Meta Just Laid Off 8,000 People… To Spend $125 Billion on AI

追加: 2026-05-24

[00:00:00]On May 20th, 8,000 people will receive a notification. Not a performance review, not a warning, a termination notice.

[00:00:07]These aren't bad employees. They aren't underperformers. Many of them have been with the company for years, but Mark Zuckerberg made a decision. And that decision wasn't about them. It was about a number, $125 billion.

[00:00:23]That's what Meta is spending on artificial intelligence this year alone.

[00:00:27]And those 8,000 jobs, they're not being cut to save money. They're being cut to fund it. Here's what nobody is telling you about what's really happening inside the biggest companies on Earth right now. And why it affects every single one of us. Welcome back.

[00:00:42]If you're new here, this channel covers the financial and technological forces reshaping the global economy. No hype, no speculation presented as fact. Just the data, what it means, and what you should know. Today, we're going inside the most important corporate story of 2026.

[00:01:00]It's not just about Meta. It's about four companies spending $725 billion on a single bet. And about what happens to the rest of us if they're right or if they're wrong. Let's get into it. $725 billion.

[00:01:16]Let me put that in perspective. That's larger than the entire GDP of Switzerland. It's more than what the United States federal government spent on Medicare last year. And it's being spent not by governments, but by four private companies. Amazon, Microsoft, Alphabet, and Meta. All of it going into one thing, artificial intelligence infrastructure. Now, before we go further, let's be precise about what AI infrastructure actually means. Because it's easy to hear that phrase and picture some abstract tech project. It's not abstract. It means data centers, millions of square feet of physical buildings filled with servers. It means graphics processing units, GPUs, the specialized chips that run AI models. It means custom silicon, fiber cables, cooling systems, and the land to put it all on. This is concrete and steel as much as it is code. And here's where it gets interesting. Meta's capital expenditure, their spending on physical infrastructure for 2026 is projected between 115 and 135 billion dollars. That number is nearly double what they spent in 2025.

[00:02:31]Now, here's a question that most analysts aren't asking. Where does that money come from? Meta's total annual payroll, every salary, every bonus, every stock grant for all of their employees worldwide, comes to roughly 27 billion dollars. That means even if Zuckerberg fired every single person at Meta tomorrow, every engineer, every designer, every executive, he would save 27 billion dollars, and the infrastructure budget is five times that. The layoffs aren't about saving money. They can't be. The math doesn't work. So, why is Zuckerberg cutting 8,000 jobs specifically now? Because the layoffs aren't a cost-cutting measure.

[00:03:13]They're a financing mechanism. The savings from labor reduction are being redirected, not to profit, but directly into the capital expenditure line, into GPUs, into data centers, into AI. He even said it explicitly to employees.

[00:03:29]The May cuts are a direct consequence of the AI infrastructure budget, not a side effect, a direct consequence. That's an extraordinary thing for a CEO to say.

[00:03:39]And most of the coverage missed what it actually means. To understand why Zuckerberg is making this bet, you need to understand what he believes the next five years look like. Meta's core business, the one that generates billions in revenue every quarter, is advertising. Facebook, Instagram, WhatsApp, Threads.

[00:04:00]The more time people spend on these platforms, the more ads they see, and the more money Meta makes. That business is already dependent on AI. Meta's recommendation algorithms, the systems that decide what appears in your feed, are some of the most sophisticated machine learning systems ever built.

[00:04:17]They process billions of data points in real time to maximize engagement. But what Zuckerberg is building now is something different in scale and scope.

[00:04:26]He's building Llama, Meta's own large language model. He's building AI agents that will eventually handle customer service, content moderation, and user interaction at a scale that would require millions of human employees to replicate. He's building a $27 billion joint venture with a company called Nebius to construct a gigawatt-scale AI data center campus in Louisiana. A gigawatt. That's roughly the power consumption of a small city.

[00:04:55]And here's what makes this a genuine strategic bet, not just aggressive spending. Meta is going open-source with Llama. They're not selling the AI as a service the way OpenAI or Anthropic do.

[00:05:08]They're releasing it publicly, making it available for developers and businesses worldwide to build on top of. Why would you give away the technology? Because the goal isn't to sell the model. The goal is make Meta's AI infrastructure the foundation that the next generation of internet products runs on. If developers build their applications using Llama, they run on Meta's infrastructure. Meta's platforms become the operating system for AI-powered software. This is the same strategic logic that made Android the world's most used mobile operating system. Google gave it away for free, and now they own mobile advertising. Zuckerberg is attempting something similar at a larger scale with more money. Is it going to work? That's the question that matters, and we'll come back to it. But first, Meta isn't alone in making this bet.

[00:05:57]Let's zoom out, because what's happening at Meta is happening simultaneously across the entire technology sector. And when you add the numbers up, the scale of what's happening becomes almost difficult to comprehend. Amazon, Microsoft, Alphabet, Meta.

[00:06:14]These four companies are projected to spend a combined $725 billion on capital projects in 2026.

[00:06:23]That represents a 77% increase compared to what they spent in 2025.

[00:06:28]To put that growth rate in context, 77% year-over-year increase in capital expenditure is not a business investment. It's a land grab. It's four of the most powerful companies in history making a simultaneous decision that the next decade of computing is being decided right now. And that whoever builds the most infrastructure fastest wins.

[00:06:51]Let's break down each company briefly.

[00:06:53]Amazon's AWS, Amazon Web Services, is the world's largest cloud computing platform. They're building AI capabilities directly into their cloud infrastructure so that every business that runs on AWS can access AI tools natively. Amazon's capital expenditure for 2026 is projected to exceed $180 billion.

[00:07:18]Microsoft made a major bet on OpenAI, the company behind ChatGPT, and has been integrating AI into every product in their portfolio. Microsoft Office, Azure, GitHub. They're spending approximately $80 billion on AI data centers in fiscal 2025 alone, with similar commitments for 2026.

[00:07:41]Alphabet, Google's parent company, has a Google Cloud backlog that nearly doubled sequentially to $462 billion.

[00:07:50]Their first quarter 2026 capital expenditure was $36 billion, up 107% year-over-year. They are building AI into Google Search, Google Workspace, and YouTube. And then there's Meta, $125 to $145 billion.

[00:08:09]The biggest single company bet in this race. Now, here's the critical question.

[00:08:14]This money has to come from somewhere, all of it. Amazon, Microsoft, Alphabet, and Meta are cutting tens of thousands of jobs. Not because business is bad, revenues at all four companies are at record highs, but because the capital has to be redirected. The tech industry has already shed more than 95,000 jobs in 2026.

[00:08:37]That's an average of 882 jobs disappearing every single day. And the companies doing the cutting are simultaneously reporting their most profitable quarters ever. This tells you something important about the nature of what's happening. This isn't a recession story. This isn't a revenue problem.

[00:08:55]This is a deliberate reallocation of capital from human labor to machine infrastructure at a speed and scale we've never seen before. The question isn't whether this is happening. It is.

[00:09:06]The question is, what does it mean?

[00:09:09]There are two ways this story ends.

[00:09:11]Scenario one, the bet pays off. In this scenario, AI infrastructure becomes the new internet. Just as the companies that built the internet's physical backbone in the 1990s captured enormous value over the next 30 years, the companies building AI infrastructure today capture enormous value over the next 30 years.

[00:09:32]Meta's advertising business becomes dramatically more efficient. AI agents handle billions of customer interactions, content recommendations, and the targeting decisions with a fraction of the human cost. Lama becomes the foundation of the next generation of software, locking in Meta as a platform dependency for millions of developers.

[00:09:53]Revenue per employee, already rising, skyrockets.

[00:09:58]The stock, already near all-time highs after a 10% jump following Q4 earnings, continues to climb. Bank of America's price target of $885 per share looks conservative in retrospect. The 8,000 people who lost their jobs in May 2026 were collateral damage in the creation of one of the most valuable companies in history.

[00:10:20]Scenario two, the bet fails.

[00:10:24]This scenario is less discussed, but it's worth understanding. $125 billion is Meta's infrastructure budget this year. Their entire payroll is $27 billion. If AI does not generate the returns Zuckerberg expects, if the advertising efficiency gains don't materialize, if Lama doesn't achieve the developer adoption they're counting on, if a competitor builds better infrastructure faster, Meta has committed a sum they cannot walk back. You can't unbuild a data center. You can't return GPUs.

[00:10:58]The CFO, Susan Li, has already warned of what she called significant acceleration in infrastructure expense growth, meaning depreciation costs and operating expenses from these new facilities will hit the income statement heavily. That hits earnings. That hits the stock. And here's the detail that almost nobody is reporting on. Meta has a preferred stock instrument called STRC, trading under the ticker stretch, that carries an 11.5% annual dividend. That's an expensive financial instrument. And to fund the dividends on it, Meta is now considering selling Bitcoin from its treasury. The company that built a $125 billion AI infrastructure bet may need to liquidate crypto assets to pay dividends on the financial instruments it used to fund that bet. If that sentence sounds complicated, good. It should.

[00:11:52]Because this is how corporate financial engineering works in 2026.

[00:11:57]And most people have no idea it's happening. The point is, the downside of this bet is real and large. If AI doesn't generate the productivity gains these companies are projecting, the write-downs and restructuring costs could be enormous. Not terminal for a company like Meta, but painful.

[00:12:17]For the broader economy, the ripple effects of 95,000 tech jobs cut, with more coming, would hit consumer spending, housing markets, and local economies in tech hubs from San Francisco to Seattle to Austin. So, where does this leave us? If you're an investor looking at this story, and you should be, regardless of whether you own tech stocks directly, here are the things that matter.

[00:12:41]First, the infrastructure beneficiaries.

[00:12:44]The companies building the data centers aren't just the tech giants themselves.

[00:12:49]They're the companies supplying what the tech giants need. Nvidia is the most obvious. Their GPUs are the critical hardware in every AI data center. But there's also the energy sector. A gigawatt-scale data center needs reliable, massive power. Utilities, natural gas suppliers, and nuclear energy companies are all seeing increased demand driven directly by AI infrastructure spending. Second, the enterprise AI opportunity.

[00:13:19]As Meta and its competitors build out AI infrastructure, they're also making it cheaper and more accessible for other businesses to use AI.

[00:13:27]Enterprise software companies, the ones selling AI tools to businesses, stand to benefit as the underlying infrastructure becomes more commoditized. Third, the risk of concentration. $725 billion going to four companies raises a question that regulators in the US and Europe are starting to ask. What happens when the infrastructure of the future is controlled by four private entities? The same antitrust concerns that broke up AT&T in the 1980s are being applied, slowly, to the cloud and AI infrastructure market today. This isn't just a philosophical question. It has regulatory and investment implications.

[00:14:08]And finally, for everyone watching this who isn't an investor, the displacement of 95,000 tech workers is the visible part of something much larger. AI is not replacing jobs by doing them better.

[00:14:22]It's replacing jobs by doing them differently, by restructuring entire business models so that fewer humans are needed at all. The workers being cut at Meta, at Snap, at Cloudflare, at Coinbase, these are not low-skill workers. These are engineers, designers, product managers, people with advanced degrees and years of specialized experience. That this is happening to them tells you something about the scope of what's coming. The $725 billion bet is the defining corporate story of this decade. Whether it succeeds or fails, and we won't know for years, the decisions being made right now are locking in the structure of the economy for a generation. If this kind of analysis is useful to you, subscribe. We cover these stories every week with the same level of depth. And drop a comment below.

[00:15:12]Do you think this bet pays off? Or are we watching the most expensive mistake in corporate history? I'll see you in the next one.