I want you to visualize two very different things. On one side, you have this plan for a gigawatt scale supercomputer in India. They're calling it Stargate. It's massive. Or city level power consumption. Exactly. And then on the other side, you just have a tweet, a warning from Andrew Yang about the, quote, disemboweling of white collar jobs. It feels like two completely separate worlds. It does. But they're not. They're actually
the same story. That huge piece of infrastructure is being built to automate the exact kind of cognitive work that keeps our offices running. Welcome to the Deep Dive. It is Wednesday, February 18th, 2026. If you're listening, you're probably what we call the learner, someone who wants to see the mechanics underneath the machine, not just the headlines. And the headlines are just... They're deafening right now. I was looking at our sources for today, and it really feels like
the tectonic plates are shifting. We're talking about physical AI infrastructure in India, a total change in how software thinks, and then, well, the economic fallout when all that hits the job market. Yeah, it's a chain reaction. So here's our plan. We're going to start in Delhi with this whole Stargate thing and what sovereign AI even means. Yeah, then we'll hit the agentic shift when tools start doing things on their own. And the cost of that intelligence, which
is... Honestly, it's pretty staggering. Yeah. Then we'll land on that warning from Adrienne about the end of the middle manager. It's a lot, but you can draw a straight line through all of it. Okay, so let's start with the hardware. OpenAI for India, we hear these big user numbers all the time, but what's the actual scale we're talking about here? The scale is kind of hard to wrap your head around. It's 100 million weekly chat GPT users. Just in India? 100 million. Yeah.
I mean, that's like a third of the entire U .S. population. But that's not even the big news. The real story is the partnership between OpenAI and the Tata Group. Which is basically the industrial backbone of India. Right. Specifically, Tata Consultancy Services, TCS, they've announced the Stargate Initiative. And this isn't just a cool name. They are literally building the infrastructure. OpenAI is the first customer for their new HyperVault data center. And they
are starting at 100 megawatts of capacity. For anyone listening who's not an engineer, 100 megawatts. Is that a lot? It's huge for a starting point. Your typical big data center might be, I don't know, 20 or 30 megs. But the plan here is to scale to one gigawatt. A gigawatt. That's a nuclear reactor. It is. It's the kind of power that lights up a major city. And all of that energy is going to be dedicated to one thing. Thinking. Processing Tokens Which brings us to the key concept here,
sovereign AI. I keep seeing that term, sovereign AI. It sounds very official, very diplomatic. What does it actually mean, though? Think of it like a national border for digital thought. For years, if you used a big AI model, your data was probably being sent to a server in Virginia or Dublin. Right. But if you're the Indian government or a huge bank there, you just can't have your sensitive data leaving the country. That's a
non -starter. So sovereign AI means the whole stack, the computers, the storage, the model. It all lives inside India's borders. Exactly. And that suddenly unlocks all these use cases that were impossible before. Mission critical government work, for example. You couldn't do that before because of security. Right. Now you can. So you see Tata rolling out chat GPT enterprise to hundreds of thousands of their own people. A's New Delhi is using it for medical education.
The data stays local. The intelligence is world class. So it's about nations controlling their own what? cognitive infrastructure? That's the perfect phrase for it. Yeah. Just like you need rows and bridges, nations now feel they need their own sovereign AI capacity. Otherwise, you're just renting intelligence from somebody else. Okay. So you have this massive engine being built in the East. Let's pivot to the software that runs on these things. The big news right now
is what's being called the agentic shift. We saw XAI drop Grok 4 .20 into public beta. And, you know, Aside from the classic Elon Musk version number, what's actually different under the hood? The whole workflow is different. I mean, until now, we've treated AI like a very smart encyclopedia. You ask a question, it gives you an answer. It's one to one. Grok 4 .2 changes that. When you give it a complex task, it doesn't just answer. It actually spins up four separate AI agents
that work in parallel. So like one agent is doing research while another one is reasoning through the logic. Exactly. It's like a little committee. One researches, one reasons, one might draft some code and another one reviews it. It's a shift from just chatting to actually collaborating. That makes sense. But my first thought is cost. If I ask one question and it kicks off four different processes, haven't I just like quadrupled my compute cost? You absolutely have. And that is
the elephant in the room. You know, we saw Apple integrating ChatGPT and Claude into CarPlay, which is great. Tap to ask is way better than yelling at your phone. Sure. But the ubiquity of it all hides the incredible expense. I was just reading the projections for Anthropic. The numbers, they're just eye -watering. They're heavy. Anthropic is projected to pay over $80 billion. That's billion with a B. $80 billion. To Amazon, Google, and Microsoft by 2029. And
that is just the cloud computing bill. That's the rent they pay just to exist. $80 billion just to keep the lights on. And look, their sales are exploding. Projected to hit like $6 .4 billion. That's amazing growth. But just... Do the math. The cost of generating this intelligence is astronomical. That explains moves like Mistral acquiring Coyab. Right. The French AI lab. They're trying to build their own infrastructure muscle because renting compute forever is just not sustainable. This
really isn't a software bubble then. It's an infrastructure arms race. Totally. Okay. So we have these huge expensive models. Yeah. How is that actually trickling down into the tools that you and I use every day? It's trickling down fast, and it's changing what the tools do. Look at something like WordPress. For 20 years, it was a blank page. You had to do the work. Now they've launched a built -in AI assistant. It can rewrite your text, change the layout, all
without you leaving the dashboard. So the tool is no longer passive. No. Or look at Flixier for video editing. It automates trimming and cutting. Or Figma. which is using Claude code to turn a UI design directly into editable code layers, the tool is becoming an active partner. I have to. I have to make a bit of a vulnerable admission here. I honestly still wrestle with prompt drift myself. Oh, yeah. As these models get smarter, I find myself kind of struggling.
Like, I get used to prompting Claude 4 .5, I learn his little quirks, and then 4 .6 drops, which it just did, and all my old prompts, my spells, they don't work the same. The model just thinks differently now. It's so different from any other software, right? When Microsoft updates Excel, SUM still means SUM. But when an AI model updates, its whole reasoning path can shift. It's like your colleague got a brain transplant over the weekend. Yeah, you have to relearn how
to talk to them. And that leads us right to the heaviest part of this. We've got hardware scaling to gigawatts. We have software becoming agentic. And that brings us to Andrew Yang. Yeah. That piece he wrote on X about the end of the office, it went absolutely viral. And the language he uses, it's visceral. He calls it the great disemboweling of white collar jobs. Disemboweling. He's not pulling any punches. No. He's trying to shake
people up. His whole argument is that we are witnessing the extinction of a very specific kind of worker. The middle manager. The coordinator. Exactly. But why the middle? For a decade, we heard that robots were coming for the blue collar jobs, for truck drivers and factory workers. Why is the target now the person in middle management? It all comes down to what he calls cost logic. Replacing a truck driver is actually really, really hard. You need robotics, sensors, insurance.
It's a mess. Right. Physical world problems. But a middle manager, what is their primary output? It's not a physical thing. In theory, it's strategy and guidance and practice. It's a lot of emails. It's summarizing reports from one team to send up to the next level. It's high cost coordination. It's routing information. And Yang's point is that the AI agents we were just talking about, they are infinitely cheaper and faster at routing information than a human who is earning 150 grand
a year. The math is just, it's brutal. It's a spreadsheet decision. He says if five managers can be replaced by one operator with a fleet of AI agents, the market will force that to happen. You know, this gets to what I call the CEO's dilemma. Think about this a lot. Let's say you're a CEO and you want to do the right thing. You don't want to fire people, but your main competitor just automated their entire middle layer and
their profit margins just jumped 20%. Suddenly your investors are on the phone screaming at you. They don't care about your morals. They care about the returns. You're trapped. You have to make the same cuts just to survive. That's the trap. It becomes a race to the bottom on labor costs. And the ripple effects, they're frightening. If you hollow out that whole middle layer, what happens to entry -level jobs? There's no next step. You hire a junior analyst to eventually
become a manager. But if that manager role doesn't exist, why hire the junior in the first place? And what does that do to the value of a degree? Yang makes the point that if knowledge is basically free and reasoning is cheap, does a generic business degree even hold its value? We could see some colleges, the weaker ones, just shut down. It's possible. We're moving from an economy of credentialism to an economy of capability. It's not about what you know. It's about whether you can direct the
machine. It's a heavy thought. It really shakes the foundation of what we've all been told is a safe career. It does. It means the safe middle is now the most dangerous place to be. And that's the thing that keeps me up at night. OK, let's try to tie all this together. We started with gigawatts of power in India and we ended with the deaths of the middle manager. How do those two things connect? They are the hardware and
the software of the exact same phenomenon. That Stargate initiative, that massive amount of compute. That is the engine. Right. It's being built specifically to run the kind of heavy reasoning models that can perform that high -level coordination work. So TATA and OpenAI are literally laying the physical groundwork. For the exact economic shift Andrew Yang is warning us about, it's a move from AI as an assistant, something that helps you work to AI as an operator, something that does the
work. The infrastructure is being built to support the replacement of cognitive labor. That's the big idea. That's it. And for you, the learner listening to this, it's not about panic. It's about recognizing that the middle is not a safe place to be anymore. You either need to be building the machine or directing the machine. Which leads to the final thought I want to leave you with. If that middle part of the corporate ladder just disappears, what does a career path even look
like in 2030? Do you just go from intern straight to executive? or do we need a completely new structure for how people contribute value? That might be the question of the decade. Maybe it's not a ladder anymore. Maybe it's more like a network of specialists. But the days of just climbing the rungs, those are probably over. So I want you to think about your own job this
week. Look at your calendar. How much of what you do is coordination, just moving information around, and how much is actual creation or direction? Because the coordination part is being solved. That's the work. Thank you for diving in with us. It's a complicated new world, but at least we can try to parse it together. See you in the deep end. See you next time.