WEBVTT 1 00:00:00.440 --> 00:00:05.919 Welcome to the debate. So imagine signing a multimillion dollar 2 00:00:05.960 --> 00:00:10.519 corporate IT contract. Okay, but you pay absolutely nothing in 3 00:00:10.599 --> 00:00:11.240 year one. 4 00:00:11.320 --> 00:00:13.400 Wait, nothing like zero zero. 5 00:00:14.439 --> 00:00:17.039 Your payment in here two depends entirely on whether the 6 00:00:17.160 --> 00:00:19.760 artificial intelligence actually saves you money. 7 00:00:20.079 --> 00:00:21.839 Oh right, the outcome based. 8 00:00:21.559 --> 00:00:26.600 Model, exactly, And that exact deal is happening right now 9 00:00:27.160 --> 00:00:30.559 in Germany. But while AI is driving the software costs 10 00:00:30.600 --> 00:00:34.280 to near zero, on the ground, it's simultaneously triggering this 11 00:00:34.960 --> 00:00:38.600 massive I mean, it's a two hundred and twenty billion 12 00:00:38.640 --> 00:00:42.399 dollar debt tsunami. Yeah, it's huge, and that threatens to 13 00:00:42.439 --> 00:00:45.200 hike the cost of borrowing for you know, the entire 14 00:00:45.479 --> 00:00:49.119 global economy. So the core question we're looking at today 15 00:00:49.240 --> 00:00:54.159 is are we entering an era of unprecedented deflationary productivity 16 00:00:54.520 --> 00:00:56.920 or are we just building an inflationary debt bomb. 17 00:00:57.119 --> 00:01:00.280 And the tension there is I mean, that is exactly 18 00:01:00.320 --> 00:01:03.320 what we are unpacking today. We're looking at this fundamental 19 00:01:03.520 --> 00:01:09.400 divergence in macroeconomic analysis. Right We're drawing on recent warnings 20 00:01:09.400 --> 00:01:13.159 from the Swiss National Bank and the IMF alongside the 21 00:01:13.200 --> 00:01:17.879 real time structural upheaval in India's IT services sector, which 22 00:01:17.959 --> 00:01:20.840 is what a three hundred and fifteen billion dollar industry. 23 00:01:21.000 --> 00:01:23.439 Yeah, three hundred and fifteen it's massive. 24 00:01:23.239 --> 00:01:24.920 Right, And of course we have to look at the 25 00:01:24.959 --> 00:01:27.959 record breaking surge in US corporate debt right now. 26 00:01:28.000 --> 00:01:31.280 Definitely, And so to lay out where I stand, my 27 00:01:31.439 --> 00:01:36.760 position is that AI is fundamentally a profound deflationary force. Okay, 28 00:01:36.840 --> 00:01:40.680 we are witnessing an unprecedented wave of productivity gains. That 29 00:01:40.920 --> 00:01:45.560 is well, it's already restructuring global service delivery. When you 30 00:01:45.640 --> 00:01:49.519 decouple human labor hours from economic output, you create this 31 00:01:49.760 --> 00:01:54.439 long term structural mechanism that drives costs down across the board. 32 00:01:54.599 --> 00:01:57.359 See, I think you're hyper focusing on the end product 33 00:01:57.359 --> 00:01:59.959 while completely ignoring what it takes to build the factory. 34 00:02:00.120 --> 00:02:03.959 How so well, I contend that the massive redirection of capital, 35 00:02:04.159 --> 00:02:07.640 the strain on physical resources, and just the sheer scale 36 00:02:07.640 --> 00:02:11.439 of debt required to actually build this AI infrastructure, it 37 00:02:11.479 --> 00:02:16.000 creates undeniable systemic inflationary pressures in the short to medium term. 38 00:02:16.080 --> 00:02:18.319 I hear that, But and the assumption that short term 39 00:02:18.360 --> 00:02:23.039 capital exhaustion just magically guarantees long term structural deflation that 40 00:02:23.199 --> 00:02:25.759 is a massive leap of faith that the data simply 41 00:02:25.800 --> 00:02:27.240 doesn't support right now. 42 00:02:27.039 --> 00:02:29.680 Well, let's break down exactly how this productivity engine is 43 00:02:29.680 --> 00:02:33.000 functioning in the real world, because it's not theoretical. Fair enough, 44 00:02:33.120 --> 00:02:36.080 Look at the Indian IT industry. For decades, this sector 45 00:02:36.120 --> 00:02:39.039 operated on a very specific mechanism, which was labor. 46 00:02:38.840 --> 00:02:40.960 Arbitrage, right hiring overseas. 47 00:02:41.479 --> 00:02:45.439 Exactly, you hired one thousand developers because paying someone in 48 00:02:45.520 --> 00:02:49.520 Bangalore was structurally cheaper than paying someone in New York. 49 00:02:50.000 --> 00:02:53.680 The business model was literally just selling human time by 50 00:02:53.719 --> 00:02:54.439 the hour. 51 00:02:54.680 --> 00:02:55.919 The billable hours model. 52 00:02:56.120 --> 00:02:59.960 Yeah right, But today that entire paradigm is being dismantled. 53 00:03:00.639 --> 00:03:04.639 Clients are breaking the traditional reliance on those billable. 54 00:03:04.240 --> 00:03:06.479 Hours because they want more for less. 55 00:03:06.960 --> 00:03:10.919 Exactly, they are demanding more for less. Major players are 56 00:03:10.960 --> 00:03:16.159 restructuring their business models toward measurable performance outcomes. So when 57 00:03:16.199 --> 00:03:19.479 you decouple human labor from output, meaning you know you 58 00:03:19.520 --> 00:03:22.879 can deliver five times the software analysis using an AI 59 00:03:22.960 --> 00:03:26.439 model without adding a single human headcount, you have achieved 60 00:03:26.479 --> 00:03:30.000 the literal textbook definition of structural productivity. 61 00:03:30.199 --> 00:03:31.080 I see what you're saying. 62 00:03:31.080 --> 00:03:34.479 But you lower the baseline cost of operations for the 63 00:03:34.520 --> 00:03:35.439 broader economy. 64 00:03:35.759 --> 00:03:40.280 That is deflationary, okay, but lowering a baseline cost in 65 00:03:40.400 --> 00:03:44.000 one specific sector does not equal macroeconomic deflation. 66 00:03:44.520 --> 00:03:47.360 Why not if it's a massive sector, because you're. 67 00:03:47.240 --> 00:03:51.599 Looking at downstream IT efficiency, which completely obscures the upstream 68 00:03:51.639 --> 00:03:54.439 macro costs of the AI buildout. We need to look 69 00:03:54.479 --> 00:03:57.280 closely at the framework laid out by Petrachudin from the 70 00:03:57.280 --> 00:03:58.919 Swiss National Banks Governing Board. 71 00:03:59.199 --> 00:04:01.039 Okay, yeah, Chudin's report. 72 00:04:00.759 --> 00:04:04.439 And also Savannah Tenrero at the IMF. Redirecting these massive 73 00:04:04.479 --> 00:04:07.960 global investment flows to build out AI, it creates immediate 74 00:04:08.039 --> 00:04:12.400 severe physical shortages. You mean like the chips and things, ships, Absolutely, 75 00:04:12.639 --> 00:04:16.560 but we're also talking about energy grids, cooling systems, specialized silicon. 76 00:04:16.920 --> 00:04:20.639 When you drain resources to fund one specific technological boom, 77 00:04:20.720 --> 00:04:24.199 prices for those underlying resources skyrocket. Well sure in the 78 00:04:24.199 --> 00:04:27.600 short term, but the capital markets are flashing red right now. 79 00:04:27.879 --> 00:04:31.199 By August of this year, AI hyperscalers. So the massive 80 00:04:31.240 --> 00:04:34.519 tech companies building this infrastructure, they issued a record two 81 00:04:34.680 --> 00:04:38.079 hundred and twenty billion dollars in corporate debt right that 82 00:04:38.160 --> 00:04:40.399 is up from just twelve point five billion in the 83 00:04:40.439 --> 00:04:41.600 same period last year. 84 00:04:41.879 --> 00:04:43.920 It's a big jump. I admit it's. 85 00:04:43.800 --> 00:04:47.639 An aggressive capital drain. And that strains credit markets, pushes 86 00:04:47.680 --> 00:04:51.439 up yields, and acts as a present undeniable inflationary force today, 87 00:04:51.680 --> 00:04:53.160 not in some theoretical future. 88 00:04:53.480 --> 00:04:56.360 But I think you are treating a temporary supply bottleneck 89 00:04:56.480 --> 00:04:58.319 like it's a permanent feature of the economy. 90 00:04:58.600 --> 00:05:01.079 Two hundred and twenty billion is a pretty big bottleneck. 91 00:05:01.360 --> 00:05:03.639 It is, But let's look at what that debt is 92 00:05:03.720 --> 00:05:06.879 actually buying and how it changes corporate contracts on the ground. 93 00:05:07.279 --> 00:05:09.680 You mentioned the downstream efficiency as if it's just a 94 00:05:09.720 --> 00:05:10.519 minor detail. 95 00:05:10.720 --> 00:05:11.800 I didn't say it was minor. 96 00:05:11.839 --> 00:05:16.680 I just it is the entire point. Historically, outsourcing was 97 00:05:16.759 --> 00:05:19.959 a numbers game. Now, as Jim at Arroya of the 98 00:05:20.040 --> 00:05:23.399 Everest Group points out, the leverage has completely flipped to 99 00:05:23.439 --> 00:05:26.639 the clients. Okay, they do not care how many human 100 00:05:26.680 --> 00:05:29.560 hours go into a project anymore. They are paying for 101 00:05:29.680 --> 00:05:31.800 algorithmic certainty. 102 00:05:31.399 --> 00:05:34.319 Which is exactly why the Nifty IT index tumbled by 103 00:05:34.360 --> 00:05:37.399 a fifth this year. Yes, it wiped out seventy three 104 00:05:37.519 --> 00:05:40.720 billion dollars in market value for its top ten constituents. 105 00:05:41.040 --> 00:05:44.519 The old human heavy model isn't just shifting, it's collapsing 106 00:05:44.600 --> 00:05:48.360 under its own weight, exactly because human labor is suddenly 107 00:05:48.439 --> 00:05:50.519 viewed as a liability, not an asset. 108 00:05:50.759 --> 00:05:54.120 It is collapsing and what is replacing it is vastly 109 00:05:54.199 --> 00:05:57.000 more efficient for the ends consumer. Let's look at the 110 00:05:57.000 --> 00:06:00.879 mechanics of these new deals, all right. Tcs's chief executive 111 00:06:00.959 --> 00:06:05.519 recently confirmed that roughly eighty percent of their finance, HR 112 00:06:05.720 --> 00:06:11.160 and business services contracts are now based on outcome performance measures. 113 00:06:11.480 --> 00:06:14.399 Eighty percent. Wow, yeah, that figure has. 114 00:06:14.279 --> 00:06:17.720 Doubled since late twenty twenty three. We are seeing deals 115 00:06:17.800 --> 00:06:21.480 like Cognizant partnering with Dailer Truck where they are literally 116 00:06:21.519 --> 00:06:25.519 splitting the AI related cost savings on the balance sheet, right, 117 00:06:25.920 --> 00:06:29.839 or HCl Tech's cloud management deal with Eon. That's the 118 00:06:29.879 --> 00:06:33.079 one I mentioned. At the start, HCl tech takes on 119 00:06:33.279 --> 00:06:36.319 all the upfront risk, they don't get paid until year 120 00:06:36.360 --> 00:06:40.279 two and only if the efficiency gains actually materialize. 121 00:06:40.399 --> 00:06:42.759 It's a bold strategy, it is. 122 00:06:43.160 --> 00:06:46.040 This isn't just a slight discount. This is a brutal, 123 00:06:46.360 --> 00:06:51.639 hyper efficient market mechanism that forces price reduction. The provider 124 00:06:51.680 --> 00:06:54.439 takes the risk, the AI does the heavy lifting, and 125 00:06:54.519 --> 00:06:56.839 the broader economy gets cheaper services. 126 00:06:56.920 --> 00:06:59.639 At the micro level. Sure Gamler Truck and Eon or 127 00:06:59.639 --> 00:07:02.920 outside sslutely saving money on their IT contracts. But let's 128 00:07:02.959 --> 00:07:08.839 apply Petrachudin's macroeconomic lens here again. Okay, she explicitly points 129 00:07:08.839 --> 00:07:11.920 out that productivity gains are not a new phenomenon. They 130 00:07:12.000 --> 00:07:15.879 happen regularly throughout history. True, but they do not automatically 131 00:07:16.000 --> 00:07:21.480 lead an economy into structural deflation. Deflation is a macroeconomic calculation. 132 00:07:21.800 --> 00:07:24.639 It is a rate of change, right. It's not a single. 133 00:07:24.319 --> 00:07:27.079 Event, right, It's annualized exactly. 134 00:07:27.639 --> 00:07:30.160 For AI to have a true deflationary effect on the 135 00:07:30.160 --> 00:07:33.480 global economy, those price declines would have to repeat themselves 136 00:07:33.560 --> 00:07:36.519 regularly on an annual basis across all sectors. 137 00:07:36.720 --> 00:07:39.240 But if inflation is the annualized rate of change, a 138 00:07:39.319 --> 00:07:43.480 permanent restructuring of how a service is delivered permanently alters 139 00:07:43.480 --> 00:07:44.240 that trajectory. 140 00:07:44.480 --> 00:07:48.240 It shifts the baseline once. A one time thirty percent 141 00:07:48.360 --> 00:07:52.160 drop in IT servicing costs is a shock to the system. Sure, 142 00:07:52.639 --> 00:07:55.600 but the following year, if that cost stays flat at 143 00:07:55.600 --> 00:07:59.040 the new lower level, inflation is zero not negative. 144 00:07:59.199 --> 00:08:00.600 Okay, point there. 145 00:08:00.720 --> 00:08:06.360 It doesn't create systemic compounding deflation. Meanwhile, the mechanism required 146 00:08:06.360 --> 00:08:08.199 to achieve that one time drop. 147 00:08:08.040 --> 00:08:10.279 Is just staggering the debt buildout. 148 00:08:10.120 --> 00:08:14.439 The investment flows. Yes, they are being aggressively redirected away 149 00:08:14.480 --> 00:08:17.560 from other productive sectors of the economy to build out 150 00:08:17.600 --> 00:08:21.720 server farms secure vast amounts of energy by silicon Right. 151 00:08:21.879 --> 00:08:26.199 Shootin warns this exact dynamic creates shortages and painful adjustments 152 00:08:26.240 --> 00:08:29.560 for the rest of the economy. The upward inflationary pressure 153 00:08:29.600 --> 00:08:33.720 of building the infrastructure entirely offsets the downstream software savings 154 00:08:33.720 --> 00:08:35.240 in the short and medium term. 155 00:08:35.440 --> 00:08:39.639 Okay, but you are evaluating this transition as a standard 156 00:08:39.679 --> 00:08:44.440 capital expenditure cycle rather than a fundamental we wiring of 157 00:08:44.519 --> 00:08:45.559 economic production. 158 00:08:46.159 --> 00:08:49.720 How is it different from any other major infrastructure build. 159 00:08:49.639 --> 00:08:52.440 Think about the introduction of the assembly line. But let's 160 00:08:52.519 --> 00:08:54.879 push the mechanics of that analogy a bit further. 161 00:08:55.240 --> 00:08:56.200 Okay, I'm listening. 162 00:08:56.519 --> 00:08:59.000 The assembly line didn't just lower the cost of a 163 00:08:59.080 --> 00:09:03.840 model T once it destroyed bloated legacy manufacturing models. True, 164 00:09:04.039 --> 00:09:07.360 but crucially, the assembly line still required you to pay 165 00:09:07.399 --> 00:09:09.919 a human and hourly wage to stand there and turn 166 00:09:09.919 --> 00:09:10.360 a wrench. 167 00:09:10.559 --> 00:09:10.759 Right. 168 00:09:11.120 --> 00:09:15.399 What TCS, Cognizant and eonor doing with outcome based pricing 169 00:09:15.759 --> 00:09:18.679 is the equivalent of buying the cars without having to 170 00:09:18.720 --> 00:09:22.240 pay for the factory floor time. They have entirely severed 171 00:09:22.320 --> 00:09:24.320 human time from economic value. 172 00:09:24.440 --> 00:09:25.960 I mean, that's a nice analogy, but. 173 00:09:26.159 --> 00:09:30.240 Wait, just look at the firms. Massive legacy firms like 174 00:09:30.320 --> 00:09:34.440 Infosis and whip Row are seeing subdued growth of you know, 175 00:09:34.600 --> 00:09:38.000 one to three percent because they're stuck in the old model. 176 00:09:38.440 --> 00:09:41.519 But look at the agile tier two firms right Persistent 177 00:09:41.600 --> 00:09:45.799 saw revenue serge sixteen percent in a single quarter. Coforge 178 00:09:45.799 --> 00:09:48.919 is sales jumped by a third. They are deploying AI 179 00:09:49.120 --> 00:09:52.600 for rabid pilot programs that automate millions of manual data 180 00:09:52.720 --> 00:09:53.279 entry hours. 181 00:09:53.360 --> 00:09:54.440 Yeah, they're moving fast. 182 00:09:54.679 --> 00:09:58.759 That brutal market efficiency is exactly the mechanism that guarantees 183 00:09:58.879 --> 00:10:03.360 compounding law long term price reductions across every industry that 184 00:10:03.480 --> 00:10:04.320 uses software. 185 00:10:04.399 --> 00:10:07.000 The assembly line analogy falls apart, though, when you look 186 00:10:07.000 --> 00:10:09.840 at the macro financial plumbing required to sustain it. How 187 00:10:10.200 --> 00:10:13.559 because the assembly line didn't require the entire global debt 188 00:10:13.559 --> 00:10:15.879 market to bend to its will before a single car 189 00:10:16.000 --> 00:10:16.879 rolled off the floor. 190 00:10:17.120 --> 00:10:19.200 That's a bit dramatic, don't you think? 191 00:10:19.720 --> 00:10:24.480 Not really, you cannot separate the software efficiencies you're praising 192 00:10:24.799 --> 00:10:28.840 from the hardware and energy realities required to run them. 193 00:10:29.159 --> 00:10:33.320 The hyperscalers, you know, the Amazons, Googles, Microsofts. They are 194 00:10:33.559 --> 00:10:38.039 triggering a debt surge that is testing the absolute structural 195 00:10:38.159 --> 00:10:39.679 limits of investor demand. 196 00:10:40.120 --> 00:10:41.159 It's a lot of debt. 197 00:10:41.399 --> 00:10:45.279 Yes, but let's examine Amazon's recent long dated bond sale. 198 00:10:45.360 --> 00:10:49.639 They raised twenty five billion dollars, right, but it priced 199 00:10:49.639 --> 00:10:52.519 at roughly one hundred and twenty basis points over treasuries 200 00:10:53.080 --> 00:10:55.000 last year that spread would have been half of that. 201 00:10:55.519 --> 00:10:58.240 Okay, but you're citing one hundred and twenty basis points 202 00:10:58.240 --> 00:10:59.919 spread as if it's an economic crice. 203 00:11:00.840 --> 00:11:02.440 It's a significant indicator. 204 00:11:02.879 --> 00:11:06.320 A wider spread just means investors want a slightly higher premium. 205 00:11:06.360 --> 00:11:09.759 It doesn't indicate a broken capital market. It just indicates 206 00:11:09.799 --> 00:11:10.960 standard supply and demand. 207 00:11:11.120 --> 00:11:14.799 It indicates severe investor fatigue and a rising fear premium. 208 00:11:15.120 --> 00:11:17.519 Let's translate what those basis points actually mean for the 209 00:11:17.519 --> 00:11:21.399 broader market. Sure Tech corporate bond spreads are currently at 210 00:11:21.440 --> 00:11:24.759 eighty nine basis points. That is, nine basis points wider 211 00:11:24.799 --> 00:11:26.639 than the overall investment grade market. 212 00:11:26.879 --> 00:11:27.200 Okay. 213 00:11:27.600 --> 00:11:30.720 Neil Sutherland at Schroeders pointed this out. Clearly, tech has 214 00:11:30.840 --> 00:11:34.399 historically traded materially tighter, meaning it was cheaper for them 215 00:11:34.399 --> 00:11:36.759 to borrow than almost anyone else because they were seen 216 00:11:36.799 --> 00:11:37.919 as incredibly safe. 217 00:11:38.200 --> 00:11:41.039 Right, they had a built in advantage exactly. 218 00:11:41.000 --> 00:11:43.960 But now they are trading wider than the average corporation. 219 00:11:44.519 --> 00:11:47.159 Alphabet had to offer a concession of ten to fifteen 220 00:11:47.200 --> 00:11:49.480 basis points just to get their recent offering. 221 00:11:49.279 --> 00:11:50.639 Absorbed, a small concession. 222 00:11:50.759 --> 00:11:53.519 But investors are demanding a penalty because the sheer volume 223 00:11:53.559 --> 00:11:56.000 of this two hundred and twenty billion dollar debt issuance 224 00:11:56.279 --> 00:11:57.879 is overwhelming the market's plumbing. 225 00:11:58.000 --> 00:12:01.639 Okay, but you're describing a temporary resupply dynamic, not a 226 00:12:01.679 --> 00:12:06.919 permanent structural flaw. The hyperscalers still carry incredibly strong corporate ratings. 227 00:12:06.960 --> 00:12:08.639 I'm not saying they're going bankrupt. 228 00:12:08.720 --> 00:12:11.600 They have fortress like balance sheets and they generate massive, 229 00:12:11.799 --> 00:12:15.360 reliable cash flows. Karen Choi, a capital group, noted that 230 00:12:15.399 --> 00:12:18.600 the investment grade corporate bond index is yielding around five 231 00:12:18.679 --> 00:12:21.960 point four percent. Right, that yield is perfectly in line 232 00:12:22.000 --> 00:12:26.919 with long term historical averages. Foreign investors, pension funds, insurance companies. 233 00:12:27.120 --> 00:12:30.639 They are still eagerly absorbing this issuance because the fundamentals 234 00:12:30.639 --> 00:12:32.600 of the underlying businesses are solid. 235 00:12:32.720 --> 00:12:34.759 They are solid. Yes, they are. 236 00:12:34.600 --> 00:12:38.440 Investing in infrastructure that has a clear, very obvious path 237 00:12:38.480 --> 00:12:39.320 to profitability. 238 00:12:39.440 --> 00:12:42.279 Like I said, no one is arguing that Amazon is 239 00:12:42.279 --> 00:12:46.320 going to default on its debt. This isn't a credit 240 00:12:46.399 --> 00:12:50.080 risk issue in the traditional sense. It's a capacity issue 241 00:12:50.120 --> 00:12:53.360 capacity in what way? Think of the institutional bond market 242 00:12:53.440 --> 00:12:57.360 as a standard garden hose. It has a rigid, fixed capacity. 243 00:12:57.639 --> 00:12:59.159 Okay, I follow what. 244 00:12:59.240 --> 00:13:01.480 The AI hyph what for scalers are doing right now 245 00:13:01.559 --> 00:13:04.159 is trying to force a fire hose volume of debt 246 00:13:04.559 --> 00:13:07.559 through that garden hose. It doesn't matter how pure the 247 00:13:07.600 --> 00:13:10.720 water is, meaning how strong their credit ratings are. The 248 00:13:10.759 --> 00:13:13.360 hose can only take so much pressure before it backs up. 249 00:13:13.600 --> 00:13:16.159 Wait, let's clarify that for a moment. Why is the 250 00:13:16.200 --> 00:13:19.879 hose rigid? Capital markets are literally designed to be dynamic, 251 00:13:20.279 --> 00:13:23.759 not entirely. If there is a profitable opportunity, the market 252 00:13:23.799 --> 00:13:27.360 expands to fund it. Why are you treating institutional investor 253 00:13:27.399 --> 00:13:29.639 demand as a hard physical limit. 254 00:13:29.679 --> 00:13:34.519 Because of strict regulatory and portfolio mandates. Karen Choi highlighted 255 00:13:34.519 --> 00:13:39.720 this exact friction point, the concentration limits. Yes, many pension 256 00:13:39.759 --> 00:13:43.240 and insurance funds have hard rules that cap their exposure 257 00:13:43.279 --> 00:13:46.279 to individual issuers at like two or three percent of 258 00:13:46.279 --> 00:13:50.279 their total assets. Right, They literally are not allowed to 259 00:13:50.320 --> 00:13:53.159 hold ten percent of their portfolio and Amazon bonds, no 260 00:13:53.200 --> 00:13:56.480 matter how attractive the yield is, the hose is rigid 261 00:13:56.559 --> 00:14:01.519 because of regulation and risk management. Okay, So as the 262 00:14:01.559 --> 00:14:05.960 same handful of tech giants repeatedly tap the market issuing 263 00:14:05.960 --> 00:14:08.840 that two hundred and twenty billion in just eight months, 264 00:14:09.360 --> 00:14:14.519 they hit those hard portfolio constraints. Right, As George Katrombone 265 00:14:14.519 --> 00:14:18.360 at DWS stated, it's not a blank check. If hyperscalers 266 00:14:18.440 --> 00:14:21.480 keep demanding capital at this rate, they have to offer 267 00:14:21.600 --> 00:14:25.000 higher yields to attract buyers outside of those traditional. 268 00:14:24.600 --> 00:14:25.679 Channels, which they're doing. 269 00:14:26.000 --> 00:14:29.360 But when their yields rise universally, it causes widespread inflation 270 00:14:29.440 --> 00:14:31.240 in the cost of capital for everyone else. 271 00:14:31.480 --> 00:14:35.919 Okay, we need some definitional precision here regarding what we 272 00:14:36.000 --> 00:14:39.399 call inflation. All right, what you just detailed, you know, 273 00:14:39.519 --> 00:14:44.639 temporary spikes in chip prices or short term portfolio constraints 274 00:14:44.679 --> 00:14:47.720 causing wider spreads in the bond market. Those are front 275 00:14:47.759 --> 00:14:49.399 loaded capital expenditures. 276 00:14:49.440 --> 00:14:50.360 They're still costs. 277 00:14:50.559 --> 00:14:53.639 They are the necessary friction cost of building out a 278 00:14:53.720 --> 00:14:59.240 new technological paradigm. It is not systemic long term macroeconomic inflation. 279 00:14:59.399 --> 00:15:02.279 I disagree. I think they bleed into the macroeconomy. 280 00:15:02.360 --> 00:15:06.279 But when the global labor market shifts structurally, replacing millions 281 00:15:06.279 --> 00:15:09.960 of billable human hours with automated AI driven outcomes, you 282 00:15:10.039 --> 00:15:14.320 create a permanent deflationary shock. That fundamental shift in how 283 00:15:14.360 --> 00:15:18.600 work is priced will easily outlive this current, admittedly aggressive 284 00:15:18.879 --> 00:15:20.759 cycle of hyperscaler debt issuance. 285 00:15:20.879 --> 00:15:25.559 You are vastly underestimating how short term capital friction becomes 286 00:15:25.559 --> 00:15:28.120 embedded into long term systemic inflation. 287 00:15:28.360 --> 00:15:30.039 How so walk me through it. 288 00:15:30.399 --> 00:15:35.320 Let's return to Petrachudin's core warning. When you redirect hundreds 289 00:15:35.320 --> 00:15:39.120 of billions of dollars into AI infrastructure, you cause immediate 290 00:15:39.360 --> 00:15:42.080 cascade difficulties for the rest of the economy. 291 00:15:42.360 --> 00:15:44.320 Okay, let's trace the mechanism. 292 00:15:44.399 --> 00:15:48.639 Then, right now, governments are still running massive deficits, meaning 293 00:15:48.639 --> 00:15:51.919 they are issuing a lot of government debt. Yes, add 294 00:15:51.919 --> 00:15:55.399 this absolute flood of AI borrowing to the market. When 295 00:15:55.480 --> 00:15:59.240 hyperscalers soak up the available capital, it pushes up treasury 296 00:15:59.320 --> 00:16:03.000 yields because the competition for every investment dollar is fierce. 297 00:16:03.399 --> 00:16:07.519 Yes, competition for capital drives up yields temporarily, that is 298 00:16:07.679 --> 00:16:11.720 exactly how markets prioritize the most vital infrastructure projects. 299 00:16:12.000 --> 00:16:14.960 But look at the collateral damage. When treasury yields are 300 00:16:14.960 --> 00:16:18.159 pushed up, the benchmark cost of borrowing rises for the 301 00:16:18.320 --> 00:16:22.519 entire global economy. Right Suppose an agricultural company needs a 302 00:16:22.559 --> 00:16:26.320 loan to build a new tractor factory. Because the hyperscalers 303 00:16:26.360 --> 00:16:30.159 have soaked up so much capital, that agricultural company has 304 00:16:30.200 --> 00:16:33.759 to offer a significantly higher interest rate just to secure 305 00:16:33.799 --> 00:16:34.240 their loan. 306 00:16:34.600 --> 00:16:38.200 So their borrowing cost goes up, their operating cost surge. 307 00:16:38.600 --> 00:16:41.039 How do they recover that cost They raise the price 308 00:16:41.039 --> 00:16:43.759 of the tractors, which raises the price of food production, 309 00:16:44.039 --> 00:16:47.000 which gets passed directly to the consumer at the grocery store. 310 00:16:47.360 --> 00:16:48.679 I see the chain you're building. 311 00:16:48.879 --> 00:16:51.679 So even if a corporate HR department is saving money 312 00:16:51.679 --> 00:16:54.679 on their IT services down the line, the cost of housing, 313 00:16:54.720 --> 00:16:58.960 food production and traditional manufacturing goes up right now because 314 00:16:59.000 --> 00:17:04.599 capital is so expened. Okay, the immediate inflationary pressure neutralizes 315 00:17:04.680 --> 00:17:10.000 those promised future efficiencies. Silvana Tenrero's research at the IMF 316 00:17:10.039 --> 00:17:14.960 confirms this. Boosting productivity does not automatically lower consumer inflation 317 00:17:15.319 --> 00:17:18.400 if the macroeconomic environment is heavily distorted by the cost 318 00:17:18.400 --> 00:17:19.759 of achieving that productivity. 319 00:17:20.079 --> 00:17:23.839 But the crowding out effect you're describing assumes that AI 320 00:17:24.039 --> 00:17:28.119 exists in a vacuum, completely separate from the agricultural company 321 00:17:28.240 --> 00:17:29.559 or the manufacturing plant. 322 00:17:29.880 --> 00:17:31.519 I'm not saying it's in a vacuum. 323 00:17:31.799 --> 00:17:34.440 History shows us that the dividends of a general purpose 324 00:17:34.480 --> 00:17:38.279 technology always spill over to benefit those exact sectors. 325 00:17:39.160 --> 00:17:39.759 Eventually. 326 00:17:40.119 --> 00:17:43.880 Maybe that tractor factory is precisely the kind of business 327 00:17:43.880 --> 00:17:47.799 that will benefit when the cost of enterprise software supply, chain, logistics, 328 00:17:47.839 --> 00:17:51.559 and back office administration plummets to mere zero thanks to AI. 329 00:17:52.000 --> 00:17:54.240 But they have to survive the capital drought. First. 330 00:17:54.559 --> 00:17:57.480 You are looking at the two hundred and twenty billion 331 00:17:57.519 --> 00:18:01.279 in tech debt and seeing a dram the system. I 332 00:18:01.559 --> 00:18:04.599 look at that exact saying two hundred and twenty billion 333 00:18:05.000 --> 00:18:08.640 and see the necessary down payment on a global operating 334 00:18:08.720 --> 00:18:13.559 system that makes every subsequent physical and digital transaction cheaper. 335 00:18:13.720 --> 00:18:16.720 A down payment that is currently straining the physical limits 336 00:18:16.759 --> 00:18:17.759 of our infrastructure. 337 00:18:18.000 --> 00:18:19.640 Well, growth requires strength. 338 00:18:19.720 --> 00:18:22.559 When tech spreads widened by nine basis points over the 339 00:18:22.599 --> 00:18:26.039 investment grade market, that is, the market's mechanism of screaming 340 00:18:26.079 --> 00:18:29.200 for a breather. The debt issuance in August looked very 341 00:18:29.200 --> 00:18:33.400 different from January. How So, in January AI linked deals 342 00:18:33.440 --> 00:18:36.839 were absorbed easily. By August, fatigue had deeply set in. 343 00:18:37.319 --> 00:18:40.799 Companies that historically issued shorter term debt are now locking 344 00:18:40.839 --> 00:18:44.000 in long dated bonds just to finance these server farms. 345 00:18:44.160 --> 00:18:45.599 Right to secure the runway. 346 00:18:46.000 --> 00:18:50.400 They are intentionally locking in higher capital costs for the 347 00:18:50.440 --> 00:18:53.319 next decade because they know the buildout is going to 348 00:18:53.319 --> 00:18:55.680 be brutally expensive for a long time. 349 00:18:56.279 --> 00:18:59.160 They are locking in those costs because the projected return 350 00:18:59.240 --> 00:19:03.440 on invested care capital for AI compute is completely unprecedented. 351 00:19:03.759 --> 00:19:07.559 Let's trace this macro theory back to the micro evidence. Okay, 352 00:19:07.759 --> 00:19:11.079 why would a massive German utility like Eon sign a 353 00:19:11.160 --> 00:19:14.680 multi year deal with HCl Tech where they pay nothing 354 00:19:14.720 --> 00:19:15.480 in the first. 355 00:19:15.279 --> 00:19:18.000 Year because HCl Tech took the risk, right? 356 00:19:18.160 --> 00:19:22.079 But why because the underlying technology, the exact AI models 357 00:19:22.160 --> 00:19:25.960 running on those expensive debt finance server farms, is capable 358 00:19:25.960 --> 00:19:30.359 of delivering such massive verifiable efficiency gains in year two 359 00:19:30.519 --> 00:19:32.440 that the provider is willing to take on all the 360 00:19:32.559 --> 00:19:33.599 upfront risk. 361 00:19:33.440 --> 00:19:35.240 It's still a huge gamble. 362 00:19:34.960 --> 00:19:39.960 That represents a fundamental structural change in how economic value 363 00:19:40.000 --> 00:19:43.480 is generated. We are entirely moving away from paying for 364 00:19:43.599 --> 00:19:44.759 human time well. 365 00:19:44.960 --> 00:19:49.119 Algorithmic certainty is a highly optimistic phrase for a technology 366 00:19:49.160 --> 00:19:52.279 that is still wildly expensive to train, run. 367 00:19:52.119 --> 00:19:54.200 And cool, but the capabilities are there. 368 00:19:54.519 --> 00:19:57.119 And again, if we look at the macro picture, a 369 00:19:57.279 --> 00:20:01.160 utility company saving money on its internal software management does 370 00:20:01.200 --> 00:20:05.240 not solve the structural inflation caused by real world resource bottlenecks. 371 00:20:05.480 --> 00:20:07.599 The resource bottlenecks will sort themselves out. 372 00:20:07.880 --> 00:20:10.640 There are physical limits on our energy grids to power 373 00:20:10.680 --> 00:20:13.599 these massive data centers. There are hard limits on the 374 00:20:13.640 --> 00:20:17.960 TSMC foundries producing the specialized silicon. The prices for those 375 00:20:18.000 --> 00:20:21.480 physical inputs will remain structurally high. That is the upward 376 00:20:21.519 --> 00:20:25.200 inflationary pressure to Chewton is warning central banks about. This 377 00:20:25.319 --> 00:20:28.200 isn't just a capital market phenomenon. It is a physical 378 00:20:28.240 --> 00:20:29.200 resource phenomenon. 379 00:20:29.359 --> 00:20:32.680 But bottlenecks and physical resources are exactly what the free 380 00:20:32.680 --> 00:20:36.160 market is exceptionally good at solving when the price incentive 381 00:20:36.240 --> 00:20:36.759 is right. 382 00:20:36.920 --> 00:20:40.119 By raising prices, the high yields. 383 00:20:39.799 --> 00:20:42.319 And widespreads you are pointing to in the debt markets 384 00:20:42.319 --> 00:20:45.599 are the exact signal needed to attract more capital to 385 00:20:45.720 --> 00:20:47.759 solve the energy and chip constraints. 386 00:20:47.920 --> 00:20:51.759 But the act of solving those constraints is inherently inflationary. 387 00:20:52.279 --> 00:20:53.400 Not in the long run. 388 00:20:53.640 --> 00:20:56.119 When you have to build new nuclear power plants or 389 00:20:56.160 --> 00:20:59.599 reactivate dormant ones just to feed data centers, the cost 390 00:20:59.640 --> 00:21:03.559 of concrete, steel and specialized labor goes up for everyone 391 00:21:03.559 --> 00:21:05.039 else in the economy. 392 00:21:04.759 --> 00:21:07.200 But they're producing more power overall. 393 00:21:07.480 --> 00:21:10.599 The Swiss National Bank hasn't changed its long term inflation 394 00:21:10.680 --> 00:21:13.920 forecast yet they still project zero to two percent through 395 00:21:13.960 --> 00:21:17.400 twenty twenty nine. But Shuden made it explicitly clear that 396 00:21:17.440 --> 00:21:20.200 their policy rate is not set in stone right. 397 00:21:20.279 --> 00:21:21.880 They have to remain flexible. 398 00:21:22.119 --> 00:21:25.640 If this AI investment cycle continues to warp the broader 399 00:21:25.640 --> 00:21:29.559 economy and suck up physical resources, central banks will be 400 00:21:29.599 --> 00:21:32.920 forced to adjust monetary policy to contain the inflation it 401 00:21:32.960 --> 00:21:33.480