As I write this, ARIN is finally down to its last /8′s worth of aggregate IPv4 address space. Depending on:
– your current and anticipated future IPv4 address requirements,
– your current IPv4 address supply, and
– your approach to things like carrier grade NAT and/or IPv6,
this is either a watershed moment… or a huge non-event. Since you’re reading this article, let’s assume you care about IPv4 address space availability. For example, maybe you operate a rapidly expanding Internet business and worry that you won’t be able to get the IPv4 address space you need moving forward, or maybe you have IPv4 address space that you don’t need and which you might be interested in transferring.
If that or a similar situation applies to you, a question that’s probably front and center in your mind is, “So what’s transferable IPv4 address space actually worth?”
It turns out that that’s a surprisingly difficult question to quantitatively answer. Ultimately, as for all things that get bought and sold, the worth of a thing is “what the market will pay and what a seller will take.”
When it comes to market prices for IPv4 address space, everyone always mentions the price that Microsoft paid for some of the former Nortel legacy space (e.g., Microsoft bought 666,624 IPv4 addresses for $11.25 per address for a total of $7.5 million).
That’s the price that people always mention when the value of IPv4 address space gets discussed, even though that price per IP is now rather dated and low, at least now that IPv4 run-out is upon us.
II. Revenue-Driven Value Ceilings
To explain why I think that per-IPv4 address price valuation is low, let’s think about this from the revenue side of things. Ultimately, at least for business users, IPv4 addresses, like any asset, are used to generate revenue. Absent cross-subsidies or other second order considerations, wholesale inputs normally won’t be purchased to be resold at retail for a loss. Revenue considerations thus let us put a cap on asset valuations.
For example, let’s arbitrarily pick an ISP, such as Comcast. Let’s pretend that Comcast needed to buy IPv4 address space for use by its business customers. Last time I looked, Comcast charged business customers $14.95/month for a single usable static IP address (or you could buy 13 for $34.95/month, which implies $34.95/13=$2.69/IP/month if you buy 13).
Multiply $2.69/IP*12 to get $32.26/IP/year, or $14.95/IP*12 to get $179.40/IP/year to get Comcast’s gross annual income per usable IPv4 address, depending on whether we’re talking a baker’s dozen or just one usable IP at a time.
Of course, IPv4 addresses aren’t a one year asset. At the same time, neither are they a twenty year asset (eventually IPv6 deployment will widely occur). To keep the math easy, let’s assume a ten year usable time horizon.
Let’s also assume that money you get later isn’t worth as much as money you’ve got “in hand.” The discount rate we should probably apply currently isn’t very high, probably just a few percent/per year, but worst case, let’s assume something absolutely crazy (like hyperinflation) happens and use a 20% discount rate.
Let’s also assume that Comcast never changes the pricing for IP addresses to business customers over the next decade (an unlikely hypothesis, but again, let’s be conservative).
In that case, an asset earning $32.26 per year for 10 years with a 20% discount rate is worth $135.25/IP today. If the real value was $179.40 per year for 10 years at a 20% discount, the value today would be $752.13.
Now naturally, a company can’t pay full retail price for an asset it’s purchasing to sell at retail — you need room to markup your costs to cover overhead and to make a profit.
A company buying address space also wouldn’t be able to use/monetize all that space in day one, either. So obviously you’d need to adjust any revenue-based transfer ceiling according to expected utilization in order to get a reasonable wholesale price.
There are other factors that should be considered, too:
– Provisioning a single usable static IP address for a customer actually requires commitment of four IP addresses from your inventory (which is one reason why it’s cheaper to buy a baker’s dozen worth of IPv4 addresses from a provider such as Comcast rather than just one … the price of that one usable IPv4 address actually needs to cover the cost of four IPv4 addresses from the provider’s inventory).
– Our earlier analysis doesn’t consider size premiums (or size discounts), either. The operational utility of a large contiguous block of network addresses is generally higher than a pile of network address “crumbs” because netblocks smaller than an IPv4 /19 (and certainly any netblocks smaller than a IPv4 /24) may not be globally routable, and in fact ARIN won’t process a transfer for a block smaller than an IPv4 /24.
– Finally, sale of legacy IPv4 addresses is not a “reputation-neutral” activity. If you are fortunate enough to have a legacy IPv4 /8 (or a smaller pile of one or more /16′s, etc.), you originally got that address space at no charge. Everyone knows that you paid nothing for your IPv4 address space. You’ve got it, and you no longer need it, while they may be desperate for it.
Given that, if you now turn around and do a designated transfer of that IPv4 address space for a substantial windfall, expect resentment of your good fortune, much as people think poorly of scalpers selling bags of ice on a hot summer day for $20/bag after a hurricane. The “optics” just aren’t good in either case.
On the other hand, if you just relinquish unneeded IPv4 address space back to ARIN for the greater good of the Internet community, that’s a very noble and altruistic gesture, thank you, but shareholders may consider that a fiduciarily irresponsible act, since you just gave away an asset that could have been transferred for millions.
Rock, hard place.
III. Transfer Timing
Timing is probably as important as pricing, sort of like selling ANY wasting asset, whether that’s legacy IPv4 address space in a world transitioning to IPv6 or selling our metaphorical ice on a hot day after a hurricane.
At some point in the future, IPv4 address space may trade at a REAL premium over current values. If you get nervous and “sell too soon,” you may be sad when the future comes around: You’ll have missed potential profits.
Eventually, though, IPv4 address space will be have no value, e.g., once the world shifts to IPv6. It’s a wasting asset with a finite shelf life. If you get greedy and wait too long (or are just indecisive and have a hard time “pulling the trigger” when the time comes) you may find yourself hording a huge cache of buggy whips when there are no buggies with horses that need whipping any more. Eventually the world will shift from IPv4 to IPv6.
I’d also note that ARIN region address space policies aren’t set in stone and are always subject to change (either for the better, or for the worse, depending on your P.O.V.). Waiting exposes you to those additional risks, too.
IV. It’s Time for IPv4 “Market Makers”
Can you imagine a stock market or commodity market where the market price of a share of a major stock or common commodity was grossly uncertain, or where it was hard for sellers to find buyers, or buyers, sellers? No. That would be unthinkable these days. Rational American financial markets ensure order, fairness, and liquidity.
Unlike the American financial markets, the IPv4 designated transfer market is currently crazy. The fact that we don’t even have broadly known and agreed upon market pricing is a symptom of that reality, as is low market liquidity: If you’re willing to pay market prices, you should be able to get IPv4 address space you really need, and if you want to sell, you should have no trouble finding buyers — at least if you’re transacting “at the market price.”
How could we fix this? Answer: At least for now, just like the financial markets, we need licensed and regulated market makers, IPv4 address “broker/dealers” who are willing and able to broker transfers of IPv4 address space while buying and selling from their own address inventories if there’s a temporary imbalance in buyers and sellers. Unfortunately, that role is currently not allowed.
In an orderly and rational market, buyers and sellers would automatically ensure that the price of IPv4 addresses reflected the totality of information known about those assets, and participants would buy IPv4 addresses if the market price was too low and sell IPv4 addresses if the prices were too high. We could even envision short sales and the purchase and sale of IPv4 options/futures. Adam Smith’s “invisible hand” would be allowed to work.
Currently that’s not possible, and as a result, buyers and sellers both experience friction when attempting to do a designated transfer of no-longer-needed IPv4 address space, and price discovery is imperfect at best, resulting in price-gouging, non-liquidity, hoarding and other symptoms of a less-than-orderly market. Isn’t it time for that to change?
One useful step would be for current holders of potentially transferable IPv4 address holders to begin accounting for their address space assets the same way they do any other asset. After all, if you control an IPv4 /8, those 16,777,216 addresses are worth something, right? If it’s Microsoft’s $11.25 per IP, that would be 16,777,216*$11.25=$188.7 million dollars. If it’s a revenue driven estimate, such as $135.25 per IP, that would be over $2.26 billion. Shouldn’t assets of that magnitude be reflected in company’s balance sheets?
When companies begin paying explicit attention to the value of the assets they control, and efficient, rational markets are allowed to operate, much of the current uncertainty about the true value of IPv4 addresses will rapidly be eliminated.
All opinions expressed in this article are solely those of the author and do not necessarily represent the opinion of any other entity.
 https://www.arin.net/resources/request/ipv4_countdown.html (retrieved April 27th, 2014)
 Because of the way ethernet works, you need both a broadcast address and an upstream gateway address in order to have at least one usable static address to assign to a host. Because subnets are binary masks, to get at least the three values you need, you need to assign an IPv4 /30, 2^2=4 IPv4 addresses.
 A prefix announced in the global routing table ends up being carried in the routing table of each default-free router on the global Internet. As a pragmatic matter, most ISPs aren’t willing to dedicate scarce routing table capacity to routes that service only a comparative handful of addresses. An IPv4 /24 (256 addresses) is about as fine grained as most ISPs are willing to go, and some won’t accept routes for anything finer-grained than an IPv4 /19 (8,192 addresses).