Amazon finds $1B jackpot in its 100 million+ IPv4 address stockpile

[zohaibahd]

The big picture: In the world of tech, IP addresses are akin to digital real estate. Just like in the physical realm where urban sprawl is an issue, the IPv4 territory is becoming increasingly crowded. There is a finite number of these 32-bit internet IDs available, and demand far exceeds supply as our online presence balloons.

Amazon Web Services (AWS) flipped the switch on its new IPv4 address pricing scheme on February 1 as it had announced months prior. The new policy means customers will pay $0.005 per public IPv4 address per hour, a seemingly negligible amount at first glance. But dig deeper and you'll find a billion-dollar revenue stream emerging for Amazon's cloud division.

The tech giant first teased the pricing change last summer, positioning it as a necessity given skyrocketing demand and administrative costs for IPv4 addresses. After all, the 32-bit protocol is tapped out at around 4.3 billion unique IDs. That may sound like a lot but in an era of proliferating smart devices, we indeed are running out.

And as IDs run out, associated costs have soared. "The cost to acquire a single public IPv4 address has risen more than 300% over the past five years," the company stated, urging users to transition to IPv6 with its vast 128-bit address pool.

But IPv4 remains widespread, and Amazon holds a trove of the sought-after addresses. An analysis by Border0 estimated Amazon controls nearly 132 million public IPv4s. Crunching some numbers, Border0 found out their eye-watering valuation – around $4.6 billion based on today's average IPv4 price tag of $35.

Of course, Amazon cannot simply cash out and offload that internet real estate. However, it can generate recurring revenue by billing active users. Border0 estimates that 30% of those IPs (79 million) are linked to income-generating AWS services. Quick calculations reveal over $1 billion per year in projected revenue from this policy adjustment. Border0 concludes that Amazon could be earning between $400 million to $1 billion annually with these new prices. Not bad at all.

The pricing upheaval underscores profound shifts in internet infrastructure. Managing that finite pool is getting trickier and costlier. Meanwhile, benefits of IPv6 (beyond its galactic address count) include supposed speed boosts and improved security.

For instance, this old blog post from Meta (Facebook back then) claimed IPv6 optimizations accelerated their site by 10-15%. And when properly implemented, the newer protocol also protects against common IPv4 vulnerabilities. Yet as of 2023, IPv4 is still widely used globally, even though its address resources are nearly exhausted.

So while some may decry the price hikes, it's fair to think that Amazon's carrot-and-stick approach may eventually push laggards to IPv6, paving the way for an internet sans address shortages.

Image credit: rawpixel

Permalink to story.

https://www.techspot.com/news/101753-amazon-finds-1b-jackpot-100-million-ipv4-address.html

 

[Theinsanegamer]

IPV6 has existed for decades at this point. I was hearing all about the transition to IPV6 in high school. It's been over a decade since then. Why are gigantic companies with millions of addresses still using IPV4?

 

[mrvco]

Mortals stopped being able to register IPv4 addresses from ARIN more than a decade ago. Go look at who is sitting on the ClassA blocks... from what I recall mostly major telcos and educational institutions.

The lack of commitment, stop-gap carrier NAT'ing implementations for smaller orgs and the overall complexity required to migrate to native IPv6 has left everyone kicking the can down the road.

And here we are, down the road. It's left us in a situation somewhat analogous to wireless spectrum. Although this one was mostly avoidable with some thought towards ease of real-world deployment, overall usability and forward-looking infrastructure investment.

I'm just surprised it has taken this long for the IPv4 monetization to kick off in earnest.

 

[toooooot]

"the company stated, urging users to transition to IPv6 with its vast 128-bit address pool"
I read almost exactly sentence 10 years ago written by some company.
Maybe we should just add ipv4+ rather than using ipv6

 

[trparky]

Maybe we should just add ipv4+ rather than using ipv6

It would require the same amount of new software and configs that implementing IPv6 would need so we might as well go for IPv6.

 

[Theinsanegamer]

It would require the same amount of new software and configs that implementing IPv6 would need so we might as well go for IPv6.

Rename IPv6 to IPv4+ AI enhanced edition.

 

[VariableSpike]

Surprise, surprise, Amazon has realised it can monetise the fact that it controls a ton of ipv4 address space, another sign along with things like CG-NAT (eww) that show we are movibg closer to ipv4 well and truly being too fully and the seitch to ipv6......maybe, to be fair, having worked with some ipv6 stuff, there are enough complexities and annoyances (as its not like ipv6 is frssh off the blocks either) that switching seems like a nightmare

 

[Lounds]

IPv4 should just exist for internal networks. I'm surprised there hasn't been a bigger push from the IT infrastructure sector in general. I think a lot of it is more to do that IPv4 is very easy to understand and follow. Maybe IPv6 would benefit from AI tools to make the whole transition easier?

 

[themastergoose]

My ISP has been promising IPv6 for over 10 years, I'm still waiting. Instead last year, they introduced a £4.99 monthly fee to retain my static IPv4 address instead.

Clear to see where their priorities lay...

 

[anastrophe]

IPV6 has existed for decades at this point. I was hearing all about the transition to IPV6 in high school. It's been over a decade since then. Why are gigantic companies with millions of addresses still using IPV4?

Because there is nobody on it. IPv6 has been one of the most spectacular boondoggles of the internet age, a non-backwards compatible NETWORKING scheme.

The day Google, Apple, Facebook, and thousands of other sites turn off their IPv4 routing, is when IPv6 will have 'succeeded'.

Don't hold your breath.

 

[human7]

IPV6 has existed for decades at this point. I was hearing all about the transition to IPV6 in high school. It's been over a decade since then. Why are gigantic companies with millions of addresses still using IPV4?

As they say, if it ain't broke, don't fix it. Thing is, lots of AWS services - and other software too - don't work in IPv6 only environments. They've made progress, but it's slow. IPv4 is just the default everyone expects. It doesn't help that local (development) environments can run happily within a private pool of IPv4 addresses - so no need to force IPv6 until, well, it's forced.

The migration will happen slowly, but I'd be surprised if IPv6 "overtook" IPv4 anytime soon. I put overtook in quotes just because the address space sizes are so much different - I mean in terms of what is considered "default", not in terms of sheer numbers.

 

[anastrophe]

IPv4 should just exist for internal networks. I'm surprised there hasn't been a bigger push from the IT infrastructure sector in general. I think a lot of it is more to do that IPv4 is very easy to understand and follow. Maybe IPv6 would benefit from AI tools to make the whole transition easier?

IPv4 and IPv6 work fine for internal networks, that isn't the issue. The problem is that there is no way live in an exclusively IPv6 world without shutting down IPv4. IPv6 is not backwards compatible nor interoperable with IPv4. It is a catastrophe that's been quietly and laughably ignored for decades. Until the most-trafficked sites/networks on the internet shut off their IPv4 routing, there will be no transition.

In other words, there will be no transition. It will merely continue to be used being NAT and CGNAT.

 

[anastrophe]

IPv4 and IPv6 work fine for internal networks, that isn't the issue. The problem is that there is no way live in an exclusively IPv6 world without shutting down IPv4. IPv6 is not backwards compatible nor interoperable with IPv4. It is a catastrophe that's been quietly and laughably ignored for decades. Until the most-trafficked sites/networks on the internet shut off their IPv4 routing, there will be no transition.

In other words, there will be no transition. It will merely continue to be used being NAT and CGNAT.

Let me amend that to be clearer: The only way there will be a global switch to IPv6 is when all IPv4 routing is turned off, and several days/weeks/and probably months for tens to hundreds of millions of people to effectively "be without internet" until someone helps them out.

Be ready to help yourself, your parents, and your grandparents when that happens. Keep in mind that in the US, those over 65 will be the largest population cohort...

 

[anastrophe]

As they say, if it ain't broke, don't fix it. Thing is, lots of AWS services - and other software too - don't work in IPv6 only environments. They've made progress, but it's slow. IPv4 is just the default everyone expects. It doesn't help that local (development) environments can run happily within a private pool of IPv4 addresses - so no need to force IPv6 until, well, it's forced.

The migration will happen slowly, but I'd be surprised if IPv6 "overtook" IPv4 anytime soon. I put overtook in quotes just because the address space sizes are so much different - I mean in terms of what is considered "default", not in terms of sheer numbers.

In reality, migration/adoption isn't working, and can't really work, because of the lack of interoperability. The only solution is a global shutdown of all IPv4 routing, as in, a massive advertising/information campaign starting today, advising that on June 30, 2026, at 00:01, the IPv4 network will shut down, and to contact your local ISP for instruction on how to switch after that date.. That's literally the only way the "switch" can occur.

If you don't believe me, have a read of this article advising how you, today, can switch to IPv6. The 7,999,500,000 people on earth who do not have the appropriate familiarity with what he's talking about are going to need a lot of help.

I spent a WEEK without IPv4 to understand IPv6 transition mechanisms

The time has come to talk about something uncomfortable to a lot of you. You’ve been using legacy methods for far too long. It’s time to move to IPv6. But, of course, there’s a lot more to IPv6 than ‘just’ switching everything over. A lot of systems in the world still haven’t adopted it after...

www.apalrd.net

 

[masaaya4]

So the big guys are cashing in on IPv4 addresses, but when some people are earning money, other people are losing money.

big companies are probably going to start looking at the IT department balance sheet when "IPv4 fees" bloat up in the coming years, and put pressure on a mitigation to IPv6.

 

[hwertz]

"Because there is nobody on it. IPv6 has been one of the most spectacular boondoggles of the internet age, a non-backwards compatible NETWORKING scheme."
Oh but there are. Unless your ISP doesn't support it, you're probably communicating with Google, and this site, using IPV6 right now. You can try going to ipv6.google.com to make sure (www.google.com uses ipv6 with an ipv4 fallback; ipv6.google.com is IPV6-only, so if you don't have ipv6 working it won't connect.)

Ridiculously, two local ISPs, Centurylink (DSL) runs IPV6rd (where they have an IPV4 network with some boxes installed in their network to push traffic through over IPV6.. which does work but their IPV4-to-IPV6 bridges are slow enough I turned it off on my connection), and Imon (fiber optics) apparently doesn't support IPV6 at all!

"IPv6 is not backwards compatible nor interoperable with IPv4."

It kind of is. ::ffff:0:0/96 is there for an IPV4 address to be represented in an IPV6 environment (for instance, Google nameserver 8.8.8.8 -- they do have IPV6 name servers of course, but that example is ::ffff:8.8.8.8 . This is probably rather commonly used internally, And there's a prefix set aside for IPV6/IPV4 translation -- 64:ff9b::/96 so in a world where end users only get IPV6 addresses and there's still IPV4 stuff to contact, it would connect to 64:ff9b::8.8.8.8 and would be connected through. Amazon is apparently using this now, so you can fire up an instance with no IPV4 (even an internal IPV4 address) and still contact IPV4 resources.

As much as I don't like Amazon, this does make sense. These addresses have been a limited resource for over a decade. IPV6 is widespread enough that users can have the choice of running IPV6 if they're equipped to use it, or pay $3.60 a month for an IPV4 if they aren't.

 

[anastrophe]

"Because there is nobody on it. IPv6 has been one of the most spectacular boondoggles of the internet age, a non-backwards compatible NETWORKING scheme."
Oh but there are. Unless your ISP doesn't support it, you're probably communicating with Google, and this site, using IPV6 right now. You can try going to ipv6.google.com to make sure (www.google.com uses ipv6 with an ipv4 fallback; ipv6.google.com is IPV6-only, so if you don't have ipv6 working it won't connect.)

I'm aware of that. I'm a semi-retired sysadmin/network admin going back to the early days of the public internet. My cellphone (as are most cell networks) is behind IPv6 CGNAT, and much of my ISP's network is the same. IPv6 is there, but 99% of the world doesn't use it directly. It is only when 99% of the world is using IPv6 directly that IPv6 will have been 'adopted'. Every site on the internet (that people actually use) that is reachable over IPv6 is reachable via the IPv4 network. The opposite is not the case however.

Ridiculously, two local ISPs, Centurylink (DSL) runs IPV6rd (where they have an IPV4 network with some boxes installed in their network to push traffic through over IPV6.. which does work but their IPV4-to-IPV6 bridges are slow enough I turned it off on my connection), and Imon (fiber optics) apparently doesn't support IPV6 at all!

More evidence of the boondogglocity of it all!

"IPv6 is not backwards compatible nor interoperable with IPv4."

It kind of is. ::ffff:0:0/96 is there for an IPV4 address to be represented in an IPV6 environment (for instance, Google nameserver 8.8.8.8 -- they do have IPV6 name servers of course, but that example is ::ffff:8.8.8.8 . This is probably rather commonly used internally, And there's a prefix set aside for IPV6/IPV4 translation -- 64:ff9b::/96 so in a world where end users only get IPV6 addresses and there's still IPV4 stuff to contact, it would connect to 64:ff9b::8.8.8.8 and would be connected through. Amazon is apparently using this now, so you can fire up an instance with no IPV4 (even an internal IPV4 address) and still contact IPV4 resources.

Outbound yes, but inbound? I don't think so; however, because of the nature of the IPv6 boondoggle, I spend as little time thinking about it as possible.
But it's the fact that there are these many oddball mechanisms in place that are necessary to use IPv6 today that just illustrate what an *****ic choice for a standard it was and is.

As much as I don't like Amazon, this does make sense. These addresses have been a limited resource for over a decade. IPV6 is widespread enough that users can have the choice of running IPV6 if they're equipped to use it, or pay $3.60 a month for an IPV4 if they aren't.

I run a not-for-profit mailserver (and webservers, and timeservers, and lions and tigers and bears oh my) for a small circle of family and friends, one of them a small business. My 'customers' exchange email with countless 'small' endpoints around the entire globe (except Antarctica) - endpoints like mine - and many times smaller. If I were to turn off IPv4, those small mailservers wouldn't be able to reach my mailserver, as far as I'm aware. So, I'll be spending another $43 or so a year on one part of my infrastructure that costs me about that much currently. In other words, double the price. In the bigger scheme of things, it's not a lot of money. But doubling my costs? That's crazy. I may move my email server to the Oracle cloud, where the base free servers are massively more powerful than Amazon's, and the additional costs for storage etc are about the same.

Of course, once I complete the migration, I'm certain Oracle will start charging for IPv4 addresses, sigh.

 

[hwertz]

I'm aware of that. I'm a semi-retired sysadmin/network admin going back to the early days of the public internet. My cellphone (as are most cell networks) is behind IPv6 CGNAT, and much of my ISP's network is the same. IPv6 is there, but 99% of the world doesn't use it directly. It is only when 99% of the world is using IPv6 directly that IPv6 will have been 'adopted'. Every site on the internet (that people actually use) that is reachable over IPv6 is reachable via the IPv4 network. The opposite is not the case however.

Well, the thing is, a lot of people are using it and just don't know it. You're right though, you really need an IPV4 address to make sure everyone can reach it.

More evidence of the boondogglocity of it all!

Well, to be fair, CenturyLink is just US West. The last of the Bell companies (who choose under the telecom act of 1996 "option B", to maintain their monopoly within their region while not competing outside their region.) They probably are still running 20 year old DSLAMs and such. Imon also has fiber going back to the McCleodUSA days (like 1999-2000) but you would have hoped they wouldn't have gear that old to prevent IPV6. Don't know if they're lazy or just still have ancient equipment in their network core.

Outbound yes, but inbound? I don't think so; however, because of the nature of the IPv6 boondoggle, I spend as little time thinking about it as possible.
But it's the fact that there are these many oddball mechanisms in place that are necessary to use IPv6 today that just illustrate what an *****ic choice for a standard it was and is.

Side note, I can't belive the forum censors out *****. OK. It's tough though -- IPV4 was out of addresses, and any mechanism that would expand them involves being incompatible. I mean, the other solution is to do carrier grade NAT with forwarding of incoming ports, so then one web server would be on port 80, one on port 81, one on port 82, etc (and I guess alternate ssh ports too). Which some people do at least on their home networks. That'd be kludgey and not too great either compared to just having a public internet address.

I run a not-for-profit mailserver (and webservers, and timeservers, and lions and tigers and bears oh my) for a small circle of family and friends, one of them a small business. My 'customers' exchange email with countless 'small' endpoints around the entire globe (except Antarctica) - endpoints like mine - and many times smaller. If I were to turn off IPv4, those small mailservers wouldn't be able to reach my mailserver, as far as I'm aware. So, I'll be spending another $43 or so a year on one part of my infrastructure that costs me about that much currently. In other words, double the price. In the bigger scheme of things, it's not a lot of money. But doubling my costs? That's crazy. I may move my email server to the Oracle cloud, where the base free servers are massively more powerful than Amazon's, and the additional costs for storage etc are about the same.

Of course, once I complete the migration, I'm certain Oracle will start charging for IPv4 addresses, sigh.

Yeah. Beats Verizon Wireless -- they don't charge a monthly fee, but (already by 2008 and still present day) charge a $500 -- yes, $500 -- one-time setup fee to get a static IP.

I personally like digital ocean myself. Don't know if the price is competitive with Amazon or Oracle but it's pretty low.

 

[anastrophe]

Well, to be fair, CenturyLink is just US West. The last of the Bell companies (who choose under the telecom act of 1996 "option B", to maintain their monopoly within their region while not competing outside their region.) They probably are still running 20 year old DSLAMs and such. Imon also has fiber going back to the McCleodUSA days (like 1999-2000) but you would have hoped they wouldn't have gear that old to prevent IPV6. Don't know if they're lazy or just still have ancient equipment in their network core.

DSLAMs. OMG. That brings up old memories...

Side note, I can't belive the forum censors out *****. OK. It's tough though -- IPV4 was out of addresses, and any mechanism that would expand them involves being incompatible. I mean, the other solution is to do carrier grade NAT with forwarding of incoming ports, so then one web server would be on port 80, one on port 81, one on port 82, etc (and I guess alternate ssh ports too). Which some people do at least on their home networks. That'd be kludgey and not too great either compared to just having a public internet address.

I hadn't even noticed they censored it - oh well. I don't even remember the word I used.

Yeah. Beats Verizon Wireless -- they don't charge a monthly fee, but (already by 2008 and still present day) charge a $500 -- yes, $500 -- one-time setup fee to get a static IP.

I personally like digital ocean myself. Don't know if the price is competitive with Amazon or Oracle but it's pretty low.

They can be much cheaper, but that's by dint of requiring a lot more elbow grease to get where you want to go; easy for me to do in order to save a few bucks, harder for others.

 

[tokyojerry]

It's all about money and politics. The technology to advance and move forward has been in place for years. When a company like Amazon owning so many millions of IPv4 addresses with a recurring income cash cow, there will be no action until the system breaks down somehow someday. When the cash cow leaves the coralled pasture for greener lands, and they can no longer function efficiently nor effectively in their daily routine, watch how quickly they will take action to change. Until then, enjoy status quo.