You are being shown one number. Mark Zuckerberg told 8,000 Meta employees in early May that their jobs were a casualty of the company’s AI infrastructure budget. Amazon has cut roughly 30,000 roles in the last five months. Microsoft has shed about 125,000 through “voluntary” departures. Alphabet is in the middle of an ongoing 1,500.
The headlines compress into a single phrase: AI is taking jobs.
That phrase is not wrong. It is just the wrong number.
$145 billion vs $27 billion: the math behind the memo
Meta’s projected 2026 capital expenditure runs $125 to $145 billion. The company’s total human compensation bill, every salary, every benefit, every stock grant, comes to roughly $27 billion. If Meta fired every single one of its employees tomorrow, it would save $27 billion against a $145 billion infrastructure check. The AI capex line is four to five times the entire payroll line. Layoffs are not the cost-cutting story here. Layoffs are the financing.
Zuckerberg told employees that the May cuts are a direct consequence of the AI infrastructure budget. He was not speaking in metaphor. He meant the budget literally, the line item, the cash. He was not saying an algorithm had replaced the work. He was saying the company chose to buy GPUs instead.
The four hyperscalers, the $725 billion bill
$725 billion. That is the upper bound of what four companies, Amazon (NASDAQ:AMZN | AMZN Price Prediction), Microsoft (NASDAQ:MSFT), Alphabet (NASDAQ:GOOGL), and Meta Platforms (NASDAQ:META), plan to spend on capital projects in 2026. It represents a 77% increase year over year and it is going almost entirely into data centers, custom chips, GPUs, robotics, and AI models. The destination is compute, not payroll, stores, or stock buybacks.
The figure comes from each company’s own 2026 guidance: roughly $200 billion at Amazon, $180 billion to $190 billion at Alphabet, $125 billion to $145 billion at Meta, and a Microsoft figure that, while not formally guided as a full-year dollar number, is running at $30.88 billion in a single quarter with management telling investors capex will keep climbing.
What It Means
The reallocation story is the story. April 2026 alone saw 83,387 announced job cuts, up 38% from March’s 60,620, with AI cited as the primary reason for 21,490 of those cuts. Year to date, the tech sector has shed 85,411 jobs, up 33% from the same point in 2025, the worst pace since 2023. The same companies running those reductions are the ones writing the biggest infrastructure checks in corporate history.
Walk through it:
- Amazon: about 30,000 employee reductions over the last five months, while Q1 2026 capex came in at $44.2 billion, up 77% year over year. AWS grew 28%, the fastest pace in 15 quarters, and the chips business is now running at over $20 billion in annual revenue.
- Microsoft: around 125,000 voluntary departures, paired with commercial RPO of $392 billion, up 51% year over year, and a stated plan to increase total AI capacity by over 80% this year and roughly double the data center footprint over the next two years.
- Alphabet: roughly 1,500 ongoing reductions, against a Google Cloud backlog of $462 billion, nearly doubled sequentially, and a Q1 2026 capex print of $36 billion, up 107% year over year.
- Meta: 8,000 layoffs planned for May, with 2026 capex guidance raised to $125 billion to $145 billion.
Meta is the cleanest case study. Its AI capex is roughly four to five times what the company spends on total human employee compensation. Even fully replacing the workforce with AI would only save about $27 billion, a fraction of the $145 billion infrastructure budget. The layoffs are a side effect of where the capital is going, not the savings story.
Market Reaction
Long-term holders have been paid to sit through this. Year to date through May 7, 2026, Amazon shares are up 17%, from $230.82 to $271.17. Alphabet is up 27%, from $312.79 to $397.99. Meta is down 6%, from $659.54 to $616.81. Microsoft is down 13%, from $482.52 to $420.77. The market is rewarding the names where backlog and AI revenue are visible (Alphabet, Amazon) and punishing names where the capex line is rising faster than investors can model returns (Microsoft, Meta).
Bull Case
Three things make this look like capital efficiency, not a bubble.
First, the backlogs are real and growing. Amazon’s AWS backlog stands at $364 billion, plus a separate $100 billion-plus Anthropic deal not included in that figure. Microsoft’s commercial RPO is $392 billion, with weighted average duration of about two years. Alphabet’s cloud backlog of $462 billion is expected to convert to revenue at just over 50% over the next 24 months. These are signed customer commitments, not aspirational forecasts.
Second, custom silicon is changing the unit economics. Amazon’s Trainium business is now at a $20 billion-plus annual run rate, with a $225 billion revenue commitment book and customers including Anthropic, OpenAI, Uber, and Meta. CEO Andy Jassy told investors that “at scale, we expect Trainium will save us tens of billions of dollars of CapEx each year and provide several hundred basis points of operating margin advantage versus relying on other chips for inference.”
Third, the cash machines underwriting all of this are still cash machines. Amazon’s Q1 operating margin hit 13.1%, the highest ever. Alphabet’s Q1 operating margin was 36.1%. Microsoft’s Q1 operating margin came in at 49%, ahead of expectations. Meta’s was 41.44%. These four are spending hundreds of billions while still earning more per dollar of revenue than most of the S&P 500.
The headcount cuts are a form of operating leverage on top of that. Jassy has described five engineers using agentic coding tools to rebuild a service in 65 days that would have previously taken 40 to 50 people about a year. CFO Susan Li said Meta is targeting a “leaner operating model” precisely to “help to offset the substantial investments we are making.” Capital is replacing payroll, then producing more output per dollar than the payroll did.
Bottom Line
For retirement-focused investors, the layoff headlines are unsettling.
The real question is whether $725 billion of concentrated infrastructure spending earns its cost of capital. The early evidence (Amazon’s $364 billion AWS backlog, Alphabet’s $462 billion cloud backlog, Microsoft’s $392 billion in remaining performance obligations, Meta’s 33% revenue growth) suggests the contracts to monetize this build already exist.
The next checkpoint is the Q2 2026 earnings report, where investors will see whether capex growth is still meaningfully outpacing revenue growth or whether the curves are starting to converge. Polymarket traders, for what it is worth, are already pricing an 89% probability that Amazon’s 2026 capex exceeds $170 billion and a 54.5% probability it tops the $200 billion guide. The capital is moving.
The only real debate is how fast it compounds.