"DDoS, FFXIV Server, and NTT Node" excuses concealing AI Data Centers?

FFXIV disconnections have been happening in NA very frequently now. Seems like every few hours. Now I hear people blame DDoS and or SE Servers. The NTT node in California closest to the FFXIV servers comes up a lot, which I think is closer to the truth. Why has no one mentioned AI and the US' deteriorating public infrastructures? US telecommunication companies are unlikely to increase bandwidth when the market is already saturated: basically most people are already customers in some manner and you only increase bandwidth if there are enough new potential customers. California also has the third highest number of AI data centers in the US and they are going to keep adding more. The internet only has so much bandwidth. Can it really keep up to the speed and bandwidth requirements of AI? Before AI, when too many people, and only people, were on at the same time some people would inevitably loose packets and or get temporarily disconnected. It's a traffic jam. Now my limited understanding of AI is that AI uses the internet and can use a lot of bandwidth at once stressing the limits. Just like AI is taking our GPUs and RAM by making prices higher, similar to electricity bills, is it not also likely that these AI data centers are using up nodes along gamers' internet routes? With 7.4 a lot more players are probably logging into FFXIV at once and combined with poor internet infrastructure and increasing number of AI data centers there is not enough road for everyone anymore. I would strongly urge SE employees to directly inquire from NTT if AI is creating bursts of additional internet usage at their US nodes. Is the "DDoS" problem just an excuse concealing another way in which AI is making our lives worse?

We can ask SE to move towards a better server host and implement DDOS mitigation tools. We can't ask them to help deal with the AI bubble and US's terrible internet infrastructure.

Quote:

---

Originally Posted by Hastatus

Why has no one mentioned AI

---

Because the DDoS have been going on since at least Stormblood release when they moved the NA data centers. It's just usually only been 1-3 attacks every several months, and only occasionally did it actually disconnect you rather than just lag.

Quote:

---

Can it really keep up to the speed and bandwidth requirements of AI?

---

Isn't AI usually just a bit of text? Our connections have reached a point where we can download movies and 100 GB games in no time at all. Needless to say, a little AI text prompt or even a still picture being produced is not taking up much.

Quote:

---

when too many people, and only people, were on at the same time some people would inevitably loose packets and or get temporarily disconnected.

---

We've had automated systems for a long time though. Just programmed visits to the internet and automated API calls. Non-intelligent robots but still robots.

The main concern would be if all possible IP addresses were being used. This concern stems from the fact the human population is bigger than the maximum number of IP addresses. To solve this, IPv6 was create but it's been a big failure due to people not adopting or caring about it.

The primary network load of AI is entirely local to the cluster running/training the model. Client requests/responses from AI services are on the scale of a few KB, maybe a few MB if image are involved. The average person probably uses more bandwidth scrolling twitter than they do using AI, as far as the greater internet is concerned.

If you compare this to the impact of something like a video streaming service, it's actually a tiny drop in the bucket in terms of traffic.