Latest / Putting the AI in MRKTNG / AI Has a Power Problem. America Isn't Ready.
Transcript
- Chris Heinemann: AI needs more compute, which requires significantly more energy. And while everyone wants the benefits of AI, nobody wants the data center in their backyard. So today we're going inside the AI data center. What they are, why they've become so controversial, and why the race to build them is the real race no one's talking about, at least until after the November elections. Joe Manuele, a longtime Silicon Valley executive and entrepreneur, now turned AI strategist and consultant to some of the largest companies in the world, joins the podcast to talk about the AI data center controversy and why they've become so politically charged and where all this is headed, particularly after the explosive comments made recently by Anthropic researcher Jacob Coxon. Joe, great to see you. Welcome to the podcast. Joe Manuele: Thank you, Chris. Thanks for having me. Happy to be here today. Chris Heinemann: All right, we're gonna start at the beginning, Joe. You know, you and I have been around. We've seen a lot of disruptive technologies that we have brought to market, ranging from IoT and blockchain and Ethernet technologies to the first WebRTC cloud-based video conferencing system. Here we are with AI. And I think it's safe to say that we have seen no other technology move at the pace that this technology is moving at. I think everybody's just trying to hang on. Joe Manuele: Yeah. Chris Heinemann: I can't even imagine what's going on in the C-suite these days, especially after some of the comments today and what we're seeing in the markets as well as as the response. Let's start at the beginning. So, the AI data center, this is where the AI lives. This is where guys like you and me and everybody who's using ChatGPT, Perplexity, Claude, et cetera, maybe Canva, they access it through the internet. So let's Joe Manuele: Mm-hmm. Chris Heinemann: let's talk about the AI data center and what it is and how it's different from the data centers that you and I were a part of in the early cloud computing days. Joe Manuele: Yeah, sure sure, Chris. And look, I'll I'll start at the risk of aging myself. In the days when I used to build my own personal computers, you know, my PC, I buy a motherboard with an Intel CPU on it. So it had a CPU, usually an Intel. 286 or X86, depending on where you were in the depending on how far back you want me to take you. there was some RAMs, some local memory, a hard drive, and a video card that would drive your graphics, right? ATI, King Company actually made made a really cool and cost effective video card back in the day. But the combination of compute, storage, and then eventually a modem for bandwidth was what you know, comprised of you know the early days of of of compute, right? You needed those three things to connect and to communicate. so obviously these systems, minis, and mainframes developed over time. And then the idea is, hey, what if we combined compute and storage and bandwidth in a central location? And and that's basically how public cloud infrastructure came to be. So honestly, only became available because of another big revolution, which was the advent of networking led by Cisco, right? So there are a lot of people compare the NVIDIA journey to the Cisco journey Chris Heinemann: Mm. Joe Manuele: when in the mid to late 90s we were connecting everyone and everything, First, businesses to each other, and then businesses to the internet, right? And then adding security layers on top of that. I mean, that was nuts. You know, that was fast. That changed so rapidly. But it pales in comparison to the Chris Heinemann: Right. Joe Manuele: the rate of change of what we're seeing right now with AI. So so if you think of the data centers that are building the traditional cloud infrastructure players. So think of Amazon with AWS, think of Microsoft with Azure, and Google with GCP, Google Cloud platform. These are very large data centers that you know have a combination of compute, storage, and bandwidth. And you can buy them in any kind of combination that suits your particular application. So when we were at Highfive, for example, Chris Heinemann: Mm-hmm. Joe Manuele: we're doing WebRTC. You know, the RTC is real-time communication. So the RT means latency mattered, bandwidth mattered. So the workloads that we bought from AWS. We paid a premium for a higher network connection with AWS and we had compute. Compute was important, but we offloaded a lot of that to our customers and then storage to store the videos, which is also a big you know undertaking. So the these data centers were built. you know, basically near water, near cold water. If you actually look at a lot of the original data centers. Microsoft has one up in the in the Pacific Northwest. And the the idea is I'm going to pump cold water, you know, into this infrastructure to keep everything cool. That was like the obvious way to cool you know, these these very large data centers. And and these data centers proliferated all over the world driven primarily by sovereignty rules like GDPR laws, right? Where in countries like Brazil, Germany, and now Canada, if you're storing any kind of personal information, it geographically needs to be in the country that you're doing business with, right? And so all of Chris Heinemann: Be there. Yeah. Joe Manuele: a sudden everybody started building these POPs connected with f fast internet POPs. And it was relatively little noise about that, actually, Chris. You know, like you think of the environmental impacts of of a of a large data center, these data centers are massive, right? But now the the the big shift towards AI data centers, that started with a couple of professors, a research project actually out of the University of Toronto, where a few researchers, and I'm gonna pronounce the names wrong, it's gonna be Alex Krizhevsky, and Ilya Sutskever, who actually became one of the founders of OpenAI, under the the under professor Geoffrey Hinton, they discovered through their machine learning research that the use of a of a GPU, which think is just a video card, graphic processing Chris Heinemann: Mm-hmm. Joe Manuele: unit, specifically one by NVIDIA, was really, really amazing and provided much more compute power in this use case than a traditional x86 or ARM architecture processor. And so that was kind of like one of the major milestones in the evolution, if you will, of this AI world that we're living in. And that wasn't that long ago. Chris Heinemann: Yeah, right. Joe Manuele: So then you think about I think about these large milestone moments, you know, in in some of the big trends. So in the internet era, Marc Andreessen worked on this project called Mosaic and basically created the first browser. And so the way you connected to the internet before the browser was using CLI, command line interface, you know, like you would basically type in, you know, connect to this guy, HTTP with your IP address, and it was basically reserved for you know geeks, effectively. And there Chris Heinemann: Mm-hmm. Joe Manuele: were some screens that started to develop and where you could do some texting back and forth and some you know rudimentary forms of messaging. A lot of these apps were developed for DARPA, by DARPA, for the US military, actually, but the creation of the browser brought it to the masses, right? Chris Heinemann: Mm-hmm. Joe Manuele: So the equivalent to that today, the parallel to that to Mosaic, is Chat GPT. So OpenAI. So Ilya, who was working on this research project as one of the founders, you know, these guys started buying GPUs and piling up GPUs. And at the time, GPUs were built specifically as graphic processing units. That's what a GPU stands for, not AI. Chris Heinemann: Were were they were they buying those from NVIDIA, Joe? Joe Manuele: And they were buying them from NVIDIA, absolutely. Yes. and so this was a big shift. Like all of a sudden, like Chat GPT comes out, and people are realizing, holy moly, like this is this is insane. Like the the capabilities, the the like the the equivalent of the Mosaic experience, if you will, is with Chat GPT. And I could use my natural language processing and NLP where I could just use my natural language and talk or type into Chat GPT and ask it. question like I would ask you a question. So Chris Heinemann: Mm-hmm. Joe Manuele: you can think of it as the evolution of you know search, right? You went from Google search and asking Google, you Google something and now you Chat GPT something. It's become a verb. And and that proliferated the use. And now there's other frontier models that everybody started to build this stuff. And the problem is that GeeP the way the on the one side it's amazing. On the other side, these GPUs were never built for this. Right? They were built for a couple of you know GPUs maybe in a computer cluster. Now all of a sudden you're the last model that was trained by OpenAI used a hundred thousand GPUs, Chris Heinemann: Wow. Joe Manuele: just to give you an order of magnitude, right? And so These GPUs just consume an inordinate amount of power. And so the biggest difference between an AI data center, very long-winded way to answer your question, Chris, and a regular, you know. Cloud data center is the addition of a GPU. So you still have all the other stuff. You still have CPUs actually in in GCP and Google Cloud and in AWS, but now you'll you'll have services and workloads that you're adding GPUs now to your workload, and they just consume exponentially more power than a regular compute, you know, x86 architecture CPU would. Chris Heinemann: Mm-hmm. Joe Manuele: And this is part of the dilemma now that we have is that these data centers on the one side, there's un unsatiable demand for them, because it's affecting absolutely everything that we do. And on the other side, we're just not built for that. Like the grid, the electrical grid in North America, actually more suited than than in Europe or in Asia. We just don't have the infrastructure to support the demand right now. It's it's actually the one thing slowing down the deployment of AI. Chris Heinemann: Interesting. Y you talk about power, excuse me. So, you know, we're hearing numbers like two hundred, three hundred, five hundred, even a you know, a thousand megawatts associated with these data centers, these facilities. Put that into perspective. I mean, how much power are we talking about? I mean, what what could you juxtap position that against, if you will? Joe Manuele: Yeah, I mean, you know, the people are throwing out I'm I'm working on a few projects, I can't really talk about who I'm working with, but you know, it's it's like if you're building a data center, the minimum size data center now is a gigawatt, you know, which is a thousand megawatts. That would probably power half a million homes. Like it's it's a ridiculous amount of power. Chris Heinemann: Mm-hmm. Mm-hmm. Joe Manuele: The biggest project that I've seen announced and it's being deployed in Europe right now is in the 200 to 300 megawatts area, right? So I'll give you a like a little bit of an order of magnitude. The latest NVIDIA platform, the NV72s, a single rack of an N V72, w which in a standard, you know, data center rack will take one megawatt of power. So That means you can only put a thousand racks of NV72s and there's seventy-two GPUs, you know, per Chris Heinemann: Mm-hmm. Joe Manuele: rack. Now each GPU is like again exponentially more powerful than the one before it, right? but the the a single rack requires a megawatt. And so if you're putting a thousand of these things, which is sounds like a lot, but it's not, Chris Heinemann: Right. Joe Manuele: in in in in based on the demand, the footprint for the GPUs is not that. That and then you need another 30 to 40 percent on top of that to power everything else, right? So you need to power your your GPUs, your your routers, your switches, you know, and and and everything that goes along around it, your rest of your cloud infrastructure. So a typical cloud AI data center being built today typically starts at about one three, one four, one five gigawatts, and this is in North America. What's interesting is that you just don't have a gig and a gig and a half of power available in most European countries. And and they all want to build their own AI data centers, right? So you're getting into this challenge right now where, you know, this is Moore's Law on steroids from a compute power perspective. The next challenge that we need to solve is the amount of power required to power these GPUs, and there's a lot of advancement in that going on right now. Chris Heinemann: Well well, let's talk about that because yeah, you know, you what you just mentioned, you know, We now understand, you know, the scale and Joe Manuele: Yeah. Chris Heinemann: and the power consumption that's going to be needed. So the controversy itself starts to make a little bit more sense. And but Communities, as you know, they're worried about the electricity prices, water usage, noise, land use, you know, strain on the grid. I mean, are the are these concerns warranted? Are they legitimate? Are they based on misconceptions? What's your take on that? Joe Manuele: I mean, look, this s some concern is warranted for sure. So the the reality is it's become a political balloon right now. And my personal opinion is that post midterm, I mean everything's a political, you know, matter right now, this close to Chris Heinemann: Mm-hmm. Joe Manuele: the midterms, right? Like should the WNBA team visit the White House if they win their championship is a political issue, right? But the you know passed the midterms, I think there'll be some saner minds. I am working with a couple of congressmen on some legislation. So helping them out. So th look, there's a couple of things. These are local issues, first of all. I I I wanna like first point out that, Chris Heinemann: Mm. I agree. I agree. Joe Manuele: You know, the local counties and the towns that this affects, you know, have to have a big say in this. And they do. You know, they they they they really, really do. And they have to outweigh the benefits and the the risks that you're that you're getting. So the concerns around water are completely unfounded. So may have been true a few years ago, but all the latest GPUs and chassis and all work on closed loop systems. So Chris Heinemann: Yeah, I want to talk about that. Yeah, let's yeah, dive into that a little bit because I don't think people realize these yeah. Yeah. Joe Manuele: Yeah, so I wanna I wanna yeah, I wanna dispel the water thing for one. Like I I saw, you know, I I don't know if I need to name the politician, but sh you know, there's a politician who you know shows up with a glass of water and it's all brown, like, look, look, they're doing more water, and it's like, my god, like it's it's not an Aaron Brekovich moment, right? actually she's actually fighting against data centers too, which is another conversation. But the the way that the latest data centers are built is water comes in, it circulates, and it completely recirculates. And so it requires a fraction of the amount of water. It's actually insignificant, right? So someone said to me like it takes something like 50 times more water to grow one almond tree in California than it does to power, you know, a data center, a gigawatt data center. So the water issue is a non-issue, right? It's been solved. It's a problem that's been solved. using some clever closed loop systems. I mean you you need to top water up and stuff, but it's you're not polluting any sources of water, right? You can and and you don't also need to pull, you know, well drinking water. Like you can use salt water. It doesn't matter. Like it's, you know, as long as it can be cooled and and you fill it once, kinda like filling a pool. You gotta top it up Chris Heinemann: Mm-hmm. Joe Manuele: once in a while, but it's not, you know, an enormous amount of water. the concerns f on, you know, are we Are we taking farmland away and are we changing zoning to let farmland become used for industrial purposes? Well, you can make that argument for anything. You know, do I build do I take this farmland and build this massive Amazon warehouse? or should I be growing cabbages and corn there instead? I mean, like, but again, these are local issues. These are issues that you know, local politicians, you know, up to the state level. We'll look at that and they regulate and keep an eye on that, make sure that we have enough agricultural land for that. Again, not a federal thing. I don't think it should be regulated at a federal level. And then the third one, and I and I think this one is one that that's interesting, and there's there's probably the most debate on, is well, if I you know go into a small community in Pennsylvania, I'm working on a project out there. Pennsylvania's got a lot of data centers because they have a lot of natural resources. If all of sudden I suck a gigawatt of power out of Pennsylvania power and that and P you know PPL, does that mean that the cost of power will go up for everybody else? Again, it's a local issue. Like PPL is regulated by the state of Pennsylvania. So this is up to Governor Shapiro to make sure that if they allow new data centers to go in, that whoever's building these data centers. Is either doing you know in front of the meter power production, which is what a lot of them are doing now, or they're adding to the grid to the point where it may actually lower the cost of electricity. So there's been an Chris Heinemann: Mm-hmm. Mm-hmm. Joe Manuele: argument that, like, hey, if I'm you know making these massive infrastructure upgrades and I'm producing, you know, five times more power than I was before for my data centers, well, maybe I'll produce 10 times more and I'll make costs of power lower for everybody else, right? and that has its own implications as well. So the question is, well, how do you produce this power, right? So one is the most popular right now in the US is liquid natural gas. I'm gonna take LNG and I'm going to burn it and produce electricity with that in a substation. By the way, if you're building an AI data center and you're converting and creating power, the electrical substation To create the power takes way more space than the actual data center itself, right? Like I can now put a rack of N V seventy twos and that that would be the equivalent of a hundred racks of previous generation. So I I need less data center space, but I need more room for these electrical substations. And that's Chris Heinemann: I I I've I've heard China's building out these what you're talking about. I mean n not hyperbole, Joe Manuele: Yeah. Chris Heinemann: but you know, the size of the country of Germany, literally, to you know, to be able to power, you know, data centers and what they're doing with AI. Joe Manuele: Right. So so China's interesting because, you know, so LNG is number one. The way we produce a lot of our electricity, I hate to say it, is we burn coal. That's some something like 30% of Pennsylvania's power comes from burning coal. they claim it's a clean coal. They have a very unique coal that produces less carbon capture issues. but there's technologies coming out now where they're capturing the carbon and so that it does protect the ozone. and you so you're you're seeing a lot of innovation in in the production of power and the production of clean power. So we talk about you know the innovation of the GPU side. what we rarely talk about, you know, in the software companies and all that, but we're what we rarely talk about is how the whole energy sector has been, pardon the pun, re-energized, right? Chris Heinemann: Mm-hmm. Joe Manuele: you see what I did there, Chris. Chris Heinemann: I did, I did. Joe Manuele: So now all of a sudden nuclear is cool, right? And like the the most efficient, cost effective way to produce this amount of energy is power plants that are powered. You have these purpose-built nuclear power plants. Like you can imagine like building a two or three gigawatt power plant would be much, much smaller, much more cost effective and cleaner from an energy perspective than burning liquid natural gas and certainly than burning coal, right? but then you you talked about, you know, not in my backyard. You know, you and I are old enough to remember Three Mile Island. if you don't remember Three Mile Island, you've probably seen Chernobyl on Netflix, right? And everything associated with that. And so we need we need to get over this fear of of you know the cleanest, you know, cheapest source of power. I think I I listened to a podcast from Andreessen, who I respect on the topic quite a bit. He said, you know, we built a thousand small nuclear power plants, we'd have all the power we would need. And and the cost of energy, by the way, would fall precipitously for everyone, for absolutely everyone, right? So this is the longer term view that I don't think people are taking, unfortunately. Chris Heinemann: Yeah, yeah. You you you you you said a lot in in in this segment and and I and I want to go back to Pennsylvania, but before I do that as well, the closed loop water system. Now and and and 'cause you're talking about nuclear right now, is that s is that those systems are they synonymous, the same kind of technology that the nuclear power plants were using to bring in cold water to to cool their facilities? Joe Manuele: not really. I mean like it well I I suppose, but on a much smaller scale, Chris. Right. I mean we're not like like Chris Heinemann: Okay, okay. Yeah, no, 'cause it that's fascinating. Joe Manuele: the amount of heat that a nuclear core produces is like you th if you I'm thinking about like we talked about you know three mile island in our previous conversation, right? Coming back up. I'm talking about much, much smaller versions of Chris Heinemann: Right. Joe Manuele: right, like not at the power generation side, but like just Using small nuclear power plants to power, specifically to power a building. And then if you have extra capacity, you put it back in the grid. And where Chris Heinemann: Yeah, yeah. Joe Manuele: the where you know you're producing more energy than you need so that the cost to the consumer goes down dramatically. and and you're producing clean power you know along the way. Now, the government obviously needs to regulate that, right? but you know, we lost a lot of our nuclear knowledge after the last meltdown and people's like, Yeah, we're not building any more nuclear power plants. unfortunately. France is probably the one country that's reversed the the that take. China's building nuclear power plants, coal power plants, you know, solar power plants are building everything. It doesn't matter, right? Like they're they're on. Chris Heinemann: Yeah, yeah. France is leasing nuclear energy out and and and making money on it for for the country. Joe Manuele: Yeah, f that it's actually it was a great investment, probably the smartest thing Macron ever did. Shout Chris Heinemann: Yeah, yeah. Joe Manuele: out to Emmanuel Macron. Chris Heinemann: I'm I'm thinking in my my mind's eye here, Chat GPT tell me in like a fifth grader, I'm I'm having visions of my gas water heater versus electronic electrical water heater to explain Joe Manuele: Yeah. Chris Heinemann: the co the the the closed loop water systems. All right, let's yeah, Joe Manuele: Fantastic analogy. Chris Heinemann: let's let's get back to let's get back to Pennsylvania because this this is there's a big play happening here. And that's Constellation Energy. So they're the largest nuclear power operator in the US. They're actually restarting Three Mile Island Unit Number One. And they've got Joe Manuele: Right. Chris Heinemann: a 20-year agreement with Microsoft. And and to be clear, I want people to understand that Three Mile Island, the unit one, is not where is not the damage unit two reactor from the the 1979 accident. So just so we're clear and I don't get a bunch of hate mail or any comments from anybody about that. So But listen, Joe Manuele: Yeah. Chris Heinemann: you know, the the plan could be operational, unit one by twenty twenty-seven, you know, subject to, you know, regulations and yada, yada, yada, and providing more than eight hundred megawatts of of carbon-free electricity to the grid. So we're hitting those numbers that you talked about. What people don't understand is that it's estimated from some of the the reports that they've done in their own audits and studies that 3,000 jobs could be created. add Joe Manuele: yeah. Chris Heinemann: $16 billion to Pennsylvania's GDP and generate Joe Manuele: Mm-hmm. Chris Heinemann: more than $3 billion in state and federal taxes. That sounds like a win to me, Joe. And it sounds like a massive inflection point too for the state of Pennsylvania as well as nuclear energy, because this clearly is not just a trend, right? Joe Manuele: Yeah. I mean that's that's you know what's wild about what you said and what jumps out at me is that you know you take a nuclear power plant and you're right, you're you're basically taking one substation and you're Chris Heinemann: Mm-hmm. Joe Manuele: powering it up and you're spending years to get it up and running. It's it's the cleanest form of energy. It's eight hundred megawatts. It's not even big enough to Chris Heinemann: Ha ha ha. Joe Manuele: To meet the requirement of an of an anthropic RFP that's out on the streets right now, right? Like, Chris Heinemann: Yeah. Joe Manuele: which basically wants almost twice that, right? So, so yeah, look, I think it's a step in the right direction. there's no doubt in my mind that like look above and beyond the jobs that it's creating, it's the types of jobs it's creating. You know, like if you're an electrician, if you're like 27, 28 years old and you're and you're you're a badged electrician in Pennsylvania, you're making a hundred sixty to a hundred eighty K. You know? Like, think Chris Heinemann: Yeah, it's big money. Yeah. Joe Manuele: about that. Like that's a serious job, right? And and Chris Heinemann: Yeah. Joe Manuele: so we're not talking about, you know, Walmart greeting jobs here, right? We're talking about jobs, blue collar jobs, electricians, carpenters, welders, you know, that are needed to to scale this. It's like a mini industrial revolution that's going on. And and it's you know, the one argument is all you build a a big data center and then once it's gone, it's all machines in there and there it doesn't create a lot of jobs. But I mean it took it it takes 10 years to to build a a gigawatt data center, right? And that's just one. Like like Chris Heinemann: Yeah. Yeah. Joe Manuele: think of the you know the the amount that we're talking about. There's a massive project project going on right now. in Childress, Texas, with a company called IREN, I-R-E-N, Australian company that started mining Bitcoin, and they signed a massive contract with NVIDIA and with Microsoft. They're building five and a half gigawatts in Texas. And it's a town. It's a like you know, I've seen pictures and movies of it. It's you know, what's already been built. The first phase was delivered actually to Microsoft, and then the next phase is being delivered to NVIDIA, but it's you you think of the life-changing opportunities for generations that you're creating. And yes, of course, they did all the environmental studies, like no no water is being poisoned, right? And you're building it in central Texas, which tells you you don't need water like you did the traditional data centers Chris Heinemann: Right. Joe Manuele: where have to build them by the Pacific Ocean so you can pump cold salt water in, right? They can bring water in once and top it up. and what you need is power, right? And and they've you know, worked with the local utilities and the power that's being generated is built purposely for them and not for the grid. And there's ex excess capacity being built. And it's being regulated heavily, but in in conjunction with local, you know, governments, towns, counties, state level and at the federal level, less than people realize actually. and Texas has built a an environment where people are you know, able to go and invest literally hundreds of billions of dollars creating countless jobs, you know, for generations. So Chris Heinemann: Mm-hmm. Mm-hmm. Joe Manuele: yeah, I thought the benefits in my mind, I mean look, the economic benefits are there. Like no one's gonna argue that. you know, I someone said to me this morning that the AI economy is gener is responsible for one percent of the US's GDP right now, right? from next to nothing five years ago. And I think it's gonna be more than that in in the not to Distant future. So I I think I think that the the concerns, Chris, that you asked me about this, the concerns about these large data centers, are they legitimate? Yes, they're legitimate. and a lot of government agencies, you know, the old Ronald Reagan joke, hi, I'm from the government, I'm here to help. You the scariest Chris Heinemann: Yeah. Joe Manuele: words you'll ever hear. and you know, like in that when I work with with different folks at different levels of government, I say, hey. It's cool to want to regulate something, but educate yourself first, right? So shout out to Sam Licardo, by the way, who's my congressman here in in Silicon Valley. and you know, he's got an a a paper that he he's had basically a law that that that a bill that they want to propose. he's doing it across the aisle with a Republican, but they're not gonna propose it till after the election. You know, it drives me nuts, right? But he's getting information, he's educating himself. He's getting input from the industry and from local folks and local government to get their concerns and figuring out, okay, regulation's important, but there are benefits here. We don't want to miss the the the the the opportunity that's in front of us here. So it's it's critically Chris Heinemann: Well Joe Manuele: important that we have American-based innovation for this technology because this is there's nothing like this, honestly. Chris Heinemann: Yeah. Joe Manuele: Like in my life. And I've seen, you know, four or five major waves. This is, you know, the internet one is probably the biggest one that I live through. This is that internet one from the late nineties on steroids. Like Chris Heinemann: Yeah. Joe Manuele: it's it's exponentially faster and more more impactful. Chris Heinemann: No, a hundred percent. So how do you what you're talking about here? This is interesting because there certainly is a China syndrome effect happening here, right? You you remember the movie with Jack Lemon and Jane Fonda Joe Manuele: huh. Chris Heinemann: when when the power plant went down. Yeah, speaking of aging ourselves, so you've you Joe Manuele: Speaking of aging ourselves, yeah. Chris Heinemann: you've got Three Mile Island, you've got Chernobyl, you Fukushima, you've got these plants that have that major, massive catastrophes and the fallout. So you've got Congressman, you know. Coming up again, we've talked about after November, we'll probably see more messaging, more campaigning around this. Joe Manuele: Sure. Chris Heinemann: Can you can you give us a little insider baseball? I what what's a what's the key message or two that they might be leaning in on to help educate the masses? Because I mean, y you know more than ever, perception is reality. you and I have been doing a lot of sales and marketing and go to market together for a long time. You know in marketing, perception matters. You're the best non-marketing guy I know, and you you could never join us because you didn't want to take the pay cut. But it's it's a real, real thing. And so f I Joe Manuele: Yeah. Chris Heinemann: think these guys have got a massive job ahead of them, but I also realize that the train has left the station. And whether or not people get on board or not, I think it's gonna be happening anyway, but there there still has to be a level of respect and communication and and educating people about this. So give us a little more inside baseball. What's what's the message? Joe Manuele: Yeah. So yeah, I look I I this one's a tough one, right? So I I'm seeing young kids like early teens saying AI's bad for you, AI data data centers are bad. I'm like, you shouldn't be using ChatGPT, you know, or you shouldn't be using Grok or whatever. And I I saw that for the first time a couple of weeks ago, like, whoa, what is going on here, right? So, like, that's the level of misinformation that's out there right now. And and whether we like it or not, something like seventy-five percent of Americans view AI data data centers. In a negative way, right? So if you Chris Heinemann: Uh-huh. Joe Manuele: look at the political spectrum here in the US, you have the Bernie Sanders of the world, right? and Bernie's basically saying, I'm calling for a complete ban of AI data centers all over the country. Ain't gonna happen, but you know, that's the one extreme. and then on the other side is Trump who's saying You know, whatever the the the drill baby drill version is of AI data centers, right? basically saying, no, we gotta move ahead and we'll figure it out later. And somewhere in between lies reality. I really think that there's there's there's safeguards that we can put together on the construction of AI data centers, mostly with local governments that protect people against the fears and the concerns that they have. And a lot of those regulations, quite frankly, are already in place. Chris Heinemann: Yeah. Joe Manuele: but the shift now is is moving a little bit towards okay, I'd like let's just say I've convinced you, Chris, that an AI data center in your neighborhood is not gonna kill you, it's not gonna pollute your water, it's you're not gonna die from electric magnetic fields and you know whatever else. but now people are worried about these models are getting really, really good. And you know, these models are getting really, really smart. and you know, are the bots gonna kill us? Right? Like like where there there's not we don't have like the there's a shift now in regulation where people are saying, no, no, we don't need to regulate the data center. No, we need to actually regulate the models, right? Chris Heinemann: Mm. Joe Manuele: and that's a fantastic debate actually because and and I think you mentioned at the beginning of the podcast, right? It was triggered By an anthropic employee who, by the way, had been with the company for like three months, right? when he issued his tweet, this is a guy who had no social media presence whatsoever. Chris Heinemann: Right. Joe Manuele: Within like 15 minutes of his tweet going out, a Wall Street Journal article just happens to drop. It gets amplified by you know every media out there. This was a controlled communication operation. This was not a random guy who worked for you know for Dario and who decided, man, like where the whole world's gonna come to an end. I better come out here and and tell people how bad it is. so be very suspicious of this particular one now. Dario, though, CEO founder of Anthropic, basically came out shortly after that and said, Hey, we're going too fast. We really need to slow this thing down. I'm really, really worried. Like, you know, we open the barn doors and the horses are out in the field, and man, like I don't think we can get them back in. And we we really need to stop innovation and development here. And then Sam Altman at Chat GPT or OpenAI says, Yes, I agree. Even Elon, even Elon, Elon is like, Yeah, I agree with with Chris Heinemann: Elon Musk, I agree. Yeah. Yeah. Yeah. Joe Manuele: Dario on this one, right? And and I'm like, holy shit, like this, okay, that I'm taking seriously now, right? Like this is this is a bit of a different moment. and then David Sacks, who is like the AI czar for President Trump and also one of the four all-in podcasters that I listened to religiously, fantastic podcast. Chris Heinemann: Ha ha ha. Joe Manuele: you know, David [Sacks] is the only one who came out and he went out on the you know the Sunday news shows saying, hang on a second, boys. Hang on. Like, if you think that what you built you can't control, you should step aside. You should stop. You should quit. Dario [Amodei], you don't think you can control what you built? Resign. Resign and get your freaking board of directors to hire someone that can do it. I mean I love that message, right? And and it's like you cannot ask the US federal government to create safeguards and to create you know boundaries around AI safeguards and safety, because at the end of the day, you what you really want is you want to keep anybody else, especially these open source models, from coming out and competing with you, right? So that is the counter argument to all this. And and Sacks is basically saying: look, if you release a model and it creates damage, you will be liable. You Chris Heinemann: Mm-hmm. Joe Manuele: under our laws, you will be liable both criminally and civilly. And you know, you have a responsibility. And yeah, you're building a really dangerous freaking product. Like you are. If I buy a Tesla and the thing, you know, catches on fire when it's charging in the middle of the night and takes my house down, Tesla's liable for that. So they put safeguards in there to make sure that that doesn't happen. So that's the counter argument to this. And you know there's there's a couple of camps being developed. There, you know, Marc Andreessen posted a great post on Twitter who's in the David Sacks camp? He doesn't believe the government should regulate these models because these models won't be regulated in China, they won't be regulated in India, they won't be regulated in, you know, you name it, you name the country, you know, like you you can't we can't regulate what the Chinese do. out of the US and we can't regulate what Canada does for that matter, you know, moving forward. So or the EU. And and and this is one of these things that quite frankly, I mean, if if you think that I don't think that politicians know much about AI data centers. Multiply that by 100 about how frontier models work, right? Chris Heinemann: Mm. Joe Manuele: Like can you imagine what legislation like that would look like? You know? And Bernie's basically saying, Well, I don't know, so let's put a pause on it until we figure it out. It's like, well, we don't have time to figure it out, Bernie. I'm sorry, right? Chris Heinemann: Yeah. Yeah. Joe Manuele: And but but the flip side of that is, you know, can you make sort of regulating the industry or or the models and and where the government says, Okay, you need to ship me the model first. I need to test it. Like imagine like, you know, in a in a software development cycle, the US federal government is the last SQA you have to pass, right? You gotta go wait, like go to the DMV and make sure that you're you know the five dot six is available now, right? Like I it it it's insane to me. So look we're we're just living in a a really, really exciting fast paced environment and there's absolutely gonna be collateral damage. There is no doubt in my mind that there's gonna be collateral damage. You rarely see this level of disruption without some collateral damage. But I'm of the camp that it's worth it. And I say that because Chris Heinemann: So this this week go ahead, sorry. Joe Manuele: I say that because it probably doesn't affect me maybe as as it as it affects others Chris Heinemann: Yeah. Joe Manuele: selfishly speaking, but I'm a tech optimist by nature, Chris, you know that. And this train has left the station, the horses have left the barn, the rockets have been launched in the air. He ain't bringing this thing back. That's basically what I'm saying. Chris Heinemann: wins." What are we gonna win, Joe? Joe Manuele: Well, I think look, there's the let's take Trump out of it for now, right? Chris Heinemann: Yeah, yeah. Joe Manuele: because you know winning for Trump might mean something different than for someone else. But look, there's a military aspect to this, right? Like let's not deny that. Like own intelligent models, You take all the drone technology, warfare has completely, completely changed. you know, the biggest defense contractors are Palantir and Andrurial, not McDonnell Douglas and Boeing, right? Like the guys that are growing the fastest are people that are implementing all these AI models to for that military use case. And you could make the argument that that we couldn't afford to lose that one, right? That's a battle that you can't afford to lose. We need to own that technology. We need to be better at that technology than anybody else, right? but then this this ends up being the nuclear, you know, debate, right? You know, if everybody has a nuclear bomb, no one's gonna use it against you. So, you know, mutual annihilation is what prevents you from dying. So eventually everyone else will catch up and have it. But it's you want to make sure you have it before Iran has it. You know, as as a as an argument, right? So there's that part of it. the other part of this is the the the the the economic impact of AI, which along the way is gonna be bumpy, but should create a level of productivity that we've never seen before. And if done correctly, a level of abundance that we should all be able to share as a country. So GDP per individual in the US should skyrocket. You know, how you make sure that it's not owned by the point zero zero zero two percent. I I Chris Heinemann: Right. Joe Manuele: that's a political conversation, but so the the economic benefits of AI, and look, we we are already the US is already twenty-five percent of the world's GDP, thirty-five percent of the world's consumer spending. and you you want to make sure that that continues. If you want to make sure that that continues, you need to make sure that this technology stays in the US first and foremost, and then you know, we become the engine of innovation for for AI and then and in all of the sub industries that have powers, right? so you know, to answer the question, what do you win? First is the military power, which you want to make sure you have before anybody else. And you know, again, you know, if you're if you're the military government. or if you're if you're the American government that has to be top of mind and then next is you want to maintain your your position as the economic superpower and if you don't own it, you you'd you'll absolutely lose that. Chris Heinemann: Yeah, yeah. For those people who are maybe listening and watching and who've been watching binging on Shameless for the last week, I wanted Joe Manuele: Ha ha. Chris Heinemann: I wanted just to reiterate what Coxon had said because it's if if you hadn't heard his statement and and there there's a podcast out that's right now on on the New York Times that he was a guest on and he said it was Joe Manuele: Yeah. Chris Heinemann: just a tweet. It was just a tweet, you know? But what the tweet said was, "The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible. But I hear the same people express fear privately. No other human activity poses this level of danger." That's what was said. So in case you missed it, I just wanted to be able to read what he had. Joe Manuele: Yeah, no no yeah, that's fair that you repeated that, Chris. Yeah. Chris Heinemann: That he tweeted out. And and and and to your point, you know, Dario Amodei, you know, he that's that he came out pretty quickly and called for the development of the slowdown. And Altman and and and Musk agreed and that they chimed in. And and in fact, as we're recording right now, you know, what time is it? it's one forty-two on Tuesday. Altman's Joe Manuele: Yeah. Chris Heinemann: actually at at Dreamforce. He's gonna be speaking as as as a keynote. So that should be an Joe Manuele: Right. Chris Heinemann: interesting. Joe Manuele: San Francisco, yeah. Chris Heinemann: So I wanna, yeah, I wanna I wanna keep going. You mentioned some numbers and you mentioned GDP. I know you said 1%. I Goldman Sacks has actually got a number of 5% of GDP, which I think is fascinating. But I but I wanna I wanna bring this all around because I think this all kind of comes full circle. And as I mentioned earlier, it's all interconnected. I want to talk layoffs. So July 2026, Microsoft almost 5,000 jobs. May 2026, Meta 8,000 jobs. Amazon 16,000 jobs 2026, January, roughly 14,000 cuts in October 2025. Just yesterday, September 14th, a bunch of folks at Oracle got the email, right? As you know. Joe Manuele: Yeah. Chris Heinemann: They had a bunch of massive layoffs. They're up to 21,000 employees over FY 2026. Joe Manuele: Right. Chris Heinemann: Here's the kicker though. Here's the kicker. The AI spending is approaching $800 billion in 2026. And Reuters is also reporting that the total AI industry spending is expected to reach nearly $800 billion this year and then $1.8 trillion in 2027. So the message is clear. You know, it's they're not eliminating the spending. That's not what's happening, I think, with Joe Manuele: Right. Chris Heinemann: the layoffs, but it's just changing where the money goes: GPUs, Joe Manuele: Right. Chris Heinemann: data centers, networking, electricity. Nuclear power. Like you said, Joe, the train has left the station. So it's Joe Manuele: Yeah. Chris Heinemann: a very interesting time that we're living in. And I think you know, I I I think this the boardrooms, the C-suites themselves are trying to figure it out. We're all trying to figure it out what's going to happen with with jobs, et cetera. But I Joe Manuele: Yeah. Chris Heinemann: don't think it's all as doom and gloom as maybe people think. And I think the discussion today has been really Joe Manuele: Sure. Chris Heinemann: interesting to kind of frame it up and and give another perspective of that. I've been cynical. I'm a journalist by nature, you know that, as my my education. And I can be jaded and cynical as as much as the next guy then and the next Dave Michaels, if you will. And I I shout out to Mr. Joe Manuele: Shout out to Dave. Chris Heinemann: Michaels and Talking Pointz. But Joe Manuele: Yeah. Chris Heinemann: we're we're we're we're at probably the greatest inflection point in in technology and and mankind that you and I have ever been around. Joe Manuele: Yeah. Yeah. So look, Chris, you know, the the the layoffs, I you know, you Microsoft how many hundreds of thousands of people does Microsoft employ, right? So th there's always a culling that happens here and there. there's also, you know, Oracle's doing a major shift. Like Oracle is is Chris Heinemann: Right. Joe Manuele: going away from you know their core business and and and spending a ton of money on CapEx and on data centers on AI. and so like yeah Chris Heinemann: Yeah, almost a hundred billion for next year. Joe Manuele: and and look, it it it really, really, really sucks if you're affected. If you've been working for Oracle for 20 years and you lose your job, you know, you're no longer a project manager for a an Oracle database transformation project or whatever. Chris Heinemann: Don't be in customer support. Joe Manuele: Yeah, I mean, yeah, customer support. like you got s you know the the the the irony here is a lot of a lot of those jobs would be disappearing and going to like, you know, Indian outsourcers, right? And now the Indian outsourcers are being, you know, replaced by agentic agents, right? Like so Chris Heinemann: Mm-hmm. Joe Manuele: like you you you're you're getting this evolution here for certain jobs. There's no doubt about that. But you know, the like the One of my favorite lines from movie Wall Street, a classic, I don't think it won any awards, but I remembered a few lines. Like, money isn't made, it's exchanged. you're not making money. And so that money is not like these jobs are are, yeah. If you're the person affected, and I've been there by the way, I've been laid off and it sucks. There's nothing scarier, especially when you have a young family and all that. if you're affected, it hurts. But the flip side of that is. There's a lot of people who are, you know, there's electricians in little towns in Pennsylvania, you know, 27, 28-year-old guys who or or maybe even former military folks who, you know, served in the military. They go back home and you know, they would have to leave because there's no jobs in the little towns that they're in all of a sudden, like, wait a second, I can get hired as a welder, or I can get hired, you know, as an electrician, and I can support my family, and I can buy a house for 250K, you know, and get a mortgage. And wow, what a concept. And so there's a shift happening and then look, as an industry, we are disruptors, right? Leave we've always Chris Heinemann: Mm-hmm. Mm-hmm. Joe Manuele: disrupted traditional industries. That's what we've always done the last, you know, my my whole career, your whole career. Chris Heinemann: Mm-hmm. Joe Manuele: We are disrupting ourselves here with AI. And that's what's scary, right? Chris Heinemann: Right. Right. Joe Manuele: And so, you know, if you're a COBOL programmer working for IBM, and you had job security because you know it's stuck in the mainframe, and Bank of America is not getting rid of that mainframe. And then some guy at Claude says, hey, guess what? I can code in COBOL now. Yeah, you're out of a job. Maybe you should have retooled 25 years ago, right? Like, I don't know. But like, I'm just saying, like, if you're a COBOL programmer, yeah, Chris Heinemann: Well just four years ago, they were they were you and I have got kids. Go code, go code, go code, learn how to code. Joe Manuele: you know, I was talking to my son Alex, he's in bioinformatics, he's a coder, and his job, he like, he's 34 years old. His job changed dramatically in the last six months. You know, he f he gets a Jira, whoops, and he he fires up Claude and watches the thing code and he says, I used to code myself and now I'm just monitoring code that Claude is producing. I go, Well, what's the biggest impact? You know, what what's the biggest thing that that that's had to your life and to your career? He said, My apartment's never been cleaner. Right, like Chris Heinemann: That's awesome. I love that. Yeah. Joe Manuele: You know, I I said, okay. So like, are they are they still paying you as much? He goes, Yeah, they gave me a big raise. All right, cool. All right, good for you. Like, what can I say? Right. So he Chris Heinemann: yeah. Joe Manuele: his job is shifting to managing multiple bots that will write code on his behalf. Like it's, and so look, this is the the who moved my cheese thing. If you haven't read the book, I highly recommend it. Chris Heinemann: Yes, change management. Yep. Joe Manuele: You know, like everyone's cheese is being moved right now, man. Like everyone's. Chris Heinemann: Yeah, yeah, yeah. Joe Manuele: I just, you know, last night watching TV, I played around with Chat GPT has a new feature called Sites. And based on chat history of all the consulting work I did, it built me a website in seven minutes, Chris. And it's Chris Heinemann: Hmm. Joe Manuele: scary good. I'll send it to you actually. And I I hope that maybe when the by the time the podcast drops, I'll have it out there. But seven minutes. Chris Heinemann: Wow. Joe Manuele: To produce a website that was like incredible, like just based on our history. Chris Heinemann: I'll I'll I'll run it through my Geo audit tool and we'll see how you score. Joe Manuele: Yeah, yeah, well we'll we'll add it to that. So look, I my my message for everyone is man, you know, if you're if you're just on the sidelines and complaining, you're gonna be a victim of this. This is very, very exciting. And I didn't want to bring my personal life into this, but you know what I went through, Chris? I went I went through eight chemotherapy sessions with my stomach Chris Heinemann: Yeah. Joe Manuele: cancer and it was six hours at a time. And guess what I did when I sat on that chemo chair? I learned about AI, Chris Heinemann: You learned AI. Yeah. Joe Manuele: you know, and I I messed around with Claude, I messed around with Chat GPT and kind of kept me sane at the same time. And it's an amazing, amazing tool. And you know, the the the effects I use it to manage my health. in a way that I couldn't possibly have imagined it before. it is such a powerful, powerful tool. The benefits are far gonna outweigh the downside. And I encourage everyone to be optimistic about it. And, you know, get on the train and it's not too late. But Chris Heinemann: Yeah. Yeah. Joe Manuele: ain't it ain't coming back to the station though. You're gonna have to catch Chris Heinemann: Yeah. Joe Manuele: up to it. Chris Heinemann: And and to leave on a positive note, even for Jacob Coxon and himself, he did say that he's looking forward to all the benefits that AI is going to be helping with with medicine and healthcare. He said those are gonna be some of the the immediate benefits. So well, Joe, Joe Manuele: Right. Absolutely. Hundred percent. Chris Heinemann: thank you so much. You know, first and foremost, I'm just happy to see ya. I I I know you're you're you're on the more than on the mend. you're you're healthy and you're you're back and just couldn't Joe Manuele: I'm back, bro. I'm back. Chris Heinemann: be happier. And I've been going on this journey with you over the last couple of years as well. So Joe Manuele: Yeah, you have. Chris Heinemann: Just really thrilled. I'd love to have you back on after the midterms as well. And and kind Joe Manuele: Sure. Let's do it. Chris Heinemann: of curious to hear what's going to be happening with the congressman that you're working with, as well Joe Manuele: Yeah. Chris Heinemann: as some of the companies that you're working with and the business transformation that they're undergoing and trying to adapt to AI and Joe Manuele: Toby. Chris Heinemann: to their businesses and what they're doing with internally as well as with their customers. So just great to see you. I wish you all the best. And one final comment. wasn't lost on me that a Montreal Canadians fan is reading Toronto News. In case anybody didn't pick up on that earlier on about fifty three minutes ago. that wasn't lost on me, Joe. Joe Manuele: I'm very excited about my habs, by the way. Very, very excited. Chris, congrats on the Chris Heinemann: It's coming soon, next month. Joe Manuele: new podcast, by the way. I love it. I think it's fantastic. I think it's Chris Heinemann: Thank you. Joe Manuele: timely. I love the approach that you're taking to it. and you know, it's it's it's a fantastic addition to the there's a lot of noise out there and and and it's it's real and it's really useful and I love it. And it's it's It's a regular on my Spotify list. So congrats on the new pod and thanks very much for having me on. I'd love to come back. Chris Heinemann: I love it. thank you. Thank you. Appreciate ya. All right. Take care, Joe. We'll see ya. Bye bye. Joe Manuele: Cheers. Thank you, Chris.