If you're an entrepreneur, it's a question of what product to build. If you work in an office, it's a question of which tasks will go to artificial intelligence. If you're an investor, it's a question of whose money will lose its power. If you're a developer, it's a question of who's going to write and maintain the software. Today I'll show you a simple framework: where the old software stays, and where artificial intelligence takes the value. And in twelve months we'll check whether companies start paying not for seats, but genuinely for artificial intelligence. Not the way it's happening now, but for real.
To set the stage, today we'll start with three changes, important changes, that are present on the market. That's an update concerning OpenAI, an update concerning Anthropic. And what is Elon Musk doing? Yes, that topic stands somewhat off to the side, and if you bring it up everywhere, it looks like hype. But let's take a real look at it. Two years ago already, when we launched the ToTheMoon channel, I said it more than once, in conversations with a huge number of people — technical people in Silicon Valley, all kinds of businesspeople, investors, anyone you like, and my partners, friends and acquaintances.
I'd say: "Listen, well, obviously, since the leaders on the market today, the leaders in models, aren't making just a chat — right now, at this moment in time, they're simply training on us, using us as guinea pigs. Obviously, once they've made AGI, or made something more serious, they'll get the ability to generate software. And then clearly — why would I need, why would I need a CRM system?" I always used a CRM system as the example. It's a simple story. It's a system that salespeople use in the course of their work. What is it even for, fundamentally?
When a salesperson comes to work, the main thing for them is to understand who to call, or who to get in touch with, or who to go see, who to talk to. Why do they need to see some sales-funnel interface? What even is that? This isn't even a question of the next task. It's a question of something completely different. A system that doesn't just interact with the layers of a CRM system, not just me and an agent — because right now everyone everywhere is talking about agents — but something completely different.
And then the question comes up: will an operating system even exist at all? Won't companies like OpenAI end up with their own operating systems? Won't they make their own mobile apps? That, by the way, is one of the reasons Apple doesn't let third-party developers in. I mean, seriously, it doesn't let them into its own infrastructure. Although with Apple it's still completely unclear what's going to happen to them in terms of their apps. How are they going to live in this market?
Won't it happen that at some point in time, people who actually use Apple will sacrifice their phone in order to get, well, some kind of super result elsewhere — which, in fact, is what I'm seeing more and more now. In March 2026, Sam Altman describes AI not as a chatbot but as — in English, in the English language there's this word, utility, right? And it encompasses everything to do with electricity, water, gas. And I think it carries a somewhat different context when translated into Russian.
But still, he describes AI, artificial intelligence, as becoming like electricity, or like water, or like gas, or like gasoline. And within that logic he says the following phrase. He first said it, by the way, in 2024, and part of that was shown in the interview with Lex Fridman. But anyway, the phrase went as follows. Back then he said it superficially. He said the time will come when we'll be able to spend, say, ten million dollars, some amount of money, to get some amount of tasks done.
That is, we'll throw it at computation, at compute. So then, in 2026 he said the following: one query can be cheap, but one extremely valuable task may require tens of millions of dollars, hundreds of millions of dollars, and someday even billions of dollars of compute. And here's where the following story comes up. So I've been developing — in recent, as it happened, in recent weeks I've shifted heavily into developing a large number of different software solutions, using Codex.
It's phenomenal, it's fundamental, I talk about it a lot on ToTheMoon, and there will be more special episodes on it. I never would have thought I'd start programming this seriously and this much, and I realize that hardly anyone from the outside could ever have solved tasks like these for me. Even though I've been in software for more than twenty years. Well, this deep in it, twenty-three, twenty-four years. And here's the thing: when I'm programming something, I'm constantly making requests for the system to write code for me, and I'm limited by that system's resources.
That is, if I give it too much input data and say I need to build, for example, an operating system, or I need to build a system for two companies to interact with each other, or a messaging system inside a company, or a task management system inside a company, it won't have enough computing power — at least not what's available to me. Even though I use the most expensive, latest models, essentially without limits. It won't have enough, and I simply won't be able to do it. I have to wait all the time.
I do everything iteratively, gradually, bit by bit. And I had one very big example of this. I said: "What if I could analyze a third-party CRM system and spend two hundred thousand dollars analyzing that system, and then spend three hundred thousand dollars building it?" So here, when Altman says you could spend two billion dollars on some project, that's already a serious question. Well, you could essentially analyze Windows, Microsoft Windows, and build — build Windows. But let's go in order. So, what happened?
Something very serious happened in the last month. And this news sort of sits somewhere off to the side, and you don't see it, but it's worth understanding in order to get to the main story. What's actually going to happen to software? And why did I say that artificial intelligence doesn't kill software, it makes it invisible? So, OpenAI announced that they're moving from the model to the concept of a deployment company. That is, OpenAI as a deployment company is essentially not just a sales department.
According to the official text, it's a company that's supposed to help organizations build and deploy artificial intelligence systems for important everyday work. What did they do? In parallel, they acquired Tamora. It's a British company, they have a hundred and fifty developers, and it has a huge number of great partners. They clearly acquired them for the partners, and said they'd also put starting capital in there, something like four billion dollars, and so they're getting into software development.
At least that's how it sounds from the outside. Like, we're sort of getting into software development. And here comes up, here comes up a very interesting aspect. So, in this case, is OpenAI going into a company as a consultant? Or do they actually want to figure out how it all works and build their own internal solutions? That is, if, say, OpenAI now comes to some company and offers to implement and use some third-party solutions and services — including database solutions, storage, text analysis systems, email sending, messaging of various kinds, some infrastructure solutions, customer task management systems, logistics and warehouse management systems, anything at all.
Or are they doing this now in order to create, going forward, companies that will develop the software entirely themselves? By the way, I want to point out that these hundred and fifty engineers they bought — that's called a forward-deployed engineer. It may seem like a technical thing, but it's important to understand. It's an engineer who doesn't sit in the OpenAI office, or doesn't just write a generic product. He goes inside the client, looks at the real processes — what data there is, permissions, chaos, what legacy there is, what people work there — and builds solutions right on site.
That is, they're genuinely building the system from the inside, while sitting in that company. This is the ToTheMoon channel. Tech news, insights from Silicon Valley. Subscribe so you don't miss our episodes. We come out several times a week, and here you'll learn not just what artificial intelligence is, not just how to bring it into your life, not just how to make money on it, but how to actually be able to keep pace with the times right now. What does Anthropic do next? Or rather, Anthropic did it a bit earlier. Anthropic, on May 4th, about a week before OpenAI announced that decision.
Anthropic, together with Blackstone, Hellman & Friedman and Goldman Sachs, announced the creation of a new AI services company. That really shook up the market of companies that are in software development today. For example, one of the companies well known to Russians — well, to Russian-speaking people — is EPAM, originally founded in Belarus. Its head office was in Belarus, and at one point it was, well, a fairly serious, big company even compared with, well, the global giants. And overall, it's a big company.
And Anthropic, entering this market — when they entered this market, I heard a lot of feedback from that market that, well, companies started worrying very seriously about what's going to happen next. At first it seems to them that... What's very interesting here is that they're taking the big, the very biggest clients for themselves all over the world, because OpenAI even announced that they're partnering up and will serve, security-wise, even European corporations, not just American ones.
There's a real example. Anthropic says, for example, there are areas concerning healthcare, and within that healthcare area doctors spend hours on various documentation, on describing different practices, what goes on in the hospital, on prioritizing different tasks and everything else. Right? And so, the team of their new company is supposed to sit down Yes, the new company is supposed to sit down, talk together with the IT staff, with the doctors, and build a tool around the already existing processes.
I'll be honest with you: I'm a patient in the American healthcare system. And in the American system — the way care actually works on the inside. I come to the doctors, and all the doctors sit with automated systems that always record our conversation and automatically produce a super-detailed record. The system inside suggests a huge number of steps. Today those steps are static. That means there's a certain clear sequence to what the doctor does. But I can see that soon, already right now, they're going to move to dynamic processes.
And if we think Anthropic comes in and talks with the IT staff and thinks about how to analyze everything, then the question is: why would they even need to talk to the IT staff? After all, they can take everything: the source code of the IT systems they have inside, or not even the source code — closed systems, how they communicate with each other. Take all the protocols that are inside, all the reports that are inside, get access to the databases that are inside, and analyze absolutely everything.
Are OpenAI and Anthropic going into companies as consultants, or are OpenAI and Anthropic actually planning to compete with the software makers themselves, the very creators of the software? Well, here I'll be very interested in what you write in the comments. For my part, of course, my opinion is well known. It's not for nothing that I talked about this two years ago, and I believe it's an incredible bet that was being placed on this. That's why I'm making a big bet, even in terms of investing.
Everything to do with infrastructure — from electricity and circuit boards to all the companies that are in any way connected with this. And there's a radical story that everyone treats, I think, as a joke. At first they treated it as a joke. It's when Elon Musk decided, said, that Wikipedia isn't real. That we need to make a copy — not a copy, but create our own Wikipedia. And that project, by the way, is very much alive, and big. Then when Elon Musk said he'd registered a trademark called Macrohard. That is, there's Microsoft, and he set macro against micro, and hard against soft.
And at first, from the outside, it looked like a kind of joke on the market. But if we really look at it, it's a direct path to what? What, essentially — look — what is Musk essentially saying? Musk is staking out the extreme position. He says there won't be agents inside a company; there will be AI-native software companies. That is, companies that are, in fact, artificial intelligence themselves. Jensen Huang — that's the head of NVIDIA — said in a recent interview, literally a month ago, that, well yes, a company like NVIDIA can't be built with artificial intelligence right now, but you can build something simpler.
But I want to say right away that NVIDIA is still worth four trillion dollars, and obviously, well, it's a very serious company. More than four trillion dollars. So what did he mean by something simpler? What kind of company? A billion, five billion, twenty billion, a hundred million dollars, a million dollars? What is 'simpler'? So, when Elon Musk sets Macrohard against Microsoft, he's clearly pointing out that we're going to create a new era of everything that's happening.
It's new computers, it's new software, it's a new operating system. And essentially, the old operating system, it... what is it even for? What's its essence? And clearly this new operating system has to be an operating system whose kernel is artificial intelligence, whose kernel isn't just AGI but real artificial intelligence, something real. And NVIDIA, for example, is building that kind of foundation together with computer manufacturers, and with various companies. They have a lot of partnerships, a huge number of projects in this.
Huang specializes in this, running hundreds of projects at once. They're making a kind of layer, like an architecture. And then on top of that goes the computer, the computer layer. And what Jensen Huang says is: "I'll sell the computer, I'll sell the compute so that you can go on and do something with it." And what does Elon Musk do? Elon Musk says: "We'll have our own compute, we'll have our own artificial intelligence, and we'll run everything ourselves." That is, inside the Tesla robot, it will itself have its own incredible computing power, so it can work offline, without the internet.
If it needs the internet, they have their own internet; if it needs some computation done externally, quickly, instantly we'll compute it externally on our servers, on our infrastructure, or connect to other infrastructure, of which we'll have a great deal, and we'll be able to do everything ourselves, fast, on our own operating system. That is, we won't need a third-party operating system; we'll create everything ourselves. I once put forward a thesis that, I think, well, for many, for many people may come across as populist, or rather misunderstood — I think, even misunderstood.
The thesis went like this: I said that if artificial intelligence learns to create its own hardware and learns to program for that hardware, then there won't be any third-party software at all. Artificial intelligence will create absolutely everything itself. Here I may be less technical, although I spent a large part of my life programming for robots and for various equipment, including very complex equipment, twenty years ago already, implementing software for fairly serious factories on the island of Taiwan, where even back then robots were moving along the walls, the ceiling, the floor, and were standing everywhere.
And the number of people wasn't that large in the production and development of complex processors, and the equipment that then serviced millions of dollars — multi-million-dollar equipment servicing the making of processors. In those years that was very serious. Numbers like that were serious in those years. So, this — this thesis of mine comes precisely from there. Because if artificial intelligence can create infrastructure, then of course it will be able to program anything at all for it. What do you think about this?
And when do you think it will happen? And we'll definitely check it together. We'll definitely check it together.