Welcome And Quick Setup

Tim 0:13

Hello and welcome to another episode of the Cables to Clouds monthly news update. Uh I'm Tim McConnaughy, your host, as usual. Uh Chris is not with us today because he's busy sleeping or something. And of course, uh with me now, uh hopefully as always, is uh our other co-host, uh Catherine McNamara. Say hi, Kat. Hi, Kat. Oh, I'm sorry, Catherine. I should have said Catherine. Anyway. All right. Uh let's get going right on these stories here. Uh we'll jump

Cisco Live News Kickoff

Tim 0:46

right into it. Obviously, the big one, uh, Cisco Live happened in June. Uh both Catherine and I were there. Chris working for Fortnite was not there. Also, he's in Australia, so but uh it was a good time. But there were some good news stories that came out, some good product launches and announcements, so we'll go ahead and cover them. Um, you know, setting aside the obvious, we both work for Cisco. This is not a marketing pitch or a sales pitch, just report in the news. So, two big things to come out of Cisco

Cloud Control Unifies Management

Tim 1:14

Live. Uh, one was uh Cisco Cloud, or sorry, it's now called Cloud Control. Sorry, it's been called internally, they've been calling it Cisco Cloud Control, and then they took off Cisco uh when they launched it because it was obviously a Cisco product. Um, but we've been calling it uh Cisco Cloud Control for so long internally, it's hard to drop it now. So Cloud Control, which is a unified uh management platform, right? So um anybody who's been following the podcast for a long period of time uh knows that I've been railing, even even when I didn't work for Cisco, when I was working for Aviatrix, I was saying a long time ago that if uh if Cisco could just unify the products, just bring together all of the products that they have in the portfolio, uh, they would be, you know, kind of unstoppable as a single vendor because no other vendor has the width, the width and breadth of the uh portfolio. So I was really excited to see that cloud controls being unveiled. Uh people have listened, you know, Cisco has listened. Uh obviously lots of customers have asked for this for a very long time, right? So this is kind of a manager of managers. Uh it's bringing together all of the different management platforms into one place and not just the not just the management platforms, but also things like telemetry and and logging and whatnot. So, you know, really, really excited to uh to see this personally. Um I'll I'll cover the next one next, but I'm curious. Do you have any any thoughts or opinions there, Catherine?

Katherine McNamara 2:41

Uh no, it just uh it looks uh awesome. Uh obviously it's early stages, so I'm curious to see how the rollout will be. Um, but yeah, I mean I've been uh I've been handling like security cloud control, uh, which was basically for the security product side of things for like the last year. And uh that part has been pretty successful. So it's interesting to see like the net operations and and sim and uh and collaboration get rolled into it. And like uh we've been actually internally calling it uh on our side the security side where I work uh security cloud control, and that's what's been rolled out my marketing-wise. So it's interesting to see it now change to you know cloud control in in general. Um and I think it was also called C3 internally too, Cisco Cloud Control, so they were shortening with C3. So that's gonna be hard to unlearn as well.

Tim 3:35

Yeah, but you know, uh certain, you know, some people really hate the acronyms, so we were told not to use the acronyms. Everybody uses acronyms anyway, right? But um, no, I completely agree. Security Cloud Control was also a really cool product, and I I mean it's just really just become cloud control. Uh what was cool about security cloud control, I thought, was the idea of unifying security policy over all the different uh security products. So hopefully they'll continue that and that will just kind of be the the marching orders forward as we move into the other products with uh with cloud control. So yeah, really, really excited to see that. Um, so go ahead.

Katherine McNamara 4:10

I was gonna say, I suspect just like with security cloud control, at first it'll be like here's the uh one place you can go to jump to the different products in inside inside of it. But what will be cool is see how they unify like the management of it. Like, I mean, they say the unified management platform, but like in reality, like I think you still have to click into like different products inside of inside of cloud control and manage it like you know, jumping from inside of cloud control. What I would like to see though eventually, and I I suspect that that's probably on the roadmap, is uh on the network side of things, kind of having like a unified policy engine as well, like what they were doing with security cloud control.

Tim 4:49

Yeah, no, 100%. And I I'm sure that's the case. Um, it's yeah, right now it's more like, hey, here's your one-stop shop to get into all of your different uh management uh planes for the different products. Uh, but I've seen, you know, they're already working on on kind of taking that feeling about security cloud control and kind of moving it forward with the other products. So yeah, I'm really excited to see what what comes out of it.

Katherine McNamara 5:12

Yeah, and I was gonna say, like, I think from the troubleshooting side of things too, like the AI canvas they also talked about uh is supposed to be helping with some of that as well, which will probably be easy to pivot from cloud control, but being able to like actually like grab troubleshooting data from different products just by asking, like, you know, like AI will be really cool too.

Tim 5:34

Yep, absolutely. So, yeah, AI Canvas should be able to grab all the go to each product that's integrated with cloud control and pull the relevant uh data for troubleshooting anyway. Sorry, starting to sound like a sales pitch, but it is something that I've been waiting for for a long time. So uh I am really excited about it. Um okay.

Multi-Cloud Fabric First Look

Tim 5:52

And then uh the second announcement, which I'm also very excited with uh about to finally get to talk about, um, you know, when I came back to Cisco back in September of last year, uh I got involved with this project. I was invited to be part of to just to get involved and help in some way with this project. Um, they Cisco just announced uh they've renamed it Multi-Cloud Fabric, which is a uh network as a service offering that connects is uh connects branches, data centers, cloud uh environments, multi-cloud, right? It's called multi-cloud fabric. So think of it um if you're if you're trying to think about where it kind of fits with the re the other multi-cloud, you know, deployment styles or vendors, it's it's a little closer to like an Alkira than like an Aviatrix or uh or well, yes, prosimo's gone, but uh, but like that. So it's very much a managed connectivity offering where you can onboard you know sites, you can onboard uh you know data centers, uh pops, whatever, but also connect your cloud workloads to the same fabric and do routing policy and uh eventually security enforcement. Now, this just got announced. It's actually, I think they're taking beta signups. So uh if this is something that isn't, again, this is not a sales pitch, but if that sounds interesting, you want to check it out, you can actually go, you know, talk to your Cisco account team, I guess. I don't know, I'm not sure, right? I'm I'm not involved with the sales part of it, but uh if you I know they're taking beta signups and you can check it out um if you're interested. So let's see, which I sorry, go ahead, Catherine.

Katherine McNamara 7:26

Uh I was just gonna say, uh yeah, obviously anyone watching this are gonna feel like, oh, we're Cisco employees, so we have to like tone down the excitement. Apologies, but we're we're not really trying to like position it as a sales pitch, which is why I mentioned that like it'll be cool to see where it goes initially, like you know, with security cloud control, like initially it was just click into each product one by one, and then they created the hybrid mesh firewall engine. So, you know, just like uh everything else, like it'll be cool to see where it goes, but like obviously we're not like throwing up powerpoints of marketing powerpoints and going through this.

Tim 8:01

Yeah, yeah, right, exactly. So, I mean, obviously, like I said at the beginning, right? We're we both work for Cisco, it is what it is. Um, Chris isn't with us, but uh at the same time, Cisco Live did just happen and they had two major announcements, so we thought it was important to cover those things on the news just as part of our normal news coverage. Um, okay, so I I won't say too much more about uh the multi-cloud fabric. I am really excited about it. I there's probably a lot more I could say, but again, um, you know, I'm not here to uh set a sales pitch. It's just I'm excited because I finally get to talk about it. But um anyway, let's go ahead and move on to the second story, which I think is really hilarious and it probably

Mythos 5 Release Backlash

Tim 8:40

is.

Katherine McNamara 8:40

Big ways to the third story. Okay, so I'll I'll cover this one. Um so uh last week, I think it was what Monday, uh, you know, uh Claude uh Anthropic's Claude Fable 5 came out with Mythos 5. It was a heavily neutered version of Mythos, by the way. I just want to say that. Heavily neutered, like they they they put some clue like anti-security um uh blocks on there. Like, so if you said something like, uh create this app for me, okay, cool, builds the app. Uh now check the code for you know that to make sure it's secure, it would block you and and downgrade down to four, uh I think it was Opus uh 4.8. 4-8, yeah. Um and it would just be like, nope, we're gonna go anything security related, even checking the code that it was producing. Like if you explicitly tell it, like, hey, do this, uh, make sure what you just created was secure, it would just downgrade you. And I had a lot of security practitioner friends who are playing around with it, uh security researchers, and they were really, really annoyed by how neutered it was. It's like, we're not even asking you to check for vulnerabilities of X, you know, some third-party code or anything like that. I'm just asking you to check to see if what you made is secure, and uh it would just downgrade it. People got really frustrated. But on the other side of things, like non-security folks I talked to said that the workflows were a lot more focused, clear, things like that, than though it did burn tokens at like a two-time, yeah, like a double rate. But they said that they really liked it. Um, it was out last week, uh, people were playing around with it, security researchers and security folks were not very happy, regular folks who are like developers, network people, uh, you know, other folks uh were very happy with it. Um, you know, as folks uh have heard from us and others, there's a lot of hype that was created around mythos. Oh, it's too dangerous to release in its unneutered form. It's it's going to you know rock cybersecurity, it's finds bugs in everything. Um and so, you know, previous story we had covered was you know, Project Glasswing was formed, which was a uh which was a um special collaboration between some large companies like Cisco, Microsoft, Cloudflare, like really big companies to uh use un you know uh unchained mythos to look for bugs and things in their code. So for like the last month or so, um all these companies have been eating tokens and and and paying to use mythos uh you know for security reasons. And so uh, but it was so hyped up, like, you know, uh the folks at Anthropic were back in the White House after being kicked out for you know refusing to unchain their AI for the Department of War. And uh, you know, there was just a lot of hype built about like how this thing was about finding bugs in uh you know, mythos was finding bugs in like, you know, 20-year-old bugs in like FreeBSD and other places, and it was just so good at finding vulnerabilities. So uh, you know, all that hype for the last month or or two, and you know, and uh, you know, mythos or anthropic getting invited back in the White House after getting kicked out of it, uh uh and all that segues to our

Export Controls And Fallout

Katherine McNamara 12:02

third story. Um, so as I said, it was a heavily neutered version of uh of Mythos that came out that you just couldn't use it for security purposes. Uh the only people who were allowed to use it for mythos for security purposes were these companies that were part of Glasswing. Well, that all changed. Uh Mythos was out for approximately, what, three or four days? And the government decided, nope, we are going to go ahead and put an export control directive uh against Anthropic to disable access for mythos for any country that's not the US or any national within or outside of the US uh from using it. Now, inside the US creates other problems, obviously. Um, Anthropic has a lot of non-US citizens working for it. So now even the people who may have contributed to the creation of mythos can't even touch our own product. A lot of the companies in Glasswing uh that are allowed to use it are um for, you know, might have H1Bs or for nationals or people dual citizenship. Now they can't use it. So all the so basically Anthropic had to pull access not only to like the Glasswing partners, but also anyone who is using uh Fable 5 or Mythos 5, the neutered public version. And that's because uh my understanding is that Amazon basically uh or AWS basically created a report about uh jailbreaking techniques and other things. And because of the hype that was created around uh Mythos, like this is where Anthropic may have shot themselves in the foot a little bit, uh, they created so much hype about how dangerous it was. US government smacked that down and and just basically said, nope, no foreign nationals at all in and outside the country. So uh now it had to get revoked from ever use from everyone in the meantime, even within Anthropic, most people can't use it now. And uh so Anthropics are pretty pissed about it. Like they're they'll say that you know, jailbreaking, the same vulnerability exists in other models and other competitors, but you know, they created their own hype and they made it sound so dangerous that it it uh it certainly is uh yeah, I I I question some of the government's motives in this case. Is it is it just their own, you know, them anthropic's own hype biting them in the butt? Is it revenge for you know yeah, retali retaliation for anthropic saying no to the Department of War? I don't know. But uh one of the things it does remind me of is back like 20, 25 years ago, we used to have export controls on really uh on strong encryption. So three folks that may not remember, like uh the only encryption we could export or use with inside and outside the United States used to be three DES, which was really easily broken. The NSA could do that in like seconds. They they basically said the government, the US government basically said years ago, this is too dangerous for uh you know, he strong encryption is too dangerous. We could steal secrets, people could do bad things with it. So they actually had an export control against it. And um it wasn't until like there was a I think the EFF was involved years ago, decades ago, and several other uh groups that finally had to like, you know, finally got the government to reverse, uh reverse that. But you know, now we're back at, you know, it feels a little reminiscent of that. Now we're back at where the government is creating export controls on technology because it might be dangerous. But meanwhile, China and other countries are gonna develop these similar models, and as long as they don't hype them up too much, what what's you know, what is the US government gonna do about it? We're pretty much stifling competition based on on more hype than anything else, I feel like.

Tim 15:50

Vibes. Uh yeah.

Katherine McNamara 15:52

Or retaliation. We we're not 100% sure. Obviously, we're not in the minds of the government, but you know, it it does feel a little bit like it could be retaliatory. Not hunt like obviously I'm not a politician. I don't I'm not in their heads, but uh it also could be just uh really bad, um uh bad uh uh you know reaction, you know, reaction without anyone really technical you know helping make that decision. I'm not a hundred percent sure, but uh obviously Anthropic's not happy. They just filed for IPO, I think, last week as well, as they released this. And now, you know, their next models basically had to be pull back. It's not you know, not a great uh great uh intro to their IPO. And I also I was gonna mention we didn't put this as a news story, but also OpenAI also um uh filed to have an IPO as well. So like I think that both happened

Does Mythos Match The Hype

Katherine McNamara 16:52

last week.

Tim 16:52

Yeah, yeah. So I mean there's a lot to unpack here, right? First of all, the the recent SD WAN vulnerabilities that were released um were all from Project Glasswing, right? They were like all that there's a ton of them that came out within the last, I think the last month. Not a ton, there was like three, three or four, but they were pretty decently, you know, impactful. They were patched really quickly because they were found via Glasswing, and therefore we, you know, Cisco had time to patch them. But just to point out that like, you know, maybe the unredacted version of Mythos is really is um stronger, certainly, than you know, I don't know if it's I don't know if it meets the hype, but certainly there were vulnerabilities that all the major vendors that were involved with Glasswing were able to find in their products and patch as a result of having access to it. So there at least is some amount of of truth to that. Now the question I would have here is if if we had taken, you know, Opus 4.8 or or the equivalent model and and pointed it at um the same code base and said find vulnerabilities, I would have loved to have seen the delta between those. Just so you really get a a sense of what the difference is, you know, how really if it really because I agree with you. I think I think they definitely shot themselves in the foot, and there was some there's definitely marketing hype and and you know, the timing of the IPO and all of this, right? It's it's it doesn't take a it doesn't take a conspiracy theorist to to to be up on the wall with the push pins and the the the the threads and the pictures and all that to to figure some of the stuff out, right? Um but you know you you do wonder was the was the mythos or of you know w the unredacted version of f of the fable model so incredibly dangerous that we couldn't have it, you know what I mean?

Katherine McNamara 18:37

So So I'm gonna eat my words a little bit from last time we did news. Um so I've actually talked to some folks who have had the chance to play around with the uh the unneutered, uh uh unchained version of Mythos. They said it did actually a pretty good job. Um they actually found a lot of vulnerabilities with it. Um and these are folks in multiple companies that I just water cooler conversation I've had online with them and stuff. Um and they these are security researchers I actually trust. And they did say it actually did a really good job. Um so I'm kind of reversing what I said before. I think the last news story I was like saying, oh, you know, I'm not I I I don't think it's quite at the hype level that they that Anthropic obviously was putting out, but it was better that from what folks I trust are telling me, it was better than what I actually originally clocked it as. It's not quite a nothing burger. It actually does a really good job at finding some vulnerabilities. So um, you know, that being said, uh, does will it replace um, you know, like bug bounty programs and threat security researchers altogether? No, um, not at all. Far from it. Um, but I I it does definitely find a lot of good stuff. And, you know, I I think though that there's a real

Bug Bounties And Real Costs

Katherine McNamara 20:00

uh and this might actually be kind of an unofficial news story as well, but like let me give an example. A lot of the security bug bounty programs, especially from big companies, we've seen, you know, in the past and even recently can be very stingy when it comes to paying for bugs. Um, there was this real big controversy that happened over the last couple weeks with um uh Microsoft. Um, somebody uh a security researcher found another really bad bug uh in like Zero Day, and uh uh and they tried to report it to Microsoft. And this is like the umpteenth millionth time this person had tried to uh I believe the researcher's name is like Nightmare Eclipse. Um and this is like the umpteenth millionth time they've tried to like report bugs to Microsoft. It they get really bad. They Microsoft had a tendency to take reported bugs and say it's not a vulnerability, whatever, and then silently patch it. And security researchers were got really upset. So this this uh I think uh this person um ended up publishing it and just saying, you know, screw you, Microsoft. And Microsoft canceled their GitHub, their GitL, and tried to like get their blog taken down. All sorts of nonsense uh in ensued. And it got really it became a public news story that uh essentially uh Microsoft is really bad about patching or about uh paying bug bounty and then yeah. Yeah. So the reason why I mentioned this is um the average from what I was reading in industry like news is that the average bug bounty cost for running Mythos to find a bug um is $36,000 per bug. So $36,000 for every every single bug. Um and that doesn't matter if it's a big bug, like a like severity, you know, like you know, CV is like a 10 versus a one. Um so you we see a lot of these programs are really stingy. They pay like 500th, a couple thousand at most, like to s for some of these like you know, security researchers who who submit their bug bounties and And and um they're not getting rich off of this. Like they could go and sell these these vulnerabilities on the black market or gray market and make a lot more money, but they're trying to do the ethical thing. And um so do I even if mythos was all that it was hyped up to be, um, and it's not it it's better than what I originally forecasted, but it's not like as bad as what Anthropic said. Uh do I feel like companies are gonna spend 36k for a bug big or small? I'm not 100% sure.

Tim 22:33

Yeah, just for any bug. You don't even know what you're gonna get, right?

Katherine McNamara 22:36

Yeah, and I had this interesting debate in uh Discord uh just recently with someone. I I don't know if you were in that conversation or not, but um, you know, somebody was saying that like uh, you know, uh the companies will pay to not have the bad reputation hit. But what we've seen in recent years is like bugs and vulnerabilities haven't been the major reason people have moved away from vendors. So I'm curious to see if how even if like mythos comes back out unrest like unchained and you know, and with the bar of the the cost being so high to run it to find every single bug, how many larger companies that have the money to do so are are they gonna continue burning through those tokens? So that's another concern. You know, we'll you know, I'm gonna you know footnote on there. But you know, I I I will say it my personal opinion, just me as a person, I don't feel like it should be uh uh uh control export controlled. I don't think it's it it's not quite the hype uh uh anthropic made it to be, but at the same time, um, and I think it's going to stifle competition. Deepseek's gonna shoot ahead eventually. And what what then? What you can't you the US has no control over export controls on Deep Seek or any other platform AI platform that happens outside the United States, so people just go and use that, and it you know, it feels like we're we're putting a uh finger on the scale against one company.

Tim 24:03

Well, I mean, there is I mean there is the the potential, and I don't know if it's true or not, but like I mean the the White House is still pretty friendly with uh with Elon and his his clan. So I mean it could be that could be uh a grok equalizer. It could be attempting to equalize things for grok. I mean, there's so many potential with this administration especially, there's so many potential angles of like why they would do this, but I I do believe it's part is the is the hype, the vibes, if you will. Um I don't know about the the AI, uh the AI um tech race, I don't know what you'd call it, is that's a whole other thing. We could spend an hour on or more on that easy. I don't know what the answer is. Like in theory, uh you know, humanity should protect humanity if we're getting to the point where we're thinking we're gonna have these super dangerous AI. I don't think we're anywhere near that, and I'm not even sure that exists, if I'm being honest. But, you know, yeah, I agree with you, just generally. It's uh you know, China's not gonna the EU now is starting to figure out that, oh, by the way, the US isn't the ally it used to be. We need to be developing our own models and and all of this other stuff too, right? So it's polarizing. And um, I'm not sure what the answer is. I I do think um yeah, I uh I'm not I'm I'm not sure that here's what I know for sure. The money that we're the $36,000 that you're paying per bug is like a fraction of what it actually costs to find those bugs, right? Like, because it's all still being being very heavy as heavily subsidized. If once the cost starts getting passed along, it's so much of the stuff is gonna be way out of reach, I I think.

Katherine McNamara 25:37

Yeah, it it it I mean we could try for better efficiency, and like some models do get more efficient depending on um depending on w how they work. Like like I I think Deep Seek is you know, it's still like behind in certain models, but it it does I believe my understanding is that it was more efficient than some some of the models that we're using heavily here in the US. And um, you know, it'll be interesting to see where it goes, but obviously I feel like um, you know, just putting export controls on AI just because they could be, you know, open to a a jailbreaking attack or uh or like a prompt injection, like every AI m LLM model you have out there, every AI model is going to be vulnerable to that. Like, are you gonna export control everything? Obviously not. Um and what happens if if uh like I mean obviously open AI, uh you know, Grok and everything like that, uh they're gonna probably not try to claim that their models are so dangerous they can't release them because they've seen what happens at Anthropic. But here's the problem with anthropic now. If they say this model is better than Fable 5 or Mythos, they're gonna get slapped with another one.

Tim 26:50

Now you've already set the bar, which is uh a bit of a problem. So yeah, I completely agree. The um all right, so let's let's move on to the next one, which is a good segue, actually.

AWS Water Efficiency Claims

Tim 27:00

So our last story, I mean there was a lot of stories, right? But we try to keep these things fairly tight. Uh the last story is about uh is about AWS kind of, but it's really about it it's about um AI's uh AI data center efficiency, right? So anybody who hasn't had their head under a rock realize you know has has read a lot of stories now about communities uh all across the nation, probably, I mean, I don't know about the rest of the world, but certainly all across this nation, pushing back on having AI data centers built in their backyard, right? There's there's an ongoing fight in Colorado, which is under already under severe drought, and you know, the Colorado River is not what it once was, about uh, you know, I forget who it is, if it's Google or who it is that's building a data center there. And then there's the one in Utah that Kevin O'Leary is famously, you know, trying to get built, and the people in Utah are pushing back against. And a lot of it has to do with just you know, water, uh, how much water these things take. So uh this this uh article just came out, you know, a couple days ago.

Speaker 1 27:58

I was gonna say water and energy as well. Well, because energy is driving up cost of the energy grade.

Tim 28:04

No, that's absolutely right. Yeah, yeah. I mean, just really resources in general, but it's really the big ones are water and energy. The two the two big ones are water and energy. Yeah, absolutely. Water being an actual physical resource that you know has to be consumed. Energy has to be consumed as well, but it's harder to track track that back because, you know, it's thousand miles away in a, you know, they they have to build more power infrastructure, and that burns coal, uses a nuclear, you know, whatever, whatever it's using. But uh anyway, so this particular article is specifically focusing on uh water efficiency, which uh as you pointed out does not do anything for the energy efficiency. Maybe it's even worse, but we'll we'll focus on the water uh for this article. So Amazon is saying basically that they are receiving hey they have achieved 0.12 liters of water per kilowatt hour in its data centers compared to the industry average of 0.84 liters per kilowatt hour, which is a 52% improvement in water efficiency over the last five years. So I'm trying to think about the math there because these these uh these data centers, these AI data centers put out gigawatts, you know, of power. They need, you know, or need sorry, not put out, uh need gigawatts of of power. So um if you have to do the math on that one, so efficiency is good, right? A 52% improvement over five years, that's great. Um, they're still using absolutely insane amounts of water, but it basically, you know, and what Amazon is saying is that, you know, hey, we we're using free air cooling, evaporative cooling, we're operating at higher temperatures, you know, we're just kind of so so we're just kind of eating the the temperature uh rise and and not cooling it to the degree we used to cool it, which you know, industry analysts are saying, okay, well, that's pretty much going to be the normal. Like any new data centers built have to do these things because the uh consumption is so high. So I what I think is interesting about this, you know, this article in general is just the the wider discussion, you know, 0.12 liters per kilowatt hour. Um less, obviously, uh a lot less than almost a liter per kilowatt hour. That's a huge improvement. But is it enough? You know, for where they're building these data centers, is this enough of an improvement to say, oh well, problem solved. Let's let's keep building data centers. Um, I don't know. I know North Carolina just uh pushed back. I think we stopped uh some data. I don't remember who it was, but uh might have been Google. Uh but building a data center, you know, here in the tr uh near the triangle, I believe. But I don't know. What are you what are you seeing out there in in California?

Communities Push Back On Data Centers

Katherine McNamara 30:36

Um well I would say that like they probably care like from what I've seen, people building their own AI infrastructure and stuff like that, they care about energy uh like being energy efficient because that's resources they have to pay for uh as well, which are more expensive than water for most of the time. But yeah, it's it's interesting. Um, you know, like these data centers are getting plopped down all over the country, uh in places I don't I don't I haven't heard of any community that was like welcoming them welcoming them with open arms because it's not like it produces a ton of new jobs unless like data center security personnel is what not exactly like the most prestigious job that you're gonna you're gonna get out of it.

Tim 31:16

It's like ten people, ten people when it's done, you know.

Katherine McNamara 31:20

Yeah, and maybe construct some construction on building it, but even then they're finding more efficient ways to build this with uh you know to pop them up and pop you know, expand them as they need to. I I think it's gonna be something that like bites us, like bite bites us as a society in the butt later on when like some of these these communities get really uh really expensive to live in because water and and energy become you know uh like we've already seen like the energy costs and water costs uh uh drive up in certain communities. Um one thing I did uh did find interesting though is uh it's kind of related to this, is like the gov US government uh is actually making a watch list now of people who are uh like AI data center resistant. Um I mean I guess the majority of the country is gonna be on a watch list because there's a lot of people that are resistant. Um but it it'll be interesting. Like I uh it it's nice that Amazon is working on on efficiency and in general, I think it's trying their attempt at trying to endear the public to them and you know get people to uh be more accepting of data centers because at the end of the day, like these communities could potentially drive drive the data center out before they're built. Um they have some local power. Like I don't know, like I not to get too political here, but when it comes to like um being able to sway things from a political level, the closer you get to the people, the less typically like the more malleable the politicians are. So like if you go to the federal level, good luck trying to get them to like do anything unless you're a multi-billionaire donor. But when you get down to the local level, state, um, city, county, um, those politicians tend to be more amend amendable to uh what their constituents says say. So I I do see I I mean I'm glad that Amazon is do making these moves, but also that they're probably also announcing these things, not just for the other techies out there, but also to to try to get the those communities to be more accepting of data centers and not like um not block them from being able to build them in certain places.

Tim 33:30

Yeah, no, I agree. I mean I've I've been keeping up with this actually. So there's there's uh I was reading so many stories. I I should have collected them. Maybe I will find go find them all and collect them all. So but um there was one where they built a data center, like there first there was a residential neighborhood, and then they build a data center basically next to it, and the people that live in that neighborhood, like they're having trouble sleeping and stuff because of the subsonic uh machinery sounds and stuff like this. So it's like there's like a whole thing. It's like, man, what would you do if you're you know you bought your house, you're living in it, you're paying off the mortgage, and then they plop a data center next to it, and all of a sudden you can't, you know, it's like impossible to live there. There's there's so much, you know, and then and then there's also I agree with you generally that you know the closer you get to people, the more politicians have to pay attention or they're gonna get voted out. Um, but there's there's also stories of you know count city councils being voted out basically because they they approved a data center above above uh public uh outcry, and then the data center gets built anyway. Like there's a lot of stuff going on here. The federal government's talking about national security reasons to like build AI data centers, you know, for the AI arms race that's heating up, like this whole thing.

Katherine McNamara 34:41

I mean the AI AI arms race that they're export controlling and neutering. Well, I mean that's probably just for certain companies, one company.

Tim 34:50

Yeah, who just who just happened to, you know, refuse them before. So anyway, let's let's not go down that rabbit hole. But that that's kind of my point, right? Is that there's there's a lot being shoved down people's throats on this thing, and making it efficient is a good thing, right? This is a good thing. Uh Amazon, I mean, there let's be honest, no, no person who makes things more efficient, generally no company that makes things more efficient is doing it because they want the you know, that's all they need is the goodwill of the people, right? They're also trying to save money and whatnot, but then you have to get into the deals that these guys are these data centers are having for like the you know, low cost energy, low cost water, like just below what the public play pays. And you start to see that just man, this is this is all kind of screwed up. But anyway, before we get too far down that rabbit hole, um, I don't know. I think we should I we're I think we're coming up on time. We'll go ahead and close this up. Uh any final any final thoughts for this week?

AI Spend Whiplash And Cloud Parallels

Katherine McNamara 35:45

It'll be interesting to see where this is all at in five to ten years. Um terrifying and interesting. Um you know, one of the ongoing things that like, you know, just even in the last month is, you know, like you know, Q1 of this year, like um, or even last year. Um, you know, a lot of companies were like, we're going do we're doing AI first. We we're gonna have leaderboards for whoever is using the most AI and celebrate them. And the the prices of AI, uh like using AI tokens, everything has gone up and people have, you know, companies are now asking people to to you know to uh stop using it as much. Not sp not spend as much money. So like this rush to build as many uh AI data centers as possible and stuff like that, it'll be interesting because we're we're kind of getting to this this point where these two colliding interests are gonna hit, like the the the need to buy like to create new AI data centers and build more resources, but also companies starting to realize this might be more exing AI might be more expensive than actually having human bodies. And so I'm wondering if i I'm I'm interested to see in five, ten years time, do these models become efficient enough that they're not no longer more expensive than human beings? Um are these all these data centers that they're building, you know, uh, you know, to try to be competitive, are they going to really be used, you know, in five, ten years if the model prices keep going up? Um and how are companies going to react if the prices do go up? Are they gonna cut back on AI and basically make this all this whole thing like void? Or, you know, it'll be interesting to see where this goes. Obviously, think this is like our our industry, like any tech tech changes all the time. Things change, sentiment changes. So it'll be interesting to see where it goes. And that's just kind of my final thought on it. Obviously, I don't have a crystal ball, neither do you. We're just kind of vibing it out as well and watching things change on like a monthly or quarterly basis. Like last quarter, use more AI. This quarter, a lot of companies are like, use less AI, wait, stop. Like Microsoft told everyone to stop uh to cut back on token usage and start canceling subscriptions because it got to be too expensive.

Tim 37:57

Yeah, yeah, yeah. So uh one thing uh you reminded me, and then we'll we'll close it out. Um, so one thing I've been saying this for a while, Chris has been saying this for a while. So, way back our our like fourth episode, maybe like this is you know almost three years ago now or whatever, we were talking about cloud, and we were talking about where are the savings that we were promised. Because part of the whole thing about going cloud first was we're gonna give you an economy of scale. Like AWS, all the people were like, we're gonna give you an economy of scale because we have all this gear, we have all this capacity, we can make it more affordable for you to run your stuff with us than you could build your own. Now that was the promise. And why the reason it never materialized, which we posited at the time, was right, it they couldn't actually lower costs because if they lowered costs, you would have so many people come in that you'd be DDoSing your own, like every customer would be be snatching the the limited capacity. And I think I've been saying again for a while that a the the the AI thing is very close to what the cloud how the cloud thing played out is that it's the same thing, right? So that what they're doing is frantically trying to build capacity, but they've run up against physics, right? There's there's not enough rare earth that you can't get the rare earths out of the ground fast enough, you can't fab the the chips fast enough, you can't build, you can't stand up the materials fast or the data centers. Can't buy RAM fast enough. Can't buy RAM fast. That's what I'm saying. You can't create it fast enough, and that's why half of these data centers have been canceled or mothballed for a while, not just because they've been pushed out, but even the ones that were accepted, simply that there's nothing to put in them, right? Like so um I think we're gonna see, and now that we are in real time seeing the same, and now people are saying things like AI first, the you know, they used to say cloud first. It was the same thing. We'll do everything with AI, we'll do everything with cloud. It's really interesting to see the cycle repeat itself with a with a different technology, and I know it's the same as it ever was. You go back and and that's always been the case, right? But it's it's really interesting to live through two cycles that are so closely put together and see the same exact type of uh behavior anyway.

Katherine McNamara 40:08

I was also gonna say, like, one of the weird things about this day and age, like cloud, like when c cloud first happened though, like it wasn't about replacing people. Like, though I did hear a lot of people saying, Oh, we're not gonna need network engineers anymore and stuff like that, and I'm like, You're still gonna have to connect to the cloud, you idiot.

Speaker 1 40:26

Yeah, yeah, yeah, exactly.

Katherine McNamara 40:27

I mean, you still are gonna need infrastructure at you know at offices. But um, yeah, cloud was you know initially gonna be some savings, and then there's like, okay, it's not savings, it's just opex cost instead of capex cost.

Speaker 1 40:39

Right.

Katherine McNamara 40:40

And then um, yeah, it it certainly hasn't made it cheaper for anyone uh at at this point. But AI first has been interesting because the promise is to replace humans in a lot of these things, or at least like the idea is, and it's interesting because the stock market is actually like 10 years ago, five years ago, if you laid off if you were a company laying off a cr a shit ton of people, uh, pardon my French, um, you basically it was a sign of instability of like you're not doing well. These it's so stock prices would potentially get hit. These days, you lay off 30,000 people, it's seen as innovative. It's innovative because you're doing so much with AI. So you're actually rewarded and you get a stock price bump. I'm more curious to see how long that'll last for.

Tim 41:24

Yeah, I agree. And I don't think I think the the chickens have to come home to a roost on that one. You can't cut, but so far without impact. So all right, let's go ahead.

Katherine McNamara 41:32

It's starting to with the the cost of AI raise uh rising.

Tim 41:35

Yeah, yeah, yeah. Agreed, agreed.

Final Thoughts And Closing

Tim 41:37

So all right, we'll go ahead and wrap it up here. Uh thanks for joining us um this month on the Cables Clouds monthly update. Hopefully next time we'll have Chris with us. And uh oh yeah, Chris wanted to make sure to apologize. I'm sorry we missed our our uh release. It's our first time ever missing a release. Catherine and I were super busy with Cisco Live, and Chris was, I don't know, I don't know, maybe he was drinking. I'm I'm not sure what he was up to, but I'm just kidding.

Katherine McNamara 42:02

He was yelling at us to get unbusy. Oh yeah, that's right, he was yelling us to record it. But no, it's it's it's all good. We actually have a bunch of people lined up for interviews and and uh and shows. We might actually end up having a backlog of uh of stuff soon enough.

Tim 42:19

Which will be nice. So, okay, we'll see everybody on the next episode, and thanks for joining us to go.