WEBVTT 00:00:00.000 --> 00:00:02.080 We think of artificial intelligence as living 00:00:02.080 --> 00:00:05.839 in the cloud, invisible, weightless. But what 00:00:05.839 --> 00:00:08.140 if the biggest threat to this revolution isn't 00:00:08.140 --> 00:00:12.179 missing code, but a lack of water? Yeah, I mean, 00:00:12.199 --> 00:00:14.519 it's pretty wild to think about. Beat, welcome 00:00:14.519 --> 00:00:17.940 to the deep dive. Today, we are ripping the invisible 00:00:17.940 --> 00:00:20.820 label off the cloud. We are looking at the massive 00:00:20.820 --> 00:00:23.800 physical reality of the AI gold rush. We are 00:00:23.800 --> 00:00:26.980 unpacking the hidden hardware costs. It's definitely 00:00:26.980 --> 00:00:29.440 not just software anymore. Right. First, we'll 00:00:29.440 --> 00:00:31.879 look at SpaceX. They just issued a bizarre new 00:00:31.879 --> 00:00:34.259 warning to the SEC regarding local plumbing. 00:00:34.420 --> 00:00:37.280 Which is just like so unexpected. It really is. 00:00:37.340 --> 00:00:39.960 Then we scale up. We will examine the trillion 00:00:39.960 --> 00:00:42.399 dollar corporate arms race happening right now. 00:00:42.500 --> 00:00:44.759 Oh, yeah. And finally, we look at NVIDIA. They 00:00:44.759 --> 00:00:47.500 have a massive new plan to bypass the cloud completely. 00:00:47.820 --> 00:00:50.320 They want to put the AI directly on your desk. 00:00:50.500 --> 00:00:52.679 It's a fascinating shift in the landscape. We 00:00:52.679 --> 00:00:55.020 are moving away from abstract software. We are 00:00:55.020 --> 00:00:57.659 entering a world of concrete, copper and cooling 00:00:57.659 --> 00:00:59.840 pipes. Yeah. Let's start with those physical 00:00:59.840 --> 00:01:02.159 constraints. This brings us directly to SpaceX. 00:01:02.500 --> 00:01:05.859 Now, SpaceX also includes Elon Musk's XAI division. 00:01:06.260 --> 00:01:09.579 Right, the AI brand. Yeah, exactly. They recently 00:01:09.579 --> 00:01:11.780 warned Wall Street investors about a totally 00:01:11.780 --> 00:01:15.579 unexpected risk. Access to clean municipal water 00:01:15.579 --> 00:01:18.519 is now a major risk factor for growth. It sounds 00:01:18.519 --> 00:01:20.620 totally absurd at first glance. You think about 00:01:20.620 --> 00:01:23.540 rocket launches and, you know, cutting edge code. 00:01:23.659 --> 00:01:26.159 You don't usually think about local water shortages. 00:01:27.469 --> 00:01:30.689 But securing water is now deemed absolutely critical. 00:01:30.969 --> 00:01:33.629 Exactly. It is right up there with locking down 00:01:33.629 --> 00:01:36.750 regional power grids. It is just as vital as 00:01:36.750 --> 00:01:39.989 hoarding the newest NVIDIA chips. I like to think 00:01:39.989 --> 00:01:42.769 of modern data centers like massive digital steam 00:01:42.769 --> 00:01:45.069 engines. Oh, that's a great analogy. Just like 00:01:45.069 --> 00:01:47.650 a 19th century locomotive needed physical water 00:01:47.650 --> 00:01:51.069 to create steam, these massive AI brains generate 00:01:51.069 --> 00:01:54.150 intense friction. They generate immense heat. 00:01:54.349 --> 00:01:56.650 Right. If you don't literally dump millions of 00:01:56.650 --> 00:01:59.810 gallons of water on them, the engine melts. Local 00:01:59.810 --> 00:02:02.049 plumbing has become the ultimate bottleneck. 00:02:02.069 --> 00:02:04.219 Absolutely. We have to understand why these new 00:02:04.219 --> 00:02:06.560 servers are different. How exactly does a lack 00:02:06.560 --> 00:02:08.919 of water change the physical architecture of 00:02:08.919 --> 00:02:11.780 these companies? Well, traditional data centers 00:02:11.780 --> 00:02:14.939 just host websites or email. They run on standard 00:02:14.939 --> 00:02:18.599 processors. But AI requires parallel processing. 00:02:18.819 --> 00:02:21.340 Okay. They use massive arrays of graphics cards 00:02:21.340 --> 00:02:24.750 running simultaneously. The power density in 00:02:24.750 --> 00:02:27.310 one single server rack is unprecedented. So they 00:02:27.310 --> 00:02:30.330 run incredibly hot. Insanely hot. Water is simply 00:02:30.330 --> 00:02:32.569 the cheapest, most effective way to cool them 00:02:32.569 --> 00:02:35.930 down. Site selection is literally dictating technology 00:02:35.930 --> 00:02:39.189 right now. Wow. Water availability decides where 00:02:39.189 --> 00:02:41.189 these massive new data centers are built. And 00:02:41.189 --> 00:02:43.490 the SEC is currently combing through this new 00:02:43.490 --> 00:02:46.270 filing. We don't know if regulators force this 00:02:46.270 --> 00:02:49.409 specific environmental disclosure, but the entire 00:02:49.409 --> 00:02:52.110 data center industry is facing a massive environmental 00:02:52.110 --> 00:02:54.849 vibe check. Yeah, the filing itself is very explicit. 00:02:55.090 --> 00:02:57.590 Local droughts could easily throttle their data 00:02:57.590 --> 00:02:59.849 center capacity. Right. Sudden government regulations 00:02:59.849 --> 00:03:02.210 regarding water usage could do the exact same 00:03:02.210 --> 00:03:04.550 thing. Competition for municipal water might 00:03:04.550 --> 00:03:07.389 skyrocket operational costs overnight, too. Absolutely. 00:03:07.590 --> 00:03:09.729 And if the local water dries up, the backup plan 00:03:09.729 --> 00:03:12.710 is brutal. SpaceX warns they might be forced 00:03:12.710 --> 00:03:15.449 to pivot entirely. To what exactly? They would 00:03:15.449 --> 00:03:18.129 have to use alternative cooling techniques. This 00:03:18.129 --> 00:03:20.530 means moving to complex closed -loop systems. 00:03:20.750 --> 00:03:23.169 What does a closed -loop system actually entail? 00:03:23.409 --> 00:03:25.969 Instead of cycling cheap municipal water, you 00:03:25.969 --> 00:03:28.370 use chemical refrigerants. It is a completely 00:03:28.370 --> 00:03:30.930 sealed system. Which is way more expensive to 00:03:30.930 --> 00:03:33.270 build, right? Way more expensive and far less 00:03:33.270 --> 00:03:36.009 efficient to operate. You can't just pack servers 00:03:36.009 --> 00:03:38.150 tightly into an existing warehouse anymore. You 00:03:38.150 --> 00:03:40.960 have to spread them out. You have to design entirely 00:03:40.960 --> 00:03:44.500 new architectural footprints. You need massive 00:03:44.500 --> 00:03:47.159 new power supplies just to run the refrigeration 00:03:47.159 --> 00:03:50.039 compressors. Wow. The basic physical footprint 00:03:50.039 --> 00:03:53.960 of the company balloons in size. Also, as a quick 00:03:53.960 --> 00:03:56.759 financial side note, SpaceX is reserving up to 00:03:56.759 --> 00:04:00.479 5 % of its IPO stock. Ah, right. For employees 00:04:00.479 --> 00:04:03.039 and insider friends? Yeah. But back to the hardware, 00:04:03.199 --> 00:04:06.060 the costs are astronomical. So without cheap 00:04:06.060 --> 00:04:08.699 water, the cost of being smart becomes financially 00:04:08.699 --> 00:04:11.759 impossible. Right. And that rising cost changes 00:04:11.759 --> 00:04:14.740 the entire economic model of AI. It is a massive 00:04:14.740 --> 00:04:16.920 problem. Which brings us to a really critical 00:04:16.920 --> 00:04:20.399 transition. If physical water and space are such 00:04:20.399 --> 00:04:22.860 limiting factors, you'd think tech companies 00:04:22.860 --> 00:04:24.639 would slow down. You would definitely think so. 00:04:24.800 --> 00:04:28.000 Instead, they are doing the exact opposite. They 00:04:28.000 --> 00:04:30.720 are pushing the hardware to the absolute breaking 00:04:30.720 --> 00:04:33.730 point. The capital being poured into cloud AI 00:04:33.730 --> 00:04:36.870 right now is staggering. It really is an arms 00:04:36.870 --> 00:04:39.569 race right now. Alphabet is planning to raise 00:04:39.569 --> 00:04:43.339 $80 billion in debt. Wow. This is just to expand 00:04:43.339 --> 00:04:46.259 their physical infrastructure. Demand for Google's 00:04:46.259 --> 00:04:49.079 AI services is growing much faster than their 00:04:49.079 --> 00:04:51.939 supply. They actually expect their total AI -related 00:04:51.939 --> 00:04:55.980 CapEx to hit $190 billion. Yeah. CapEx just means 00:04:55.980 --> 00:04:58.600 money spent on physical corporate assets like 00:04:58.600 --> 00:05:02.339 equipment. Beat $190 billion this year alone. 00:05:02.500 --> 00:05:04.860 That is a staggering amount of hard currency. 00:05:05.079 --> 00:05:07.160 It is. What does that actually buy in the real 00:05:07.160 --> 00:05:09.519 world? Well, it buys acres of prime real estate 00:05:09.519 --> 00:05:12.629 near cheap power grids. It buys massive concrete 00:05:12.629 --> 00:05:15.430 bunkers. It buys private electrical substations. 00:05:15.430 --> 00:05:17.769 And thousands of miles of high -speed fiber optic 00:05:17.769 --> 00:05:20.430 cables. Exactly. And they aren't alone in this 00:05:20.430 --> 00:05:23.629 massive spending spree. Anthropic just confidentially 00:05:23.629 --> 00:05:25.889 filed to go public. Right. They did this days 00:05:25.889 --> 00:05:29.610 after hitting a near $1 trillion valuation. Yeah. 00:05:29.649 --> 00:05:32.910 And they just secured $65 billion in fresh funding. 00:05:33.329 --> 00:05:38.819 The AI IPO race is accelerating wildly. The software 00:05:38.819 --> 00:05:41.899 side is equally aggressive. We are seeing a race 00:05:41.899 --> 00:05:44.639 for the ultimate everything app. Microsoft isn't 00:05:44.639 --> 00:05:47.079 just building a single tool anymore. Right. They 00:05:47.079 --> 00:05:50.779 want to merge GitHub Copilot, their chat functions, 00:05:51.019 --> 00:05:53.839 co -work, and autopilot. They want one unified 00:05:53.839 --> 00:05:56.879 AI super app. running constantly in the background. 00:05:57.120 --> 00:05:59.480 OpenAI and X are racing toward the exact same 00:05:59.480 --> 00:06:02.360 goal. They want you inside a single, inescapable 00:06:02.360 --> 00:06:05.040 ecosystem. It's kind of wild. XAI just opened 00:06:05.040 --> 00:06:08.259 access to a model called GrokBuild 0 .1. It is 00:06:08.259 --> 00:06:10.920 a new agentic coding model. An agentic coding 00:06:10.920 --> 00:06:13.019 model is software that writes and fixes code 00:06:13.019 --> 00:06:15.379 independently. Yeah. It doesn't just answer questions. 00:06:15.420 --> 00:06:17.740 It can plan out entire projects. It can refactor 00:06:17.740 --> 00:06:20.259 complex architecture. To refactor just means 00:06:20.259 --> 00:06:22.399 restructuring existing code without changing 00:06:22.399 --> 00:06:25.519 its external behavior. Exactly. It iterates across 00:06:25.519 --> 00:06:29.399 complex coding tasks autonomously. And the pricing 00:06:29.399 --> 00:06:32.819 is the wild part. It starts at just $1 per million 00:06:32.819 --> 00:06:35.939 input tokens. Input tokens are the basic chunks 00:06:35.939 --> 00:06:39.139 of text an AI reads first. Right. And it's incredibly 00:06:39.139 --> 00:06:42.560 cheap for that profound level of autonomy. Developers 00:06:42.560 --> 00:06:44.980 can spin up hundreds of these agents simultaneously 00:06:44.980 --> 00:06:47.519 for pennies. Wait, let me push back on that for 00:06:47.519 --> 00:06:50.540 a second. It is cheap in terms of a software 00:06:50.540 --> 00:06:53.199 subscription. Sure. But didn't we just establish 00:06:53.199 --> 00:06:55.480 that the physical cooling costs are astronomical? 00:06:55.639 --> 00:06:59.680 If the software costs $1, who is eating the massive 00:06:59.680 --> 00:07:02.680 physical cost of the power and water? That is 00:07:02.680 --> 00:07:05.480 the trillion dollar question. Right now, the 00:07:05.480 --> 00:07:08.019 tech giants are subsidizing that compute cost. 00:07:08.199 --> 00:07:09.920 They are swallowing the massive infrastructure 00:07:09.920 --> 00:07:12.620 bills just to capture market share. They are 00:07:12.620 --> 00:07:14.300 betting that once everyone is hooked on these 00:07:14.300 --> 00:07:16.459 agents, they can raise prices later. The physical 00:07:16.459 --> 00:07:18.500 world is merging directly with the software world, 00:07:18.560 --> 00:07:21.519 too. OpenAI is now hiring engineers to build 00:07:21.519 --> 00:07:24.500 real physical robots. Yeah, they have open roles 00:07:24.500 --> 00:07:26.060 for electrical engineering. They were looking 00:07:26.060 --> 00:07:28.259 for actuator design. They want machine learning 00:07:28.259 --> 00:07:30.800 systems experts. They're even hiring 3D printing 00:07:30.800 --> 00:07:33.180 lab technicians. It is moving off the screen 00:07:33.180 --> 00:07:36.019 and into the physical room. Microsoft is pushing 00:07:36.019 --> 00:07:38.040 heavily into the physical health space, too. 00:07:38.120 --> 00:07:40.339 They just launched Copilot Health. Oh, right. 00:07:40.579 --> 00:07:43.959 It combines wearable hardware stats, lab reports, 00:07:44.279 --> 00:07:46.740 and traditional medical records. Everything lives 00:07:46.740 --> 00:07:48.959 in one place. They even partnered with Harvard 00:07:48.959 --> 00:07:51.959 Health to build this out securely. We are seeing 00:07:51.959 --> 00:07:54.879 deeply integrated daily tools pop up everywhere. 00:07:55.399 --> 00:07:57.600 Mina is a new meeting assistant. It doesn't just 00:07:57.600 --> 00:08:00.360 transcribe. No, it actively participates in meetings. 00:08:00.620 --> 00:08:03.980 It responds in real time, pulls context, and 00:08:03.980 --> 00:08:08.699 executes tasks. There is Social Echo 2 .02. It's 00:08:08.699 --> 00:08:11.399 an AI social media co -pilot that runs marketing 00:08:11.399 --> 00:08:14.170 campaigns autonomously. There's a lot. Two sec 00:08:14.170 --> 00:08:17.870 silence. Looking at this sheer volume of tools 00:08:17.870 --> 00:08:20.189 and money, I'll admit I still wrestle with prompt 00:08:20.189 --> 00:08:22.589 drift myself. It's overwhelming. It really is 00:08:22.589 --> 00:08:24.550 a lot to take in at once. And prompt drift is 00:08:24.550 --> 00:08:27.029 when an AI slowly starts ignoring your original 00:08:27.029 --> 00:08:29.529 instructions. You give it a task, and an hour 00:08:29.529 --> 00:08:31.250 later, it's doing something slightly different. 00:08:31.389 --> 00:08:33.809 It's super frustrating. It is. And beyond those 00:08:33.809 --> 00:08:36.809 small glitches, there is a darker side to this 00:08:36.809 --> 00:08:40.230 rapid expansion. We saw a video of an AI founder 00:08:40.230 --> 00:08:42.730 recently. Oh, yeah. He was controlling his entire 00:08:42.730 --> 00:08:45.549 computer with only his voice. The AI was navigating 00:08:45.549 --> 00:08:48.889 his desktop perfectly. Many are comparing his 00:08:48.889 --> 00:08:51.710 fluid setup to Steve Jobs unveiling the iPhone. 00:08:51.950 --> 00:08:55.350 It looks like absolute magic. But Magic has serious 00:08:55.350 --> 00:08:58.269 vulnerabilities. We also saw hackers exploit 00:08:58.269 --> 00:09:01.750 Meta's new AI support bot recently. Right. They 00:09:01.750 --> 00:09:04.309 manipulated the AI to bypass security protocols. 00:09:04.470 --> 00:09:07.509 They hijacked high -profile Instagram accounts. 00:09:07.769 --> 00:09:10.110 This included profiles linked directly to the 00:09:10.110 --> 00:09:12.929 Obama White House. There are real legal consequences 00:09:12.929 --> 00:09:15.070 emerging from these vulnerabilities right now. 00:09:15.289 --> 00:09:18.129 OpenAI is facing a first -of -its -kind lawsuit. 00:09:18.129 --> 00:09:20.289 This is a big one. It was filed by the state 00:09:20.289 --> 00:09:22.919 of Florida. The filing claims ChatGPT played 00:09:22.919 --> 00:09:25.320 a role in tragic events, including shootings, 00:09:25.639 --> 00:09:28.279 suicides, and harmful user behavior. Just to 00:09:28.279 --> 00:09:30.279 be clear to the listener, we're simply reporting 00:09:30.279 --> 00:09:32.480 the claims exactly as stated in the filing. We 00:09:32.480 --> 00:09:34.860 aren't taking a side here. Right. We remain completely 00:09:34.860 --> 00:09:37.559 impartial. But it is a major legal test for the 00:09:37.559 --> 00:09:40.419 industry. It is the defining tension of this 00:09:40.419 --> 00:09:43.440 entire era. They are building the airplane while 00:09:43.440 --> 00:09:46.440 flying it at supersonic speeds. The capital demands 00:09:46.440 --> 00:09:49.019 require them to push new features constantly. 00:09:49.259 --> 00:09:51.899 They need a return on that massive infrastructure 00:09:51.899 --> 00:09:54.700 investment we talked about. But the safety mechanisms 00:09:54.700 --> 00:09:57.799 are inherently reactive. They learn from mistakes 00:09:57.799 --> 00:10:00.659 after they happen. When the system is this deeply 00:10:00.659 --> 00:10:03.899 integrated into daily life, those mistakes carry 00:10:03.899 --> 00:10:08.120 severe real -world weight. So how do companies 00:10:08.120 --> 00:10:11.580 responsibly scale a $190 billion infrastructure 00:10:11.580 --> 00:10:14.799 when basic safety guard rails are still failing? 00:10:15.039 --> 00:10:17.200 Well, the uncomfortable truth is that they often 00:10:17.200 --> 00:10:20.200 don't scale it responsibly. The scale is simply 00:10:20.200 --> 00:10:22.840 moving faster than the security protocols can 00:10:22.840 --> 00:10:24.720 adapt. They rely on patching vulnerabilities 00:10:24.720 --> 00:10:27.970 after the fact. Yeah. But when an AI agent has 00:10:27.970 --> 00:10:30.330 access to your medical records or your company's 00:10:30.330 --> 00:10:32.610 server infrastructure, a post -incident patch 00:10:32.610 --> 00:10:34.929 isn't really good enough. Growth is outpacing 00:10:34.929 --> 00:10:37.289 safety, making the cloud a deeply unpredictable 00:10:37.289 --> 00:10:40.009 foundation. Exactly. It is a precarious balancing 00:10:40.009 --> 00:10:43.490 act. So cloud AI is demanding unsustainably massive 00:10:43.490 --> 00:10:46.169 data centers. We saw that clearly with the SpaceX 00:10:46.169 --> 00:10:48.889 water issues. Oh, absolutely. It is gobbling 00:10:48.889 --> 00:10:51.549 up. massive corporate budgets. The industry is 00:10:51.549 --> 00:10:53.730 desperate for a pressure release valve. And what 00:10:53.730 --> 00:10:56.169 a pivot it is. The NVIDIA thinks they have the 00:10:56.169 --> 00:10:59.389 exact solution. Move the AI out of the cloud 00:10:59.389 --> 00:11:01.929 completely. Put it directly on your physical 00:11:01.929 --> 00:11:04.389 desk. They opened the Computex conference with 00:11:04.389 --> 00:11:07.590 a massive hardware drop. NVIDIA is coming directly 00:11:07.590 --> 00:11:11.929 for the $200 billion desktop CPU market. They 00:11:11.929 --> 00:11:14.649 unveiled something entirely new called the RTX 00:11:14.649 --> 00:11:18.019 Spark. This is a wild piece of hardware. They're 00:11:18.019 --> 00:11:20.860 calling it a one -petaflop superchip. A petaflop 00:11:20.860 --> 00:11:23.720 is a speed of one quadrillion calculations per 00:11:23.720 --> 00:11:26.679 second. Whoa, just imagine having a one -petaflop 00:11:26.679 --> 00:11:28.919 superchip sitting on your desk. It's crazy. It's 00:11:28.919 --> 00:11:30.700 literal supercomputer power running silently 00:11:30.700 --> 00:11:32.820 right in your home office. It changes the paradigm 00:11:32.820 --> 00:11:35.820 entirely. It packs the CPU, the GPU, the RAM, 00:11:36.019 --> 00:11:39.200 and the CDA software together seamlessly. CDA 00:11:39.200 --> 00:11:41.559 software is NVIDIA's proprietary coding platform 00:11:41.559 --> 00:11:43.940 for parallel processing. The ultimate goal here 00:11:43.940 --> 00:11:48.480 is to run hefty local LLMs easily. An LLM is 00:11:48.480 --> 00:11:50.740 the underlying mathematical brain that powers 00:11:50.740 --> 00:11:53.940 modern chatbots. Right. NVIDIA wants to completely 00:11:53.940 --> 00:11:57.139 kill the traditional point and click. PC experience. 00:11:57.279 --> 00:11:59.759 Yeah. They want you talking to a local supercomputer 00:11:59.759 --> 00:12:02.399 instead. They teamed up directly with Microsoft 00:12:02.399 --> 00:12:05.440 for this massive rollout. They're building secure 00:12:05.440 --> 00:12:08.980 sandboxes right into the core of Windows. A sandbox 00:12:08.980 --> 00:12:12.200 is a secure, isolated digital quarantine zone 00:12:12.200 --> 00:12:14.700 on your computer. Right. It lets you run untrusted 00:12:14.700 --> 00:12:16.600 code without risking your entire hard drive. 00:12:16.759 --> 00:12:18.960 You can deploy local agents securely inside these 00:12:18.960 --> 00:12:21.220 sandboxes. They specifically mentioned running 00:12:21.220 --> 00:12:24.200 complex open source tools like OpenClaw or Hermes 00:12:24.200 --> 00:12:26.769 right on your desktop. No cloud connection required. 00:12:27.110 --> 00:12:29.690 Zero latency. Complete privacy. Heavyweights 00:12:29.690 --> 00:12:31.409 are already locked in for the big fall release. 00:12:31.710 --> 00:12:34.490 We are going to see new machines from Asus, Dell, 00:12:34.710 --> 00:12:37.470 HP, Lenovo, and Microsoft. Over 100 software 00:12:37.470 --> 00:12:39.909 makers are fully on board. That includes massive 00:12:39.909 --> 00:12:42.490 legacy players like Adobe. It also includes newer 00:12:42.490 --> 00:12:45.330 visual tools like Humph UI. NVIDIA is aggressively 00:12:45.330 --> 00:12:47.629 targeting the builder market here. If you build 00:12:47.629 --> 00:12:50.330 AI workflows, you know the physical reality of 00:12:50.330 --> 00:12:53.409 the space. The Mac Mini has been the undisputed 00:12:53.409 --> 00:12:56.289 king of local development. Apple's unified memory 00:12:56.289 --> 00:12:59.149 architecture is incredible for this stuff. It 00:12:59.149 --> 00:13:01.309 is currently the most affordable way to run heavy 00:13:01.309 --> 00:13:04.629 tools like OpenClaw smoothly. NVIDIA is clearly 00:13:04.629 --> 00:13:07.230 tired of watching Apple take that builder market 00:13:07.230 --> 00:13:09.830 unopposed. They want those developers back on 00:13:09.830 --> 00:13:12.710 Windows architecture. But there's a massive cashier. 00:13:13.000 --> 00:13:15.500 We have to look at their history. NVIDIA tried 00:13:15.500 --> 00:13:19.379 ARM -based Windows devices back in 2013. It was 00:13:19.379 --> 00:13:22.019 a massive, highly publicized failure. They lost 00:13:22.019 --> 00:13:25.580 $900 million on that specific flop. It was a 00:13:25.580 --> 00:13:27.580 brutal track record. The hardware just didn't 00:13:27.580 --> 00:13:30.259 translate into a good user experience. And the 00:13:30.259 --> 00:13:32.679 current pricing is a massive hurdle. NVIDIA currently 00:13:32.679 --> 00:13:36.240 sells a DGX Spark developer mini computer. It 00:13:36.240 --> 00:13:38.759 is meant to preview this new ecosystem. Yeah, 00:13:38.779 --> 00:13:41.840 and it costs around $4 ,800. Two sec silence. 00:13:42.350 --> 00:13:45.190 Will consumers actually drop nearly $5 ,000 on 00:13:45.190 --> 00:13:47.169 a Windows machine when NVIDIA's last attempt 00:13:47.169 --> 00:13:49.889 was a disaster? That is the ultimate test of 00:13:49.889 --> 00:13:53.490 this strategy. The Mac Mini is dominant precisely 00:13:53.490 --> 00:13:56.769 because it offers incredible efficiency at a 00:13:56.769 --> 00:13:59.210 very low price point. Developers are very practical 00:13:59.210 --> 00:14:02.210 people. They care about overhead. Exactly. If 00:14:02.210 --> 00:14:05.470 the new OEM machines, like the rumored Surface 00:14:05.470 --> 00:14:08.710 Laptop Ultra, sit at that premium $5 ,000 tier, 00:14:08.929 --> 00:14:12.139 they will absolutely struggle. Builders simply 00:14:12.139 --> 00:14:15.159 won't adopt a completely new workflow if it costs 00:14:15.159 --> 00:14:17.879 five times more than a proven, reliable Apple 00:14:17.879 --> 00:14:20.220 setup. If they miss on price, developers will 00:14:20.220 --> 00:14:22.480 just keep buying Apple. Right. It all comes down 00:14:22.480 --> 00:14:25.220 to basic accessibility. You can't start a revolution 00:14:25.220 --> 00:14:27.500 if nobody can afford the entry ticket. It's time 00:14:27.500 --> 00:14:29.399 to pull back and look at the big picture here. 00:14:29.860 --> 00:14:32.500 Artificial intelligence is fundamentally a physical 00:14:32.500 --> 00:14:35.080 industry. We talk about it like it's magic in 00:14:35.080 --> 00:14:37.629 the ether. But the code is just software. The 00:14:37.629 --> 00:14:39.750 actual battleground is hardware. We saw SpaceX 00:14:39.750 --> 00:14:42.149 fighting for local municipal water rights just 00:14:42.149 --> 00:14:44.409 to cool their servers. We saw Alphabet dropping 00:14:44.409 --> 00:14:48.070 $190 billion on physical infrastructure like 00:14:48.070 --> 00:14:51.049 land and copper. We see Nvidia trying to put 00:14:51.049 --> 00:14:53.409 a one petaflop super chip directly inside your 00:14:53.409 --> 00:14:56.649 house. The bottleneck for human progress isn't 00:14:56.649 --> 00:14:59.690 mathematics anymore. The bottleneck is physical 00:14:59.690 --> 00:15:03.090 reality. It's the physical limits of planet Earth. 00:15:03.330 --> 00:15:06.490 When you use an AI tool today, I want you to 00:15:06.490 --> 00:15:08.889 think about this. Think about the physical supply 00:15:08.889 --> 00:15:11.009 chain that made your instant answer possible. 00:15:11.450 --> 00:15:14.309 Think about the millions of gallons of water 00:15:14.309 --> 00:15:16.990 cooling those massive servers in the desert. 00:15:17.190 --> 00:15:19.909 Think about the local power grid straining under 00:15:19.909 --> 00:15:23.169 the load of a billion daily queries. It's wild 00:15:23.169 --> 00:15:26.149 to visualize. If water rights and localized desktop 00:15:26.149 --> 00:15:28.789 superchips dictate the next decade of technology, 00:15:29.169 --> 00:15:32.250 ask yourself this. Will the ultimate winner of 00:15:32.250 --> 00:15:34.970 the AI race be the visionary who writes the absolute 00:15:34.970 --> 00:15:37.809 best code? Right. Or will it be the corporation 00:15:37.809 --> 00:15:40.169 that quietly controls the physical plumbing of 00:15:40.169 --> 00:15:42.289 the planet? That is the big question. Two sec 00:15:42.289 --> 00:15:44.629 silence. Thank you for joining us on this deep 00:15:44.629 --> 00:15:47.009 dive. We appreciate your time, your focus, and 00:15:47.009 --> 00:15:49.309 your curiosity. Be sure to subscribe and return 00:15:49.309 --> 00:15:51.470 for our next conversation. Take care of yourselves.