By Meta, Amazon, Microsoft, and Alphabet will collectively spend $725 billion on AI infrastructure in 2026—up 77% from last year’s record $410 billion. That’s not the surprising part. What’s jarring is they’re cutting roughly 50,000 jobs while doing it. Meta executes 8,000 layoffs on May 20, Microsoft offered buyouts to 8,750 employees last month, and Amazon has eliminated 30,000 corporate roles since October. Big Tech just made its calculation: GPUs are a better investment than engineers.
If you’re a developer working at or aspiring to work at one of these companies, this is your new reality. Your salary is being weighed against the cost of an H100 cluster, and the cluster is winning.
The Numbers: Company by Company
Each of the Big Four is following the same playbook—massive capex increases paired with workforce reductions. Meta increased AI spending to $115-135 billion (up from $72.2B in 2025) while cutting 10% of its workforce and closing 6,000 open roles. Moreover, teams are being reorganized into AI-focused “pods” with new role categories like “AI builder” and “AI pod lead.” If your title doesn’t have “AI” in it, you’re in the danger zone.
Amazon’s spending up to $185 billion on AI infrastructure while CEO Andy Jassy insists the layoffs are “not really financially driven and it’s not even really AI-driven.” That’s a fascinating claim when you’re announcing $200 billion in AI spending while cutting 30,000 jobs. Consequently, actions speak louder than earnings call scripts.
Microsoft is spending $190 billion—with $25 billion attributed solely to rising memory and chip costs. They’re offering voluntary buyouts to 8,750 employees under the “Rule of 70” (age plus years of service). Furthermore, it’s the company’s first voluntary retirement program in 51 years. The buyout costs $900 million but saves $1.3-1.75 billion annually. The math is blunt.
Google is spending $180-190 billion while Sundar Pichai pursues a goal of making the company “20% more efficient.” Despite Q1 revenue hitting $109.9 billion with 22% YoY growth, hundreds of roles are being cut in Cloud, platforms, and devices. Profitable companies don’t cut jobs for profit—they cut them for priorities.
Where the $725B Goes (And What It Replaces)
The money is flowing to data centers, GPUs, memory chips, and power infrastructure. Meta’s building a $27 billion joint venture with Nebius for a gigawatt-scale AI facility in Louisiana. Microsoft is spending $25 billion on memory chips alone because High Bandwidth Memory is in short supply. Additionally, Nvidia H100 GPUs cost $25,000-40,000 each, and training a GPT-4-scale model requires tens of thousands running for months. AI data centers require 10-100x more power per rack than traditional servers.
Here’s the calculation that matters: 50,000 workers at an average Big Tech salary of $175K costs $8.75 billion annually. The $725 billion in capex represents 83 years of those workers’ salaries. In other words, these companies are betting that AI infrastructure will deliver more than 83 years worth of human productivity. That’s not efficiency—that’s a hypothesis.
The Justifications Don’t Add Up
Company messaging contradicts observable reality. Meta’s HR memo said the layoffs help “run the company more efficiently and offset the other investments we’re making.” Translation: we’re cutting you to pay for GPUs. Similarly, Amazon’s Jassy claims cuts aren’t “AI-driven” while spending $200 billion on AI. Microsoft’s CFO says “AI tools can augment productivity” as a justification for reducing headcount by 8,750 people. Meanwhile, Google’s Pichai insists AI will “create demand for more employees” while cutting hundreds.
Meanwhile, Big Tech’s cash flow is at its lowest level in over a decade despite record revenue. Spending is outpacing revenue growth, and investors are starting to demand proof of ROI. However, no company has demonstrated that AI productivity gains justify the human cost. This is speculation dressed as strategy.
What This Means for Developers
The skills that survive this transformation aren’t the ones being cut. Traditional software engineering roles—generic CRUD apps, frontend development, QA testing—are being automated or eliminated. In particular, mid-level generalist positions are especially vulnerable because AI can handle routine tasks. What remains valuable: AI/ML engineering, infrastructure and DevOps, distributed systems architecture, security, and senior strategic roles that require judgment AI can’t replicate.
Developer communities on Reddit, Hacker News, and Blind are blunt about the situation. “If Meta can cut 10% after record revenue, no one is safe,” one engineer wrote. Another: “They’re spending my salary on GPUs that will replace me.” As a result, the sentiment is clear—Big Tech’s golden age as a stable career path is over. AI engineer job postings are up 300% year-over-year while traditional SWE roles are down 20%.
Developers are responding practically: upskilling in AI/ML and RAG systems, pivoting to smaller companies with less AI focus, pursuing geographic arbitrage through remote work, or trying entrepreneurship. The calculation is simple—if layoffs are inevitable, taking control beats waiting for the axe.
The Unanswered Question
Will $725 billion in AI infrastructure generate more value than 50,000 skilled employees would have created? Nobody knows. The dot-com bubble saw massive infrastructure spending on fiber optic cables based on “build it and they will come” thinking. Consequently, the cables sat unused after the crash. The mobile revolution also required companies to rebuild everything, but mobile had clear consumer demand from day one. In contrast, AI’s value proposition is still theoretical for most use cases.
Watch these signals: if cash flow stays negative through 2026, if Copilot and other AI products fail to gain adoption, if Q3-Q4 layoffs accelerate, or if data center projects get delayed. Those would indicate the bet isn’t paying off. Bull case analysts say AI transforms productivity and justifies the spend. On the other hand, bear case analysts call it “garbage thesis” destined to pop. The realistic case is messier—some companies will win, others will stumble, and developers will bear the adjustment costs either way.
Big Tech chose $725 billion in infrastructure over 50,000 people. That’s not transformation—that’s a gamble on a future that hasn’t arrived yet. Therefore, developers watching this unfold should plan accordingly. The industry isn’t eliminating software engineers, but it’s definitely repricing them against the cost of compute. Make sure you’re on the right side of that equation.
