically addressed servers to interact with dynamically addressed devices behind NAT. This is how you can remotely control your office computer from home, despite neither side having a static ip address. Only the server that coordinates the connection needs to be globally addressable, and that server can be kept more secure than the clients. Consider the Internet of Things. Can every “Thing” be kept secure in a changing environment, or wouldn’t it be more likely those “Things” be kept behind a NAT and communicate only outbound to a server tasked with recording data and allowing secure access to it?
Finally the IPv4 transfer market allows addresses to find their highest and most efficient purpose, as all markets intend. And the market works, moving unused and underused IPv4 blocks into the hands of those with an ability to profit from the utilization of those blocks. Market analyses confirm that old legacy blocks make up the largest part of those supplying the transfer market. And the data from pricing indicates a maturing but fairly stable market, with prices rising from around $7 in 2011 to around $18 today. And the rise has been fairly predictable, with several plateaus which last a few months. As the prices rise, we see more addresses becoming available for transfer, which is what we would expect, as sellers become more efficient to free up space for sale. We have seen the same block sold multiple times, as projects spin up and down, and businesses fail and thrive.
As long as it continues to work, the motivation to change away from IPv4 is limited. At some point, given the flexibility of NAT, which has already functionally increased the IPv4 address space by several factors (way more than 10 billion devices and only 4 billion IPv4 addresses) the market may decide that IPv4 will work effectively for the foreseeable future.
If the market decides IPv4 but the technorati decides IPv6, what will happen? My money is on the market, but I don’t discount a future strong-arm approach by large players who may unilaterally turn off IPv4 to drive IPv6 adoption at some point. Then we will all feel the difference, if it is feelable, between a limitless Internet and a bounded one.