[Fiveplus]

The collapse in IPv4 transfer prices is what caught my eye here, dropping from a ~$55 peak in 2021 to a mean of $22 in early 2026 (figure 12).

This validates my hypothesis that the run-up in 2020–2022 was an artificial scarcity bubble driven largely by hyperscalers. AWS was right up there stockpiling before they shifted their pricing model. Once AWS introduced the hourly charge for public IPv4 addresses (effectively passing the scarcity cost to the consumer), their acquisition pressure vanished. The text notes Amazon stopped announcing almost 15M addresses in Nov 2025. I think they have moved from aggressive accumulation to inventory management.

We are seeing asset stranding in real-time. The market has realized that between the AWS tax and the efficacy of mobile CGNAT, the desperate thirst for public v4 space was not infinite. I'm curious to hear more takes on this.

[JulianHart]

The CGNAT point is underrated. Carriers have zero incentive to move away from it - thousands of users per public IP, no transition cost.

The interesting downstream effect is on IP reputation systems. Traditional detection assumed 1 IP = 1 user. CGNAT breaks that entirely - platforms can't aggressively filter mobile carrier IPs without blocking legitimate customers by the thousands.

Makes sense the IPv4 price dropped once mobile networks proved you can serve massive user bases with relatively few public addresses.

[patmorgan23]

Expect CG-NAT boxes are expensive, and introduce another point of failure into the network. Most mobile carriers are running IPv6 first networks these days anyway.

Like you said, CG-NAT does have the benefit of making v4 address reputation less reliable, which means it's not as big a deal for the transition to v6.

[pixl97]

>CG-NAT does have the benefit of making v4 address reputation less reliable

heh, less reliable is doing a lot of heavy lifting there. You mean "complete and total trash". We need to get to the point where Cloudflare/AWS/some other big sites just block CG-NAT nodes for a day going this IP address is a risk.

Instead if you're a website, instead of doing an easy block by IP, you're left filtering out AI crawlers, spammers, and lots of other crap hiding behind a single IP with thousands of other users behind it, and ISPs that don't really give a shit about doing anything about it.

We need to push the value of IPv4 to nearly zero and finally move away from that crap.

[wcfields]

Anecdotally on how this affects the day to day user experience: I just deployed T-Mobile 5G Business Internet to a temporary pop-up art space (it's only active for a few months) and I'd say twice daily I get a CAPTCHA challenge on Google search.

[pixl97]

And I hope it gets worse for users behind CG-NAT to the point that websites and ISPs move to IPv6.

[kalleboo]

I would have thought that a 5G connection would have IPv6 and wouldn't need CGNAT to Google properties

[SchemaLoad]

I wonder if all these new tools that punch through CGNAT like tailscale will end up breaking it when they force these NAT boxes to maintain tons of long lived connections.

With the uptake in smart home and internet connected CCTV by consumers, things could dramatically shift.

[anyfoo]

I personally hate CGNAT, but I cannot deny that nowadays, the overwhelmingly vast majority of customers most likely does not care (and much less know) that they are behind CGNAT, so this is valid.

Come to think of it, for my use cases, I would probably be fine to be behind IPv4 NAT as long as I also have an un-NATted IPv6 prefix. But a big part of the question here of course is whether IPv6 adoption is worthwhile...

[zokier]

It is noteworthy that in 2020 AWS had very limited ipv6 support, but these days they have at least some support in the most critical services.

[WorldMaker]

> efficacy of mobile CGNAT

At driving the majority of mobile traffic to IPv6? Otherwise, it seems hard to describe mobile CGNAT as efficacious to me.

[inemesitaffia]

Amazon LEO

Aka Kuiper

>stopped announcing almost 15M addresses in Nov 2025

[dlcarrier]

As someone with a background in electronics who doesn't manage any internet-connected equipment but has multiple embedded devices connected to a WAN, I'm glad that IPv4 still seems to have a bit of life left in it.

When IPv6 was developed, over 30 years ago, connecting everything to the internet seemed like a great idea. I know that IPv6 can be made secure, but I don't have the background or research time to learn how to do so, and the NAT-by-default of IPv4 effectively means that I get the benefit of a default-deny security strategy that makes it impossible to accidentally directly connect anything to the internet.

I'm hoping I can keep using IPv4 until IPv8 or IPv4.5 or whatever comes next is developed with the modern proliferation of cheap insecure IoT in mind.

For some background on why IoT products are so insecure:

Hardware manufacturers don't really comprehend the idea of updates, let alone timely of security patches. Hardware has to work on the day of release, so everything is documented and tested to verify it will work. I have hardware with a TCP/IP stack that was released 20 years, (https://docs.wiznet.io/Product/Chip/Ethernet/W5500) and doesn't have a single errata published, despite widespread use. This is expected for every single component, for even the smallest 1-cent transistor, which has dozens of guaranteed performance characteristics laid out over several pages of documentation (https://en.mot-mos.com/vancheerfile/files/pdf/MOT2302B2.pdf).

When manufacturers venture into a product that runs software, they don't realize that for a given complexity, working through undocumented or, worse yet, incorrectly documented APIs takes more time than the equivalent hardware development and documentation. I've worked on multiple projects where software bugs were fixed with hardware workarounds, because it's faster, cheaper, and easier to develop, test, document, retool, and add a few cents of bill-of-materials cost per product, than to get reliable output from the already-written library that's supposed to provide the functionality.

The hardware TCP/IP stack that I linked to was developed at a time when it was the cheapest way to connect a low-power embedded system to a network. Modern low-power embedded systems have multiple cores running at hundreds to thousands of MIPS making the resources to run a softtware TCP/IP stack trivial, but the product still sells well, because when security is an absolute must, the hardware development and maintenance cost for the functionality is still cheaper than through software, even when there's no marginal cost to run the software.

[radlad]

> the NAT-by-default of IPv4

IPv4 is not NAT-by-default. The reality of the world we live in today is that most home networks have a NAT, because you need multiple devices behind a single IP.

That said, I agree: it's quite unknowable how many services I've turned on on local machines with the expectation that a router firewall sat between me and potential clients.

But that doesn't go away with IPv6 - the NAT does, the router doesn't, and the firewall shouldn't either. For example, the default UniFi firewall rules for IPv6 are: 1. Allow Established/Related Traffic (outbound return traffic), 2. Block Invalid Traffic, 3. Block All Other Traffic

You must explicitly open a firewall rule for inbound IPv6 traffic. NAT is not the firewall.

[cyberax]

> NAT is not the firewall.

NAT _is_ a firewall. And a much safer one than IPv6 firewalls, because NAT will fail safe if misconfigured.

[radlad]

NAT is not a firewall: all it does is rewrite packets, it does not drop them.

[jonathanlydall]

The article actually remarks on this kind of argument.

While you are technically correct about NAT not being a firewall, it is in practice a widely used front-line defense which even if not “perfect”, it has indisputably proven to be quite effective against a lot of malicious activity.

Against highly determined malicious actors you will of course want a proper firewall, but for 99% of people, NAT is enough to keep from being bothered by run of the mill malicious actors.

Kind of like physical home security, a lot of it is very easy to bypass, but it’s good enough for the common threats.

[radlad]

> Against highly determined malicious actors you will of course want a proper firewall, but for 99% of people, NAT is enough to keep from being bothered by run of the mill malicious actors.

Maybe, maybe not, but regardless 99% of people are not protected by a NAT. They are protected by a "proper firewall," which happens to support NAT (and typically, is enabled for IPv4 networks.)

That is to say, while most home routers support NATs, they also ship with a default-deny firewall turned on. Typically, enabling NAT mappings also configures the firewall for users. But they are not the same thing and we need to stop conflating them because it causes a lot of confusion when people think that IPv6 is "open by default" and that IPv4 is "protected by NAT." It's not. They are both protected by your router using the same default-deny firewall.

[tyingq]

You have to squint a little and see they mean that most consumer routers don't map inbound unsolicited packets to anything internal unless the user specifically configured it to. Which is basically a firewall.

[jcalvinowens]

That's not true in my experience, consumer grade routers will often happily route packets with rfc1918 destination addresses from the WAN to the LAN interface all day. The "firewall" is only that nobody can get packets with those destination addresses to the home router's WAN interface through the internet.

[radlad]

This is because most consumer routers have a firewall, which is separate from the NAT. Creating NAT mappings also creates firewall entries.

Otherwise, the router would happily pass the packet along to any IP address it finds in a packet it receives. That's the job of a router, after all.

[dlcarrier]

A NAT will drop all packets, until something upstream opens a port. Dropping packets is the default behavior of a NAT.

[radlad]

Nope, it's the default behavior of a typical firewall. NAT rewrites packets but it never drops packets. An un-rewritten packet may fail to route (i.e. "destination unknown".) But that depends on the destination in the packet.

[simoncion]

> I know that IPv6 can be made secure, but I don't have the background or research time to learn how to do so, and the NAT-by-default of IPv4 effectively means that I get the benefit of a default-deny security strategy that makes it impossible to accidentally directly connect anything to the internet.

To get the "unsolicted traffic is rejected or dropped" behavior of the typical IPv4 NAT, forward inbound traffic that's related to an established connection and drop or reject the rest.

You can also use the exact same NAT techniques you use for IPv4 addresses with IPv6 addresses. The only differences are that instead of you using RFC 1918 Private Internets addresses (10./8 and friends) you use RFC 4193 ULA addresses (fd00::/8), and you need the usual NAT rules on your edge router, except for IPv6, rather than IPv4. Remember that IPv6 is still IP, just with larger addresses.

It's recommended that you generate your ULA subnet rather than selecting one by hand, but absolutely nothing stops you from choosing fd::/64. If you're statically assigning addresses to your LAN hosts, then your router could be -say- fd::1 and you count up from there. Also note that DHCP exists for IPv6 [0] and is used by every non-toy OS out there except for Android.

> I'm hoping I can keep using IPv4 until IPv8 or IPv4.5 or whatever comes next...

IPvnext is not happening in either of our lifetimes. You're either going to have to buy edge gear that's set up with a "reject or drop unsolicited inbound forwarding traffic" firewall, or learn how to set it up yourself. Either path is not hard. Well, I guess there's secret option #3: "Die without doing either.". That's also not hard.

[0] It has been around for nearly twenty-three years.

[dlcarrier]

Yeah, that's the kind of stuff that I know how it works from a network protocol standpoint, but have no clue how to configure on any given system, let alone verify I configured it correctly. I installed DD-WRT on my router, hoping it would be easier to set up. The user interface was much easier to navigate, but the labels of the settings were so sparse that I couldn't tell what anything was referring to, even knowing the terminology for the the lower layers of network protocols. I wouldn't be surprised if I never get around to working on it in my lifetime, as long as I can play around with electronics projects.

Regarding Android OS, I'm not convinced it isn't a toy OS. I feel like they threw in the Linux kernel, but didn't bother including most of the useful features, and pat themselves on the back whenever they add one back. It took almost a decade before they figured out that you could install fonts without reinstalling the operating system. If they ever discover DKMS, we can stop throwing our phones away every few years, and have some actually useful hardware. Then again, it took Apple two years to add copy and paste to a phone, so maybe it's an industry-wide problem. If I could buy a modern Jornada 700 series running Linux or BSD, I'd never need to pick up an Android or iOS device again.

[simoncion]

> DD-WRT

Since you're in the mood for experimentation, you might try OpenWRT. They even have a somewhat-fancy-shmancy configuration GUI called LuCI.

[themafia]

I don't think you even need a stateful firewall. If it's an IoT device that's not meant to provide services to the internet then it seems to me you can just drop all non local subnet originated traffic and get most of the security you would expect with NAT.

[oasisbob]

If you want to drop all non-local subnet originated traffic, you need to keep state. Otherwise, how can you tell which side originated the flow?

Even that is only a partial solution - UPNP hole punching exploits holes in this logic to allow peer-to-peer traffic into a network which otherwise has a default-deny ACL.

[ianburrell]

IPv6 is just as secure as IPv4. NAT usually combines address translation with a stateful firewall. I remember when they were separate things. IPv6 has the stateful firewall, all the same security but without the mess of address translation.

Also, if you have devices connected to WAN, then they are insecure because they are not NATed.

[dlcarrier]

Oops, I meant to say LAN, not WAN.

[huslage]

NAT is not a security measure at all. It just obscures what's behind a firewall, but that is leaky and not reliable from a security perspective. It might make you feel better, but that is not security.

[dlcarrier]

A firewall has nothing to filter, if nothing is routed to it. My IoT devices communicate with a server running in my network. As long as I am behind an IPv4 router, their communications to that server will never make it to the internet, and any communications from the internet have no way of addressing any device on my network. I literally can't add any security to a firewall because there's no communications to handle. Sure, I have personal computers on the same network, which aren't on a separate VLAN because I'm not familiar enough with my router to set that up, so a compromised PC could forward attacks to my IoT devices, but the firewall would be useless at that point.

If I have an IPv6 router, I can miss-configure it in a way where all of my internal communications between IoT devices work as expected, but they also have discoverable addresses on the internet. This would give the firewall something to do, but I'd rather there be no route in the first place.

Also, if I trusted myself to properly configure my router for IPv6, I would put all of my IoT equipment on ULAs, which much like an IPv4 NAT would leave me with nothing to configure in the firewall.

If I were to take your claims at face value, using GUAs with packet filtering is far more reliable and secure than ULAs, and that seems preposterous.

A properly configured firewall for sure adds security, but isolation always wins out.

[pixl97]

Yea, people consider NAT a firewall, but at best it stops direct connections from outside. People use this as a rationale to non secure individual devices on the network. Then the moment a single device on your network is compromised (do you really trust that Chinese IOT device?) every host that doesn't have its own firewall is at risk.

With IPv6 you at least say "Holy crap, anyone could connect to this, I better secure it from outside and inside attacks" which is how actual security works.

[immibis]

For some background why IoT products will stop being insecure: if you sell one in the EU, you're liable for all the damage your botnet causes.

Luckily, common EU home routers have firewalls, even for IPv6. And it's so much easier to punch holes on purpose! Instead of messing with port forwarding and internal and external IP addresses, you can just say "this device is a server, please allow traffic on port 80 and 443, thank you"

[dlcarrier]

I don't see how the logistics for that would work. Even when you know what devices are part of a botnet, which itself is no easy task, each device in a botnet is only doing cents worth of damage, and mostly to the target, but product liability only applies to the owner of the product.

Also, everyone I know that lives in Europe (although most of them not within EU countries) imports their IoT controllers directly from China or the US, because there is very little available from manufacturers in Europe.

[newsoftheday]

When AWS rolled out plans to start charging for IPv4 addresses:

https://aws.amazon.com/blogs/aws/new-aws-public-ipv4-address...

"As you may know, IPv4 addresses are an increasingly scarce resource and the cost to acquire a single public IPv4 address has risen more than 300% over the past 5 years. This change reflects our own costs and is also intended to encourage you to be a bit more frugal with your use of public IPv4 addresses and to think about accelerating your adoption of IPv6 as a modernization and conservation measure."

Their move disgusted me and I moved from AWS to OCI.

[knollimar]

What disgusted you about it? I'm out of the loop

[jdsully]

They hadn't bothered to add ipv6 support to most of their services and the ones that did have it usually were only dual stack - still requiring an ipv4 address.

[huslage]

They didn't require you to have a public IPv4 address. Just an IPv4 address.

[pixl97]

Which requires dual-stack and all the issues that come with it, especially with private addresses.

[knollimar]

That sounds like a failure in every direction. I see why you moved

[newsoftheday]

It was clearly a corporate money grab, not an altruistic motion as they made it sound.