WEBVTT 00:00:00.000 --> 00:00:03.819 We're seeing this profound paradox emerge right 00:00:03.819 --> 00:00:06.599 now in the technology world. On one hand, the 00:00:06.599 --> 00:00:08.439 sources we're looking at say that something like 00:00:08.439 --> 00:00:12.980 95 % of AI projects, they're failing. Yeah, just 00:00:12.980 --> 00:00:14.980 burning through cash with almost nothing to show 00:00:14.980 --> 00:00:18.519 for it. Exactly. And yet, at the exact same time, 00:00:18.640 --> 00:00:22.579 we're tracking a $3 trillion infrastructure investment 00:00:22.579 --> 00:00:26.219 cycle. Right. Flowing into this exact same industry. 00:00:26.420 --> 00:00:29.589 That's a staggering figure. Yeah. It suggests 00:00:29.589 --> 00:00:32.990 someone, somewhere, is absolutely minting money. 00:00:33.210 --> 00:00:35.710 It does. And the core insight we pulled from 00:00:35.710 --> 00:00:37.929 the source material is, well, it's a bit counterintuitive. 00:00:38.210 --> 00:00:40.869 The real durable profit isn't in the virtual 00:00:40.869 --> 00:00:43.109 penthouse, you know, the flashy apps on your 00:00:43.109 --> 00:00:46.030 phone. Nope. It's deep down in the physical plumbing, 00:00:46.170 --> 00:00:48.350 the blue -collar trades that actually power the 00:00:48.350 --> 00:00:50.310 whole machine. And that's what we're here to 00:00:50.310 --> 00:00:53.109 do. Strip away the hype. Welcome to the Deep 00:00:53.109 --> 00:00:55.509 Dive. Our mission today is to walk you through 00:00:55.509 --> 00:00:57.789 the economics of this industry using this six 00:00:57.789 --> 00:00:59.899 -tier framework. We're going to focus on where 00:00:59.899 --> 00:01:02.960 consistent long -term profit is really likely 00:01:02.960 --> 00:01:06.439 to flow in 2026 and beyond. To really get this, 00:01:06.480 --> 00:01:09.260 let's use the gold rush analogy. Historically, 00:01:09.319 --> 00:01:12.140 most miners went completely broke chasing gold. 00:01:12.299 --> 00:01:14.780 Yeah. The lasting wealth, it was built by the 00:01:14.780 --> 00:01:17.000 people who sold the necessities, the picks, the 00:01:17.000 --> 00:01:20.019 shovels, the jeans. We're here to identify the 00:01:20.019 --> 00:01:22.879 modern picks and shovels for the AI boom. We're 00:01:22.879 --> 00:01:24.420 going to walk you right up the entire stack. 00:01:24.560 --> 00:01:26.750 We'll start at tier zero. the absolute foundation 00:01:26.750 --> 00:01:31.269 and climb all the way to tier five that uh overcrowded 00:01:31.269 --> 00:01:33.930 hype -fueled penthouse and as we climb you'll 00:01:33.930 --> 00:01:36.290 see exactly why the odds of building something 00:01:36.290 --> 00:01:40.209 that lasts diminish significantly okay so let's 00:01:40.209 --> 00:01:42.090 unpack this starting at the lowest layer the 00:01:42.090 --> 00:01:44.530 bedrock this is tier zero energy infrastructure 00:01:44.530 --> 00:01:46.930 right now this is not a sexy topic it's slow 00:01:46.930 --> 00:01:49.950 it's complicated kind of boring and it is utterly 00:01:49.950 --> 00:01:53.239 essential yeah because at its root AI runs on 00:01:53.239 --> 00:01:55.579 immense constant power. Immense is the right 00:01:55.579 --> 00:01:57.439 word. This is where the massive demand shock 00:01:57.439 --> 00:01:59.540 is hitting the U .S. economy the hardest. The 00:01:59.540 --> 00:02:02.359 U .S. power grid is, frankly, it's aging. It 00:02:02.359 --> 00:02:04.700 is. It was designed for, you know, residential 00:02:04.700 --> 00:02:07.500 sprawl and some factories, not for the endless 00:02:07.500 --> 00:02:10.139 needs of massive data centers. A single one of 00:02:10.139 --> 00:02:12.240 these can use as much power as a small city. 00:02:12.400 --> 00:02:14.159 When you look at the numbers, it's just shocking. 00:02:14.400 --> 00:02:17.039 S &P Global noted that U .S. data centers are 00:02:17.039 --> 00:02:19.719 going to need 22 percent more power by the end 00:02:19.719 --> 00:02:22.439 of 2025. Just by next year. Just by next year. 00:02:22.479 --> 00:02:26.199 And by 2030, consumption will nearly triple. 00:02:26.599 --> 00:02:29.139 That's a crisis for utilities, but it's a staggering 00:02:29.139 --> 00:02:31.159 opportunity for anyone who can help solve it. 00:02:31.300 --> 00:02:33.819 So where does the opportunity hide? Well, it's 00:02:33.819 --> 00:02:36.180 in the capacity upgrades. Yeah. Electrical contracting 00:02:36.180 --> 00:02:38.319 companies are getting just flooded with work. 00:02:38.400 --> 00:02:40.840 I bet. And because these data centers demand 00:02:40.840 --> 00:02:44.710 redundancy. They absolutely cannot fail. There's 00:02:44.710 --> 00:02:47.409 this huge demand for energy storage. Yeah. We're 00:02:47.409 --> 00:02:49.770 talking high capacity batteries, backup generators 00:02:49.770 --> 00:02:52.710 and something called a UPS, an uninterruptible 00:02:52.710 --> 00:02:55.330 power supply. A business built on zero downtime. 00:02:55.610 --> 00:02:58.030 Exactly. And we can't overlook the sustainability 00:02:58.030 --> 00:03:00.270 pressure either. Right. Because they use so much 00:03:00.270 --> 00:03:02.590 power, they get a ton of pressure to offset their 00:03:02.590 --> 00:03:05.810 carbon footprint. 100 percent. That means massive, 00:03:05.889 --> 00:03:10.129 specialized solar and wind farms built specifically 00:03:10.129 --> 00:03:14.050 to feed. these new facilities it's also creating 00:03:14.050 --> 00:03:16.909 opportunity in grid management software you know 00:03:16.909 --> 00:03:19.770 helping utilities manage this surging unpredictable 00:03:19.770 --> 00:03:22.810 load so it's not just about plugging in it's 00:03:22.810 --> 00:03:25.050 about rebuilding the whole system the entire 00:03:25.050 --> 00:03:27.590 system and all that power ultimately feeds the 00:03:27.590 --> 00:03:31.419 brain which brings us up to tier one chips and 00:03:31.419 --> 00:03:34.000 manufacturing. Yeah. The industrial machine that's 00:03:34.000 --> 00:03:35.879 churning out the GPUs. You know the players. 00:03:36.120 --> 00:03:38.219 Yeah. NVIDIA, for instance, controls something 00:03:38.219 --> 00:03:41.580 like 92 % of the GPU market right now. It's massive 00:03:41.580 --> 00:03:43.500 leverage if you're one of those giants. But this 00:03:43.500 --> 00:03:45.400 isn't a realistic playground for most people. 00:03:45.539 --> 00:03:48.039 I mean, building a semiconductor fab costs billions 00:03:48.039 --> 00:03:50.879 of dollars, takes years, and it's just insanely 00:03:50.879 --> 00:03:53.419 complex. So if direct manufacturing is locked 00:03:53.419 --> 00:03:56.080 down, how do we sell the picks and shovels here? 00:03:56.280 --> 00:03:58.139 You serve the factories. You serve the supply 00:03:58.139 --> 00:04:00.240 chain and the infrastructure around them. Think 00:04:00.240 --> 00:04:03.180 about clean room construction, incredibly specialized 00:04:03.180 --> 00:04:06.659 sterile building processes. Or think about logistics, 00:04:07.000 --> 00:04:10.479 moving delicate, priceless equipment. You might 00:04:10.479 --> 00:04:12.740 only do a few shipments a year, but each one 00:04:12.740 --> 00:04:15.259 is worth millions and needs extreme precision. 00:04:15.539 --> 00:04:17.779 And quality assurance, right, testing those parts 00:04:17.779 --> 00:04:19.959 before they even get integrated. Exactly. These 00:04:19.959 --> 00:04:22.079 are boring, high -margin, specialized services 00:04:22.079 --> 00:04:26.269 that the behemoths absolutely rely on. I still 00:04:26.269 --> 00:04:28.389 wrestle with this, though. I have to admit, the 00:04:28.389 --> 00:04:30.910 tier one chip, this abstract piece of silicon 00:04:30.910 --> 00:04:34.379 versus... the concrete measurable power demand 00:04:34.379 --> 00:04:36.959 of tier zero, you can almost feel the heat from 00:04:36.959 --> 00:04:39.620 the grid, the strain. It just feels more urgent. 00:04:39.839 --> 00:04:42.819 The tier zero demand is immediate and physical. 00:04:42.959 --> 00:04:45.060 We talked about the capital flowing in. But if 00:04:45.060 --> 00:04:47.279 the grid is so strained, what's the single biggest 00:04:47.279 --> 00:04:49.360 bottleneck that money can't instantly solve? 00:04:49.579 --> 00:04:51.819 It's the skilled labor shortage. Our sources 00:04:51.819 --> 00:04:55.459 show the U .S. needs 140 ,000 more skilled electricians 00:04:55.459 --> 00:04:57.899 just to meet demand by 2030. That's the real 00:04:57.899 --> 00:05:00.019 bottleneck. And that labor shortage leads us 00:05:00.019 --> 00:05:02.360 perfectly to the core opportunity. This is tier 00:05:02.360 --> 00:05:04.819 two, data centers and physical infrastructure. 00:05:05.040 --> 00:05:07.439 The real sweet spot. This is what we're calling 00:05:07.439 --> 00:05:10.199 the real sweet spot. It's the most accessible 00:05:10.199 --> 00:05:13.540 and durable opportunity for people looking for 00:05:13.540 --> 00:05:16.220 a high moat trade. We talk about the cloud like 00:05:16.220 --> 00:05:18.439 it's some mystical thing, but AI doesn't live 00:05:18.439 --> 00:05:22.019 in the sky. It lives in these massive, hot, noisy 00:05:22.019 --> 00:05:25.000 warehouses, these hyperscale facilities. And 00:05:25.000 --> 00:05:27.480 they are being built everywhere right now. Driving 00:05:27.480 --> 00:05:30.420 constant demand. This tier is all about its physical 00:05:30.420 --> 00:05:33.300 defensibility. No software update is going to 00:05:33.300 --> 00:05:35.720 destroy your contract to fix a cooling system. 00:05:35.800 --> 00:05:38.740 Never. So let's talk about the needs driven by 00:05:38.740 --> 00:05:41.680 that extreme heat. HEAC and cooling systems are 00:05:41.680 --> 00:05:44.180 just these silent foundational moneymakers. If 00:05:44.180 --> 00:05:47.480 a cooling system fails for just 10 minutes, millions 00:05:47.480 --> 00:05:50.600 in hardware can literally melt. This is a high 00:05:50.600 --> 00:05:53.879 -mote, trade -based business. If you run an industrial 00:05:53.879 --> 00:05:56.939 HEAC company, pivoting to data centers is, well, 00:05:56.980 --> 00:05:59.120 it's a must. Which means high risk, but very 00:05:59.120 --> 00:06:01.660 high return. Absolutely. And electrical insulation 00:06:01.660 --> 00:06:04.639 is the same story. Your normal residential electrician 00:06:04.639 --> 00:06:07.019 can't touch the high voltage, redundant systems 00:06:07.019 --> 00:06:09.300 these buildings need. Specialized contractors 00:06:09.300 --> 00:06:11.420 are booked solid and they're commanding premium 00:06:11.420 --> 00:06:14.399 rates. And once it's built, there's the constant 00:06:14.399 --> 00:06:18.579 grinding upkeep. The facilities management? It 00:06:18.579 --> 00:06:21.699 never, ever stops. We're talking plumbing, cooling 00:06:21.699 --> 00:06:24.879 maintenance, and even incredibly focused cleaning. 00:06:25.279 --> 00:06:27.420 The cleaning part is fascinating. Why cleaning? 00:06:27.680 --> 00:06:31.519 Because dust, even microscopic dust, can settle 00:06:31.519 --> 00:06:34.060 on server components, cause them to overheat, 00:06:34.060 --> 00:06:36.959 and eventually just destroy them. Wow. Yeah, 00:06:36.959 --> 00:06:38.660 look at a company like Promera, which used to 00:06:38.660 --> 00:06:41.699 be DataClean. They specialize in this. They now 00:06:41.699 --> 00:06:43.879 service over 100 million square feet globally. 00:06:44.379 --> 00:06:47.300 They have this deep, specialized knowledge of 00:06:47.300 --> 00:06:49.839 how to maintain these extreme environments. And 00:06:49.839 --> 00:06:51.959 then there's just the sheer scale of the construction 00:06:51.959 --> 00:06:54.120 itself. It's not just a big building. It's a 00:06:54.120 --> 00:06:57.699 nonstop, hyper -accelerated, giant project. Turner 00:06:57.699 --> 00:07:00.279 Construction reported a 43 % revenue increase, 00:07:00.579 --> 00:07:03.759 driven almost entirely by these hyperscale projects. 00:07:04.160 --> 00:07:06.379 So we're talking specialized structural work, 00:07:06.600 --> 00:07:09.240 fire suppression, security. And intensive site 00:07:09.240 --> 00:07:12.199 prep, heavy groundwork, getting foundations stable 00:07:12.199 --> 00:07:14.199 enough for these enormous... sensitive machines. 00:07:14.500 --> 00:07:17.139 And here's the beauty of tier two, the trade 00:07:17.139 --> 00:07:19.740 skill advantage. You do not need a computer science 00:07:19.740 --> 00:07:22.399 degree. Not at all. Success here needs commercial 00:07:22.399 --> 00:07:25.720 driving licenses, specialized HVAC certification, 00:07:26.139 --> 00:07:29.470 electrical licenses, welding expertise. Skills 00:07:29.470 --> 00:07:32.449 that take months, not years, to learn. And right 00:07:32.449 --> 00:07:34.829 now, they're worth far more than many entry -level 00:07:34.829 --> 00:07:37.269 tech jobs. While entry -level coders are fighting 00:07:37.269 --> 00:07:39.970 over saturated roles in tech hubs, specialized 00:07:39.970 --> 00:07:43.089 tradespeople are signing high -figure, long -term 00:07:43.089 --> 00:07:46.350 contracts. The market is desperate for them. 00:07:46.490 --> 00:07:49.750 Whoa. Just imagine scaling a physical maintenance 00:07:49.750 --> 00:07:52.050 business based only on the exponential increase 00:07:52.050 --> 00:07:54.750 in power consumption across a single state. The 00:07:54.750 --> 00:07:56.670 sheer scale of that build -out is incredible. 00:07:56.910 --> 00:08:00.290 And you can see how T0, the power, and T2, the 00:08:00.290 --> 00:08:02.790 building, are completely interconnected. It's 00:08:02.790 --> 00:08:05.689 a physical feedback loop of demand. Which brings 00:08:05.689 --> 00:08:07.750 us to the real point. What makes these physical 00:08:07.750 --> 00:08:10.050 businesses in tier two so much safer than building 00:08:10.050 --> 00:08:12.730 software? It's the physical defensibility. No 00:08:12.730 --> 00:08:15.870 AI update or open source model can suddenly copy 00:08:15.870 --> 00:08:17.990 or destroy this work. Okay, let's move up the 00:08:17.990 --> 00:08:19.709 stack past the foundation. We hit tier three, 00:08:19.810 --> 00:08:22.350 foundation models. Or as the sources accurately 00:08:22.350 --> 00:08:24.810 label it, the billionaire's playground. Yeah. 00:08:24.930 --> 00:08:27.829 Simply put, you can't play here unless you are 00:08:27.829 --> 00:08:31.170 already a giant. This is the realm of creating 00:08:31.170 --> 00:08:35.129 the core. AI brains, your GPTs, your Geminis, 00:08:35.149 --> 00:08:37.250 your Claudes. And the capital required is just 00:08:37.250 --> 00:08:40.049 immense. Billions. You need billions in capital, 00:08:40.210 --> 00:08:43.730 thousands of top tier GPUs, and an army of PhD 00:08:43.730 --> 00:08:46.870 researchers. And the GPU supply is not exactly 00:08:46.870 --> 00:08:49.789 a free market, is it? No. It's entirely constrained. 00:08:50.090 --> 00:08:52.409 The supply goes straight to a handful of giants, 00:08:52.509 --> 00:08:55.710 Google, Microsoft, Meta. This leads to what's 00:08:55.710 --> 00:08:57.830 called circular financing, where they fund startups. 00:08:58.299 --> 00:09:00.500 but then demand those startups spend the money 00:09:00.500 --> 00:09:03.279 immediately on their own GPU clouds. It keeps 00:09:03.279 --> 00:09:05.279 the money in the club. We saw that case study 00:09:05.279 --> 00:09:07.139 from Elizabeth Jin of Hustle Fund, which really 00:09:07.139 --> 00:09:09.039 put the burn rate into perspective. Oh, yeah. 00:09:09.100 --> 00:09:11.879 She was seeing Series A AI companies burning 00:09:11.879 --> 00:09:14.419 half a million dollars a month. Sure. Per month, 00:09:14.500 --> 00:09:17.259 just on GPU infrastructure. With zero profit 00:09:17.259 --> 00:09:19.000 margin. Zero. They're basically just paying rent 00:09:19.000 --> 00:09:21.080 on the playground. This room is locked and the 00:09:21.080 --> 00:09:24.000 key is held by like four companies. So we pivot 00:09:24.000 --> 00:09:27.080 our attention to Tier 4, orchestration and tools. 00:09:27.549 --> 00:09:29.610 This is the technical middle ground. And this 00:09:29.610 --> 00:09:31.710 is where technical founders can find some real 00:09:31.710 --> 00:09:34.429 leverage. This is where you take those raw models 00:09:34.429 --> 00:09:36.870 from tier three and you make them useful for 00:09:36.870 --> 00:09:39.590 a business. So building the wiring and the skeleton 00:09:39.590 --> 00:09:42.669 for the AI brain. Exactly. Vector databases, 00:09:42.990 --> 00:09:45.389 workflow automation, prompt engineering tools, 00:09:45.690 --> 00:09:48.950 security compliance guardrails. Companies like 00:09:48.950 --> 00:09:51.309 Langchain and Zapier live right here. This sounds 00:09:51.309 --> 00:09:55.330 like classic B2B software. stable, high margin 00:09:55.330 --> 00:09:58.029 if you do it right. It is. If you have coding 00:09:58.029 --> 00:10:00.250 skills and you actually understand how a real 00:10:00.250 --> 00:10:02.769 world business operates, compliance, integration, 00:10:03.070 --> 00:10:05.389 all that, you can build tools that customers 00:10:05.389 --> 00:10:08.029 will happily pay for. The margins are way higher 00:10:08.029 --> 00:10:10.309 than the consumer app layer. OK, so what's the 00:10:10.309 --> 00:10:12.490 catch for a tier four founder? It seems relatively 00:10:12.490 --> 00:10:15.730 stable. Speed. Speed is everything. This space 00:10:15.730 --> 00:10:18.870 moves incredibly fast. Competitors and free open 00:10:18.870 --> 00:10:21.090 source alternatives pop up almost instantly. 00:10:21.309 --> 00:10:24.169 Success requires not just tech skill, but deep, 00:10:24.269 --> 00:10:26.889 deep industry vertical knowledge to solve a very 00:10:26.889 --> 00:10:29.470 precise, urgent problem. So if tier three is 00:10:29.470 --> 00:10:31.950 locked down by GPUs, what's the core asset tier 00:10:31.950 --> 00:10:35.120 four founders need to survive? Speed, intense 00:10:35.120 --> 00:10:37.919 focus, and a precise understanding of industry 00:10:37.919 --> 00:10:42.000 -specific pain points. And finally, we arrive 00:10:42.000 --> 00:10:44.740 at the top floor, tier five, the applications. 00:10:45.139 --> 00:10:49.639 This is the penthouse where 99 % of people try 00:10:49.639 --> 00:10:52.399 to build. And where sources suggest 99 % will 00:10:52.399 --> 00:10:54.700 ultimately fail. Yep. This is your generic writing 00:10:54.700 --> 00:10:57.919 tools, your image generators, your chat GPT wrappers. 00:10:58.440 --> 00:11:01.139 I was really fascinated by the analysis of why 00:11:01.139 --> 00:11:05.019 apps fail up here. What was the first key failure 00:11:05.019 --> 00:11:07.460 mechanism they pointed out? Commoditization. 00:11:07.480 --> 00:11:09.519 It hits like a freight train. Just think about 00:11:09.519 --> 00:11:12.000 AI transcription services. Companies were making 00:11:12.000 --> 00:11:14.379 good money on this until OpenAI released Whisper 00:11:14.379 --> 00:11:16.879 for free. Half the market just got wiped out 00:11:16.879 --> 00:11:18.940 overnight. The feature became standard. And second, 00:11:19.100 --> 00:11:21.860 the lack of a moat. Right. Most apps are just 00:11:21.860 --> 00:11:24.019 thin wrappers around existing models you can 00:11:24.019 --> 00:11:26.480 already access directly. Replication is ridiculously 00:11:26.480 --> 00:11:29.379 easy. Why would you pay $30 a month for an app 00:11:29.379 --> 00:11:30.919 when you can often do the same thing for free 00:11:30.919 --> 00:11:33.059 with the base model? And finally, the battle 00:11:33.059 --> 00:11:36.159 for attention is just brutal. Oh, it's an ocean 00:11:36.159 --> 00:11:39.100 of tools. To stand out, you need a massive marketing 00:11:39.100 --> 00:11:42.000 spend. If you don't have a giant ad budget or 00:11:42.000 --> 00:11:45.100 a pre -existing audience, most apps just drown. 00:11:45.279 --> 00:11:48.340 It truly is hard mode building here. But there 00:11:48.340 --> 00:11:50.460 are those rare exceptions that do survive. What 00:11:50.460 --> 00:11:52.840 defines their success? They build for a tiny 00:11:52.840 --> 00:11:55.740 specific niche. AI for orthodontists analyzing 00:11:55.740 --> 00:11:58.720 x -rays has potential because it solves an acute 00:11:58.720 --> 00:12:01.100 problem. They also succeed if they own their 00:12:01.100 --> 00:12:03.879 own distribution or if they combine the AI with 00:12:03.879 --> 00:12:06.440 some kind of specialized human service or proprietary 00:12:06.440 --> 00:12:09.500 data. So the final judgment on Tier 5 is pretty 00:12:09.500 --> 00:12:12.820 clear. admire it from a distance, but focus your 00:12:12.820 --> 00:12:15.740 energy on the lower tiers. So beyond being niche, 00:12:15.899 --> 00:12:18.379 what's the single most vital competitive advantage 00:12:18.379 --> 00:12:20.779 for a tier five startup that actually makes it? 00:12:20.860 --> 00:12:23.279 Speed. They have to launch, learn, and pivot 00:12:23.279 --> 00:12:25.299 faster than the big companies can react and just 00:12:25.299 --> 00:12:27.679 absorb their ideas. So let's bring this all back 00:12:27.679 --> 00:12:30.259 down to the ground. The smartest money, the most 00:12:30.259 --> 00:12:32.720 rational money, it flows toward the path with 00:12:32.720 --> 00:12:35.240 the highest odds of long -term success, not just 00:12:35.240 --> 00:12:37.460 the path with the loudest hype. And we've established 00:12:37.460 --> 00:12:39.740 a pretty clear hierarchy of opportunity for you. 00:12:39.789 --> 00:12:42.049 you, the person who isn't trying to raise VC 00:12:42.049 --> 00:12:45.070 money or get a PhD in machine learning. Right. 00:12:45.230 --> 00:12:48.070 Tier two data centers stands out as the most 00:12:48.070 --> 00:12:51.190 accessible and durable opportunity. Physical 00:12:51.190 --> 00:12:53.850 businesses are hard to copy. The worker shortage 00:12:53.850 --> 00:12:56.250 is critical and the contracts are for years, 00:12:56.389 --> 00:13:00.789 not months. And tier zero energy is that strong, 00:13:00.830 --> 00:13:04.590 long term patient play. Regardless of which chat 00:13:04.590 --> 00:13:07.649 bot wins, the grid has to be rebuilt and governments 00:13:07.649 --> 00:13:09.759 are pouring money into that. And then tier four 00:13:09.759 --> 00:13:11.320 is where the technical builders can find their 00:13:11.320 --> 00:13:13.539 leverage. It all comes back to that picks and 00:13:13.539 --> 00:13:16.299 shovels concept. We're profiting from the gold 00:13:16.299 --> 00:13:18.279 rush without actually having to bet on which 00:13:18.279 --> 00:13:21.360 miner finds the gold. Apps come and go with incredible 00:13:21.360 --> 00:13:23.820 speed, but the physical infrastructure stays. 00:13:24.080 --> 00:13:27.100 Think about the original gold rush. Levi Strauss 00:13:27.100 --> 00:13:30.000 sold durable denim pants to the miners. And built 00:13:30.000 --> 00:13:32.299 a company that's still compounding wealth 175 00:13:32.299 --> 00:13:35.159 years later. Exactly. The backbone businesses 00:13:35.159 --> 00:13:37.639 are durable because everyone needs the same foundation. 00:13:38.110 --> 00:13:40.029 The conclusion here is that the real money in 00:13:40.029 --> 00:13:42.590 AI is in the foundation, the physical plumbing, 00:13:42.750 --> 00:13:45.490 and the power. It's not really in the algorithms 00:13:45.490 --> 00:13:48.049 themselves. While the loudest tech arguments 00:13:48.049 --> 00:13:50.769 rage on social media about which model is smarter, 00:13:51.049 --> 00:13:53.710 the electrical contractor who just signed a three 00:13:53.710 --> 00:13:55.669 -year maintenance deal with a hyperscale server 00:13:55.669 --> 00:13:58.549 farm That's the one actually building generational 00:13:58.549 --> 00:14:01.570 wealth. The AI future, this trillion dollar build 00:14:01.570 --> 00:14:04.169 out, is being constructed with specialized wrenches 00:14:04.169 --> 00:14:06.730 and welding corches, not just algorithms. And 00:14:06.730 --> 00:14:09.029 that leaves you with a thought to consider. When 00:14:09.029 --> 00:14:11.769 you look at your own skill set, what physical 00:14:11.769 --> 00:14:14.250 high moat skill could you acquire that artificial 00:14:14.250 --> 00:14:16.309 intelligence can never truly replicate?